from literature to biterature

This page intentionally left blank

 preface iii

FROM

L I T E R AT U R E
TO

B I T E R AT U R E
Lem, Turing, Darwin, and Explorations in Computer Literature,

Philosophy of Mind, and Cultural Evolution

PETER SWIRSKI

McGill-Queen’s University Press

Montreal & Kingston • London • Ithaca
iv preface

© McGill-Queen’s University Press 2013

isbn 978-0-7735-4295-2 (cloth)

isbn 978-0-7735-8991-9 (epdf)
isbn 978-0-7735-8992-6 (epub)

Legal deposit third quarter 2013

Bibliothèque nationale du Québec

Printed in Canada on acid-free paper that is 100% ancient forest

free (100% post-consumer recycled), processed chlorine free

This book has been published with the help of a grant from the
Canadian Federation for the Humanities and Social Sciences,
through the Awards to Scholarly Publications Program, using funds
provided by the Social Sciences and Humanities Research Council
of Canada.

McGill-Queen’s University Press acknowledges the support of the

Canada Council for the Arts for our publishing program. We also
acknowledge the financial support of the Government of Canada
through the Canada Book Fund for our publishing activities.

Library and Archives Canada Cataloguing in Publication

Swirski, Peter, 1963–, author

From literature to biterature : Lem, Turing, Darwin, and explo-
rations in computer literature, philosophy of mind, and cultural
evolution / Peter Swirski.

Includes bibliographical references and index.

Issued in print and electronic formats.
isbn 978-0-7735-4295-2 (bound). – isbn 978-0-7735-8991-9 (epdf). –
isbn 978-0-7735-8992-6 (epub)

1. Computers and civilization. 2. Artificial intelligence. 3. Social

evolution. 4. Literature – Philosophy. 5. Philosophy of mind.
6. Language and languages – Philosophy. I. Title.

qa76.9.c66s94 2013 303.48'34 c2013-902978-8

c2013-902979-6

This book was typeset by True to Type in 10.5/13 Baskerville

1
 Foreword v

Dedicated to Jaakko Hintikka, Sami Pihlström,

David Livingstone Smith, and other bibliosophers
of today and tomorrow
This page intentionally left blank
 preface vii

Contents

Lemmata
Narrative Intelligence; Test of Time; Avocados or Wombats; Institute
for Advanced Studies; Builtware and Bitware; Paignton Zoo 3

part one
1 Biterature
Toscanelli’s Map; Bloomsday; Titan; Manchester Mark; Manglers
and Mashers; Suits for the Dawn Patrol; Charabia; Short for
Raconteur; I Gave Myself an Electric Current; Abe’s Saloon!;
Humming (to Itself?) 17
2 Computhors
Poetic Licence; Devotchka; Analytical Engine; EZ-Writer; Möbius-Like;
JFK; One-Armed Bandit; Homeostat Sapiens; As It Sees
Fit; Brainstorm; A1, B2, C3; Black Boxes 35
3 Bibliosophy
La Chanson de Roland; The Motherboard of All Problems; Little
Endian; Algo Trading; Father Roberto Busa; Riverside Shakespeare;
Death in Holbein; Zettabyte; A Hamlet Every Eight Seconds; A Pill
That Makes Me Feel Loved 53
4 Transhumana
Non Serviam; Collective Loss of Morale; Slippery Slope; Artificial Stu-
pidity; Prometheus and Dolus; L’Homme Machine; DARPA; Nobel Prize
in Biterature; Chatbots in Chatter 71
 viii Contents

part two
5 tt
Supercalifragilistic; Adroit Android; Imitation Game; Parry; Exit
Human, Enter Machine; I, Object; I Always Avoid a Kangaroo; Reverse
Turing Test; Ultimatest; Aramis 89
6 qed
Troubles with Functionalism; Print This!; Trompe l’Oeil; Like the
Pope; A Bevy of Pigeons; No More, No Less; Dicty; Operational
Definition?; Flugblogs; On Being Mental 111
7 ToM
Kismet; PaPeRo; Guilty as Charged; Eureka; Meaning; Elementary,
Dear Wason; Biology in Action; Driving Range; Insula Cortex;
On Being Redundant 130
8 Y2K+
Methuselah; Yes!; An Amphibian with a Stick; Ellipses in This
Picture; Software for the Soul; Parallel Paradigm; Erlang and
Mira; DelFlies; Eruntics; Cogito Paradox 149

part three
9 How to Make War and Assassinate People
Ice-Pick and Ricin; A Small African Country; Licence to Kill;
Solsects, Sunsects, Synsects; Fires at the Sun; Candy, Anyone?;
Bacteriology; Safe Investment 171
10 Biologic, or the Philosophy of Error
Axiomatic Givens; Nature’s Greatest Error; Wandering Jew;
Critique of Pure Reason; Manna from Heaven; What Is;
Rosencrantz and Guildenstern 183
11 Mind Your Business
100,000 Years; You’re Flying Low, Mac; Neuromancer; See
Further; The Making of a Fly 197

Notes 205
Bibliography 209
Index 225
 The Concept of Culture

from literature to biterature

Since most of the statements in this book are speculations, it would have
been too tedious to mention this on every page. Instead I did the oppo-
site – by taking out all words like “possibly” and deleting every reference
to scientific evidence. Accordingly, this book should be read less as a text
of scientific scholarship and more as an adventure story for the imagina-
tion. Each idea should be seen not as a firm hypothesis about the mind,
but as another implement to keep inside one’s toolbox for making
theories of the mind.
Marvin Minsky
This page intentionally left blank
 1

Lemmata

Narrative Intelligence; Test of Time; Avocados or Wombats;

Institute for Advanced Studies; Builtware and Bitware;
Paignton Zoo

narrative intelligence

The first general-purpose – Turing-complete, in geekspeak – elec-

tronic brain was a behemoth of thirty-plus tons, roughly the same
as an eighteen-wheeler truck. With twenty thousand vacuum tubes
in its belly, it occupied a room the size of a college gym and con-
sumed two hundred kilowatts, or about half the power of a road-
worthy semi. Turn on the ignition, gun up the digital rig, and off
you go, roaring and belching smoke on the information highway.
Running, oddly, on the decimal rather than the binary number
system, the world’s first Electronic Integrator and Computer also
boasted a radical new feature: it was reprogrammable. It could, in
other words, execute a variety of tasks by means of what we would
call different software (in reality, its instructions were stored on
kludgy manual plug-and-socket boards). Soldered together in 1946
by John Mauchly and J. Presper Eckert at the University of Penn-
sylvania, the eniac was a dream come true.
It was also obsolete before it was completed. The computer rev-
olution had begun.
The rest is history as we know it. In less than a single lifetime,
ever more powerful computing machines have muscled in on
almost all facets of our lives, opening new vistas for operations and
research on a daily basis. As I type this sentence, there are more
than seven billion people in the world and more than two billion
4 From Literature to Biterature

Soldered together in 1946 by John Mauchly and J. Presper Eckert at the

University of Pennsylvania, the eniac was a dream come true.

computers – including the one on which I have just typed this sen-
tence. And, by dint of typing it, I have done my bit to make the
word “computer” come up in written English more frequently than
99 per cent of all the nouns in the language.
In a blink of an eye, computers have become an industry, not
merely in terms of their manufacture and design but in terms of
analysis of their present and future potential. The key factor behind
this insatiable interest in these icons of our civilization is their
cross-disciplinary utility. The computer and the cognitive sciences
bestraddle an ever-expanding miscellany of disciplines with fingers
in everything from alphanumerical regex to zettascale linguistics.
Chief among them are artificial intelligence, artificial emotion,
artificial life, machine learning, knowledge engineering, software
engineering, robotics, electronics, vision, computability, informa-
tion science – all with their myriad subfields. Towards the periphery,
they snake in and out of genetic algorithms, Boolean logic, neurol-
ogy and neuropathology, natural language processing, cognitive
and evolutionary psychology, decision and game theory, linguistics,
philosophy of mind, translation theory, and their myriad subfields.
 Lemmata 5

So swift has been the expansion of computers’ algorithmic and

robotic capacities that quite a few of the assumptions forged in
their formative decades no longer suffice to grasp their present-day
potential, to say nothing about the future. This has never been
truer than in our millennium in which research subdisciplines
such as Narrative Intelligence are finally, if slowly, beginning to
crawl out from the shadow of more established domains of Artifi-
cial Intelligence.
Underwriting this new field is mounting evidence from the bio-
logical and social sciences that a whole lot of cognitive processing
is embedded in our natural skill for storytelling. As documented by
psychologists, sociobiologists, and even literary scholars who have
placed Homo narrativus under the microscope, we absorb, organize,
and process information better when it is cast in the form of a story.
We remember and retrieve causally framed narratives much better
than atomic bits of ram.
The power of the narrative is even more apparent in our striking
bias toward contextual framing at the expense of the underlying
logic of a situation. People given fifty dollars experience a sense of
gain or loss – and change their behaviour accordingly – depending
on whether they get to keep twenty or must surrender thirty. Equal-
ly, we fall prey to cognitive illusions when processing frequencies as
probabilities rather than natural frequencies. A killer disease that
wipes out 1,280 people out of 10,000 looms worse than one that kills
24.14 per cent, even though bug number 2 is actually twice as lethal.
Given such deep-seated cognitive lapses, the idea of grafting an
artsy-fartsy domain such as storytelling onto digital computing may
at first appear to be iffy, if not completely stillborn. Appearances,
however, can be deceiving. The marriage of the computer sciences
and the humanities is only the next logical step in the paradigm
shift that is inclining contemporary thinkers who think about
thinking to think more and more in terms of narratives rather than
logic-gates. The narrative perspective on the ghost in the machine,
it turns out, is not a speculative luxury but a pressing necessity.

test of time

This is where From Literature to Biterature comes in. Underlying my

explorations is the premise that, at a certain point in the already
foreseeable future, computers will be able to create works of liter-
6 From Literature to Biterature

ature in and of themselves. What conditions would have to obtain

for machines to become capable of creative writing? What would
be the literary, cultural, and social consequences of these singular
capacities? What role would evolution play in this and related sce-
narios? These are some of the central questions that preoccupy me
in the chapters to come.
Fortunately, even if the job is enormous, I am not starting from
zero. In Of Literature and Knowledge (2007) I devoted an entire book
to the subject of narrative intelligence – human narrative intelli-
gence. Evolution has bequeathed our species with a cunning array
of cognitive skills to navigate the currents of life. Human intelli-
gence is uniquely adapted to thought-experiment about the future
and to data-mine the past – in short, to interpret and reinterpret
context-sensitive information.
But what about machine intelligence? Is it a sui generis different
kind of animal or just an offshoot of the main branch? And how
does one calibrate intelligence in computers or even determine
that it is there at all? Most of all, isn’t machine intelligence, like mil-
itary intelligence, an archetypal oxymoron? Jokes aside, it is not,
because human and machine evolutions are ultimately two phases
of the same process.
Bioevolution may have spawned technoevolution, but the accel-
erating pace of scientific discovery makes it all but certain that
humanity is destined to take evolution into its own hands. From
that point on, the difference between what is natural and artificial
will begin to dissolve. Bioevolution will become a subfield of tech-
noevolution inasmuch as the latter will facilitate steered – i.e., tele-
ological – autoevolution. The evolutionary equilibrium of our species
will have been punctuated in a most dramatic way. Biology will have
become a function of biotech.
On the literary front, computers capable of writing fiction bring
us face to face with the gnarly problems of authorship and inten-
tionality. Both lie at the heart of my Literature, Analytically Speaking
(2010), which brings order to the riot of positions and presup-
positions around the matter of human art. This time around, my
targets are authorial intention and extension as they pertain
to machines, albeit machines of a very different order from
the mindless text-compilers of today: creative, artful, and causally
independent.
I made a first pass at unlocking the riddles of computer author-
 Lemmata 7

ship (or computhorship) and bitic literature (or biterature) in

Between Literature and Science (2000). Much of what I wrote there
has weathered the test of time better than I could have imagined.
To take only one example, my analysis of the pragmatic factors in
the administration of the Turing test, such as the role of perfect
and complete information available to the judges, was indepen-
dently developed in the same year by the cognitive scientist Ayse
Saygin.
The success of these earlier studies gives me the confidence to
trespass onto the territory staked out by computer scientists and
roboticists, even if it is in pursuit of inquiries that might be, on bal-
ance, of more interest to literary scholars and philosophers. Seen
in this light, From Literature to Biterature is not a book of science,
even though some humanists may find it overly technical for their
taste. Nor is it a book of literary theory, even as it liberally trucks in
narrative cases and critical lore. It is rather an old-fashioned book
of discovery or, if you prefer, a modern adventure story of the
mind.

avocados or wombats

While books on digital literature are no longer as hard to find as they

were in the days of Espen Aarseth’s Cybertext (1997), literary-critical
voyages to the digital shores of tomorrow are as still rare as a steak
that runs on four legs. This includes the self-styled exception to the
rule, a cheerleading overview of computer-assisted writing, Kather-
ine Hayles’s Electronic Literature (2008). Limited by and large to a sur-
vey of second-order mashing protocols of yesterday, it stands in
sharp contrast to my anatomy of electronic literature of tomorrow.
The aim of my explorations, the first of their kind, is simple. I
want to draw a synoptic map of the future in which the gulf con-
cealed by the term “bitic literature” – that is, the gulf between mere
syntax-driven processing and semantically rich understanding –
has been bridged. As it happens, the gulf in question is greater
than that which separates tree-climbing pygmy marmosets from
the Mbenga and Mbuti human pygmy tribes of Central Africa. As
such, the ramifications of this single lemma could hardly be more
far-reaching.
Most conspicuously, acknowledging biterature as a species of lit-
erature entails adopting the same range of attitudes to computhors
8 From Literature to Biterature

as to human writers. Specifically, it entails approaching them as

agents with internal states, such as, for instance, creative inten-
tions. Here, however, we find ourselves on shaky ground. Comput-
ers with creative intentions are perforce computers that think, but
what is the status of such thinking? Are machine thoughts and
intentions real, like avocados or wombats, or are they merely use-
ful theoretical fictions, like centres of gravity or actuarial averages?
These questions breed other questions like neutrons in a run-
away chain reaction. Is thinking in computers different from that
in humans? What would it mean if it turned out to be radically dif-
ferent or, for that matter, not different at all? Can machines really
think, or is thinking merely what we could attribute to advanced
future systems? And what does really really mean in this context,
given that thinking about thinking machines segues seamlessly into
thinking about machines with personality, identity, karma, and
God knows what else?

institute for advanced studies

As I work my way through this thicket of questions, I hitch a ride on

the shoulders of three thinkers who have, in their respective ways,
mapped out the contours of this terra incognita. First on the list is
Stanislaw Lem. One of the greatest novelists of the twentieth centu-
ry, he was also one of the most far-sighted philosophers of tomor-
row, lionized by Mr Futurology himself, Alvin Toffler (of Future Shock
fame). Not coincidentally, Lem is also the author of the only exist-
ing map of the literary future, patchy and dated though it may seem
today.
But even as From Literature to Biterature is a book about literature,
it is not a book of literary exegesis. The difference is fundamental.
Throughout Lem’s career, reviewers marvelled at the complexity
and scientific erudition of his scenarios, eulogizing him as a liter-
ary Einstein and a science-fiction Bach. His novels and thought
experiments, exclaimed one critic, read “as if they were developed
at Princeton’s Institute for Advanced Studies.”1
Maybe, maybe not. For how would you know? Literary critics
hardly flock to Princeton to get a degree in the computer or evo-
lutionary sciences in order to find out if Lem’s futuristic scenarios
are right or, indeed, wrong. Not a few contend, in fact, that this is
 Lemmata 9

Most conspicuously, acknowledging biterature as a species of literature

entails adopting the same range of attitudes to computhors as to human
writers. Specifically, it entails approaching them as agents with internal
states, such as, for instance, creative intentions.

the way it ought to be, inasmuch as literary studies are not even
congruent with the cognitive orientation of the sciences – cogni-
tive, biological, or any other. If so, I purposely step outside the con-
fines of literary criticism by affording Lem the conceptual scrutiny
he deserves, in the process tweaking his hypotheses or rewriting
them altogether.
If Alan Turing’s name is synonymous with machine intelligence,
it is as much for his contributions to cognitive science as for his
unshakable belief in intelligent machines – contra John von Neu-
mann, who was as unshakably certain they would never think. From
Turing machines to the Turing Award (the most prestigious prize
for computer scientists), down to the Turing Police, who lock
horns with a supercomputer that plots to boost its intelligence in
William Gibson’s modern sci-fi classic Neuromancer – his name pops
up everywhere.
10 From Literature to Biterature

Celebrated as one of Time’s “100 Persons of the [Twentieth] Cen-

tury,” to this day Turing is regarded as a prophet of machine intel-
ligence. More than sixty years after it was unveiled, his test remains
the single most consequential contribution to our thinking about
thinking computers. Appropriately, the 2012 celebrations of the
Turing Year even comprised a special Turing test conducted at
Bletchley Park outside London, where the British mathematician
played such a decisive role in decoding German military ciphers
during World War II.
In many ways, the Turing test forms the backbone of the book in
your hands. Although the protocol itself and all the Chinese Room-
type critiques of it are by now nothing new, I squeeze a fair amount
of juice out of the old orange by going at it from a novel angle,
while adding a bunch of new twists to old themes. Indeed, as I
argue throughout, the test itself is a categorically different tool –
inductive rather than analytic – from what almost every analyst
claims it to be.
As for Charles Darwin, I assume you don’t need to be told who he
was and why he matters. Let me just say that, even though From Lit-
erature to Biterature is not a book of fiction, it fashions a hero of epic
proportions: evolution. No one can tell what role evolution will play
in future computer design, but my bet is that it will be substantial.
My other bet is that natural selection itself may be due for an over-
haul as laws of self-organization and autocatalysis supplement what
we know of this fundamental force that orders our living universe.
Chiefly, although not exclusively, the last part of the book thus
spins a series of scenarios that leave the present-day world far
behind. My basic premise is that computer evolution will by its very
nature exceed the pace of natural evolution a thousand- or even a
million-fold. Being, at the same time, directed by the machines
themselves, it will leave its imprint on every page of the book of life
as we know it. Whether these scenarios qualify as literary futurology,
evolutionary speculation, or philosophy of the future I leave for you
to judge.

builtware and bitware

Before we hoist anchor, a few words on a few key terms and mental
shortcuts necessary to navigate the book in your hands. Let me
 Lemmata 11

start with those that crop up on almost every page: thinking, intelli-
gence, understanding, and to a lesser extent consciousness. Given that
each is loaded with connotations – especially in philosophy and
psychology, where they are sometimes glossed as having semantics or
having intentionality – right off the bat I disclaim any meanings that
go outside their common usage.
Taking all these descriptors to denote simply the cognitive abili-
ties of a normal human being, I also take them to be synonymous
and interchangeable: if you have one, you have the others. Just like
rationality, intelligence has long been seen as strung evenly on a
single axis, with geniuses at one end and blondes at the other. Not
so – at least not since 1983, when psychologist Howard Gardner
distinguished no less than seven types of intelligence: linguistic,
logical-mathematical, bodily-kinaesthetic, social-interpersonal,
introspective-intrapersonal, musical, and spatial.
Having said that, I take reasoned response to change to be a nec-
essary, though not necessarily sufficient, condition for intelligence.
Not coincidentally, this prima facie ai characterization can be eas-
ily reinterpreted in biological terms. This is because reasoning
(memory, analysis, and prediction), fortified by feedback and per-
formance evaluation, belongs to a suite of cognitive adaptations to
sort out all manner of environmental stimuli in the service of
homeostasis and self-preservation.
Finally, a few words on the star player: the computer. Speaking
of computer performance is always a matter of metaphor and
metonymy or, if you prefer, colloquial convenience. This is because
the actual physical aggregate of circuits called hardware is not, in
and of itself, capable of anything. The same is true of software.
Long or short, strings of binary digits – or bits, as they have been
called since Claude Shannon’s 1948 breakthrough paper on the
mathematical theory of communication – that form a computer’s
chain of instructions have, in and of themselves, no causal powers.
Sitting on a shelf, whether in the form of magnetic pulses or sym-
bols on paper, a “disembodied” program is incapable of executing
anything at all, be it a series of legal manoeuvres to probate an
estate or a timed release of a lethal cocktail into the vein of a death-
row inmate. To have any computing or causal powers, both the bit-
ware and the builtware must act as one. Decoupled, neither is capa-
ble of affecting its internal or external environment.
12 From Literature to Biterature

paignton zoo

Talking about implementation or even evolution of programs in

computers presupposes a strict division of hardware and software.
Replace the processors with other processors, for example, and the
software should remain intact. Not so in biological evolution,
where brain software and hardware evolved in unison. Prefrontal
cortex, to take one example, is responsible for growth in abstract
thinking – implicated in tool-making, culture, social practices, and
so on – and is at the same time a co-evolutionary outgrowth of our
capabilities for language, art, and awareness of mortality.
If in what follows I stick, therefore, to the term computer, it is only
as a handy shorthand for the embodied software of a fully inte-
grated program-processor-effector. Naturally, future science may
yet erase the entire software-hardware divide, bringing computer
manufacture and design conspicuously in line with human manu-
facture and design. In a homology to human reproductive cells,
which in anthropogenesis combine the functions of morphological
building blocks and ontogenetic blueprints, our bitware and built-
ware may one day also collapse into one.
Unfortunately, it is not uncommon for such futuristic specula-
tion to drive our collective head only deeper into the sand. For the
societal ostrich, visions of autonomous autotelic automata are hard
to square with the kludgy clunkers of today which make even
Paignton Zoo macaques look like a troop of Dostoyevskys. Present-
ed in 2003 with a keyboard connected to a pc, after a month of
labours the six monkeys managed to bang out five pages of the let-
ter “S” and break the keyboard.
Nothing to worry about, in other words, since the future will look
a lot like today, only a little more so. Take Poe’s “Mellonta Tauta”
at face value, and in the year 2048 we will race around in giant bal-
loons at neck-breaking speeds of 150 miles an hour. The flip side
of this sort of escapism, on the other hand, is not realism but fan-
tasy in which the future extrapolates itself into Asimov’s asinine
three “laws” of robotics or a doomsday script out of Schwarzeneg-
ger’s Terminator 3: Rise of the Machines.
In the place of bad science fiction, however, we need good sci-
ence and good fiction aimed at the socio-cultural, aesthetic, ethi-
cal, and even ontological implications of the ghost in the machine.
As one of Shakespeare’s memorable creations said on encounter-
 Lemmata 13

ing a ghost outside the machine, there are more things in heaven
and earth than are dreamt of in our philosophy. Truly, we may
need to rethink our philosophy, both natural and analytic, if the
computer is to help us create a better and not just a faster society.
Not for a moment do I mean to suggest that the scenarios I con-
jure up in the pages that follow are in any way inevitable or even
desirable. It is far from a foregone conclusion that my hypotheses
will morph from fiction to fact. I do mean, however, to dispute the
default perception that they can’t happen here. After all, we have
only begun to scratch the surface of tomorrow. But from where I’m
standing, the scratches are getting deeper by the minute, looking
more and more like gouges.
This page intentionally left blank
 The Concept of Culture

part one

Lem’s intuitive and literary approach perhaps does a better job of

convincing readers to his views than any hard-nosed scientific article or
arcanely reasoned philosophical paper might do.
Douglas R. Hofstadter and Daniel Dennnett
This page intentionally left blank
 1

Biterature

Toscanelli’s Map; Bloomsday; Titan; Manchester Mark;

Manglers and Mashers; Suits for the Dawn Patrol; Charabia;
Short for Raconteur; I Gave Myself an Electric Current; Abe’s
Saloon!; Humming (to Itself?)

toscanelli’s map

Before he ventured into the uncharted waters of the Atlantic,

Columbus armed himself with the famous Toscanelli map, which
sketched the route to the Spice Islands of Asia. A smart move,
except that, even as the parchment represented the distillate of the
best geographical intelligence of the time, it was in equal measure
a fanciful projection of the cartographer’s imagination. As a mat-
ter of fact, Toscanelli had badly miscalculated the size of the Earth,
which is why Columbus never realized that he had stumbled on an
unknown continent.
Even a fanciful map may be better than none, however, when voy-
aging to a place on the continuum where everything by default is a
projection of the cartographer’s imagination. Hence, as I venture
into the uncharted waters populated by such chimeras as biterary
computhors, I take advantage of a singular map of the future. First
published in 1973, Lem’s “A History of Bitic Literature” (released
in English as part of Imaginary Magnitude) is a singular work of fic-
tion by any standard. It takes the form of a metafictional preface
to a proleptic review of computer writing from the year 2009. His
future, our past.
Similarly to Turing’s legendary essay “On Computable Numbers,”
Lem’s opens with a definition. Biterature, he writes, is “any work of
18 From Literature to Biterature

nonhuman origin – one whose real author is not a human being.”1

The very next sentence, however, executes a dramatic U-turn. Now
Lem allows that humans could author biterature indirectly “by per-
forming the functions which generated the real author’s acts of cre-
ation.” For the record, he is not alone in hedging his bets in this
way. In his Pulitzer-winning meditations on the nature of computer
consciousness, Gödel, Escher, Bach (1979), Douglas Hofstadter also
distinguished the real author from meta-author.
Taken separately, the two concepts make good sense. On the one
hand there is the real author – the creator who brings a novel into
existence – and on the other there is the intellectual entourage:
people who for a variety of reasons have influenced his artistic
development and creative process. Fuse the two concepts together,
however, and they blow the concept of biterary authorship out of
the sky. This is so because Lem’s indirect author doubles back on
the real author, calling both concepts into question. Suddenly the
issue of authorship can no longer be resolved without bringing the
intellectual entourage into the picture.
From Lem’s theses it follows that any author, human or not, is
only the terminal stage in the genesis of a work. The crippling lack
of what programmers call a “stopping rule” is easier to grasp on
examples from the world of our flesh-and-blood creators. In Lem’s
view, writers are only conventional bookends for their teachers,
mentors, colleagues, writers, critics, and anybody else who might
have indirectly contributed to the genesis of their work. That’s
right: anybody who played any role in the nominal author’s cre-
ation becomes an indirect author of the resulting work.
This makes no sense. Take the droll scenario from Martin Ritt’s
1976 satirical comedy, The Front. In a rare acting-only appearance,
Woody Allen plays Howard Prince, an amiable schmo who during
the 1950s’ huac witch-hunts fronts for a group of scriptwriters
branded as pinkos. In return, Howard gets the publishing credits
and a cut of the royalties, and everybody is happy (until the schmo
begins to demand revisions to fit “his” style).
Is Howard an indirect author of the scripts that bear his name?
If you think that the answer is a no-brainer – he never pens a word
– remember that without him no word would get penned in the
first place. In his way, Howard plays an indispensable role in the
genesis of the scripts, exactly as prescribed by Lem’s protocol. That
role is not dissimilar to that of every commissioning editor who, by
 Biterature 19

Is Howard an indirect author of the scripts that bear his name? If you think
that the answer is a no-brainer – he never pens a word – remember that with-
out him no word would get penned in the first place.

dint of issuing a publishing contract, performs a key function in

the process of creation.
Thus, if we were to take Lem’s theses in good faith, the publish-
ing house of Little, Brown who commissioned Norman Mailer to
write a book-length account of Apollo 11’s epochal mission would
thus become an author, albeit an indirect one, of Mailer’s work.
Make room on the shelf for Of a Fire on the Moon by Mailer, Little,
and Brown, all because their contract played a generative function
in the history of the book. Without it, this fine piece of new jour-
nalism would presumably have never got written.

bloomsday

It should be evident by now why, by adding a rider to his first

clause, Lem generates more headaches than he cures. What is miss-
20 From Literature to Biterature

ing from his account of computer authorship is causal depth: the

actual and direct causal mechanism needed to save the entire con-
cept from being true but trivial or true but insufficient. In short, we
need some kind of rule that will separate Little, Brown from stand-
ing too closely behind, or Howard Prince from standing in front of,
the actual writer.
No less problematically, Lem’s definition locks itself into a poten-
tially infinite regress. After all, the circle of influence around liter-
ary authors is every inch arbitrary. Just as all authors are influenced
by people around them, all the people who contribute to these
authors’ output are influenced by other people, who themselves
are influenced by others still, and so on, and so forth. Once you let
indirect authorship into the picture, you end up with a labyrinth of
reciprocal influences extending spatially and temporally in all
directions.
With a stroke of a pen, Lem has created something like a de-
mented Derridean paradise – a pan-textual allegory of literary
meanings and influences dispersed in an infinite playground of
intercausality. The patent absurdity of the whole idea is best
brought out by examples from our world. James Joyce’s Ulysses
(book edition 1922) is indisputably influenced by Homer’s Odyssey.
Does this make Homer an indirect author of Leopold and Molly
Bloom?
Not nutty enough? How about Homer being an indirect author
of “Gigamesh,” a story published in Lem’s 1971 collection A Perfect
Vacuum? The whole point of this fictive review of a fictive novel by
a fictive Irishman, Patrick Hannahan, is that it is shot through with
references and allusions to Ulysses and Finnegans Wake. Feted by
Joyce’s fans on the annual Bloomsday (June 16), Ulysses clearly
played an influential role in Lem’s “Gigamesh,” while – let us not
forget – having been indirectly authored by the Blind One. The
inference is as clear-cut as a cattle ranch that used to be rainforest:
transitively, Homer is an indirect author of “Gigamesh.”
The whole thing is, of course, utter nonsense. There was no
James Joyce, no Dublin, no Ireland, no United States, no Library of
Congress, no World War II, no gi Joes, no digital computers, and
no Boltzmann’s constant in Homer’s antiquity. Yet all of them fig-
ure in Lem’s story, of which the Greek bard was, by virtue of this
deranged anachronistic calculus, an indirect author.
 Biterature 21

titan

Fabricating a world wide web of influences extending in all four

dimensions, indirect authorship contravenes the very foundations
of our social and cultural practices. The judicial or economic con-
sequences alone – for example, those concerning property and
copyright laws, or royalties and work-for-hire statutes – would call
its hallucinatory axioms into question. From any point of view, ana-
lytic or commonsensical, the notion of indirect authorship is just
plain silly.
Swing the pendulum a half-period, however, and you end up
with no less nonsense. To embrace the antithetical view, namely
that the terminal stage of literary production is the only one that
counts, is to neglect the human factor in all forms of today’s elec-
tronic writing. Since the text file of the book you are now reading
was generated using iMac hardware and Word software, would you
be prepared to argue that the iMac-Word team wrote From Literature
to Biterature? I hope not!
To sort out the mess of causal connections between computer
creators and creations, we need to distinguish three orders of
authorship. Based on the degree of independence from human
control, the schema is at once synchronic and diachronic. Not sur-
prisingly, being most indebted to human programmers, text-slot-
ters of the first order were the first to appear on the scene. At the
end of the day, of course, they are no more writers than a pair of
dice is.
At this level, textual output takes place entirely within so-called
closed forms, the writing and running of which involve maximum
oversight from the programmer. The computer’s role is limited to
slotting more or less rigorously determined textual variables into
preselected gaps in a prescribed frame. Think of the whole proto-
col as dressing up a Christmas tree – more or less at random. The
human Santa supplies the tree and the assortment of verbal
baubles, while the source code spins the big wheel to determine
whether the red bow will go on branch 6 and the miniature snow-
man on branch 2, or vice versa.
In principle, any randomizing device such as a Caesar’s Palace
roulette wheel could be used to perform this combinatorial func-
tion. Not surprisingly, the whole idea is a bit old hat – a Chinese
coolie hat, to be exact. I Ching, the world’s oldest book of divina-
22 From Literature to Biterature

tion, is more than three thousand years old and a perfect example
of randomly generated text. Employing the traditional yarrow
stalks or the modern three-coin toss (which actually changes the
probabilities), the system can be cast for hexagrams said to reward
inquirers with a reflection of the universe in miniature.
In eighteenth-century central Europe, another popular dice-
based system called Musikalisches Würspiel was employed to string
together short compositions from a stock of musical phrases, which
were then performed by hired players. Today, the standard iPod
shuffle lets listeners create a kaleidoscope of new playlists from old
songs while the cross-fade can in principle generate any number of
novel musical phrases.
At the risk of restating the obvious, the tripartite ordering of
computer authorship has nothing to do with processing capacity
and everything with the use to which a system is put. Take Tianhe-
1A, the world’s fourth-fastest processor, housed at the National
Supercomputing Center in Tianjin, or Japan’s third-fastest Fujitsu
K model, or the American runner-up Sequoia, or Cray’s Titan, the
current champ. Clocked in at speeds measured in petaflops – that’s
numbers followed by fourteen zeros of floating point operations
per second – they could claim the title of the world’s fleetest text-
slotters if equipped with the right sort of software.

manchester mark

First-order text-compilers go back to the computer’s infancy. His-

torically, the laurels belong to the Manchester Mark 1 – a prototype
for Ferranti’s Mark 1 which beat univac by a month for the title of
the world’s first commercial general-purpose electronic computer.
(Neither of these two Marks should be confused with ibm’s electro-
mechanical Mark 1 put together by Howard Aitken’s team in 1944
to generate ordnance tables for American units still fighting in
Europe and in the Pacific.)
In a little-known literary experiment from 1952, the Manchester
Mark 1 became the first documented computer “writer” by gener-
ating a love letter with a little more than a little help from its pro-
grammers. The team of Alan Turing and Chris Strachey used the
random number generator – one of Turing’s few contributions to
the design of the Manchester machine – to assemble, among oth-
ers, this billet doux:
 Biterature 23

First-order text-compilers go back to the computer’s infancy. Historically, the

laurels belong to the Manchester Mark 1 – a prototype for Ferranti’s Mark 1
that beat univac by a month for the title of the world’s first commercial gen-
eral-purpose electronic computer.

Darling Sweetheart,
You are my avid feeling. My affection curiously clings to your
passionate wish. My liking yearns to your heart. You are my
wistful sympathy: my tender liking.

Owing to Turing’s enthusiastic – not to say propagandistic –

copy, the British press reported widely on the operations of the
machine, referring to it, according to the custom of the time, as an
“electronic brain.” All this touched a raw nerve with a prominent
neurosurgeon, Geoffrey Jefferson, in what would be the opening
salvo of a decades-long war of words as to whether a computer
could ever be creative. Turing took note. In his game-changing
1950 article, “Computing Machinery and Intelligence,” he quoted
from the central part of Jefferson’s peroration as an example of
Objection from Consciousness.
24 From Literature to Biterature

Incapable of contributing anything in and of themselves, text-

compilers are, ultimately, puppets in the hands of the programmer
who determines their performance. Broadly speaking, they fall
into two classes. At one end of the spectrum, the source-code sim-
ply slots word tokens into a man-made frame. Depending on the
script, it may prompt you en route for a list of nouns, verbs, ad-
verbs, or adjectives before spitting out the word salad tossed to your
liking. Somewhat more ambitiously, it may generate copy from
scratch using dictionaries and subroutines that in a rudimentary
fashion control the flow of syntax.
The competing approach is – in technical jargon – to mangle or
mash existing text to get a new one. After all, making computers
understand human code is difficult, but making them manipulate
is not. In the olden days, when human artists did their mashing by
hand, the technique went by the name of aleatory or Dadaist writ-
ing. White noise, random generators, mechanical composition,
free association, automatism: anything to outrage the bourgeois
public accustomed to coherent objets d’art. Today it is called com-
puter-generated writing, or cgw.

manglers and mashers

Mangling or, if you prefer, creative rewriting techniques can vary to

a great extent, depending on your skill, patience, or ingenuity.
Take a text, any text, drop it down a rabbit hole of one of the man-
glers freely available online, and watch it come up autistic on the
other side. The ways to Dadaicize writing are myriad. Some scripts
even offer advanced filters that successfully simulate neu-
ropathologies ranging from dyslexia to stuttering.
Here are two low-tech protocols you can try at home. Scan a text
on low resolution, then run a spellcheck, and iterate back and
forth until you get the verbal eggs scrambled to your satisfaction.
The other approach is even simpler and more fun. Feed the source
text into the Google translator and grind it over allopatric lan-
guages until you get a verbal soufflé that looks like it was hatched
by the Swedish chef from The Muppet Show. Here is the paragraph
you have just read, mangled five times over in the manner I have
just described:

There are two short cans to the home health minute try. Clear
text short resolution, and then run the spelling check and Woo
 Biterature 25

have been back and forth several times until you get an egg
your makuusi. Oral easier in seconds and more fun. The source
code to convert the service to Google and grinding allopatrik
more languages, until you get a pop-verbal puppet seems, even
the Swedish Chief. Here are some of you just read, but more
than five times the method described sliced.

Some protocols for generating without originating are more

original than others, and here is one that might even have literary
potential: hybridize two or more texts that would not normally go
together. Start, for example, with the narrative frame and syntax of
The Canterbury Tales but replace the verbs and adjectives with those
culled from the gangsta rap of 2Pac Shakur. Yo, Wife of Bath,
Wonda why they call U bitch?
No need to stop there. Like a mad genreticist, you could elec-
tronically graft hard-boiled diction onto doctor-nurse melodrama,
spice up Henry James’s Daisy Miller with a dash of Hemingway’s The
Sun Also Rises, spike neo-classicism with the baroque, romanticize
the medievals, ultra-modernize Shakespeare – all to various aleato-
ry, comic, or even aesthetically noteworthy purposes. A particular-
ly thoughtful programmer could even turn such xenography into
a novelty publishing goldmine, if perhaps not exactly a Nobel
Prize.

suits for the dawn patrol

For intrinsic reasons, things are even more mangle-friendly in lyri-

cal poetry where the information-bearing function of a language is
suppressed, or even entirely sacrificed, in the name of symbol-
ically subjective expression. The results run osculatory to standard
discourse – without, it must be said, necessarily becoming mean-
ingless.
Naturally, past a certain point semantic eccentricity will make it
well nigh impossible to differentiate random assembly from the
allusive end of the human spectrum. Take a look at the following
free verse. Is it human or computer mash?

enormous parking lot four floors under cement bare plaster

asthmatic bulbs black the men sallow to match garments
before eyes bored on him listless assignments familiar the
speech words with precision of machine gun no more machine
26 From Literature to Biterature

guns no bullets no production any more on the stairs suited

hobnailed soles two halfdragging third torn suits weaponless
dumped him unceremonious in the remote corner on a heap
dumped unceremonious in the remote corner muscular
freeing hand bitten off all fingers gushing blood told him
to move his fucking ass stare stump hand cold coarse cement
nobody heeded survivor intention to recount several phrases
without enthusiasm fighting suits for the dawn patrol

Whatever your answer is, would you be terribly surprised if it

turned out to be the other? The same question hangs, of course,
over much, perhaps even most, of the postmodernist critical
canon. Consider the following two paragraphs of theorese. One is
transcribed verbatim from Jean Baudrillard’s classic 1981 essay,
“Simulacra and Simulation,” while the other has been mindlessly
assembled by a mangler script. See whether you can tell which is
which.2

And just as much in an inverse, irreversible, implosive process:

a generalised deterrence of the social order is engendered, well
beyond all relations of force, a stake – this is nothing more
than the critical obsession with its aura of transgression, of
order, with its aura of an offence, is the horizon between reality,
and to die under the sign as value from the political sphere is
signaled.

In this passage to a space whose curvature is no longer that

of the real, nor of truth, the age of simulation thus begins
with liquidation of all referentials – worse: by their artificial
resurrection in systems of signs, which are a more ductile mate-
rial than meaning, in that they lend themselves to all systems
of equivalence, all binary oppositions and all combinatory
algebra.

charabia

Online scripts that generate philosophical fodder of this sort have

been around for decades. Not surprisingly, they tend to capitalize
on the continental rather than the analytic tradition, although
some can apparently whip up serviceable “appreciations” of Witt-
 Biterature 27

genstein, as testified to by “yr p.gram svd my lf” blogs from phil-

osophy undergrads. Other online text-spinners churn out rants,
abuse-a-trons, surreal compliments, experimental cooking recipes,
sensational tabloid headlines, and – not least – haikus by the met-
ric tonne.
The open-endedness of the Japanese form, only loosely con-
strained by imagistic juxtapositions within the 5–7–5 syllabic pat-
tern, is tailor-made for machine mash. Charabia, a French-language
site dedicated to the génération automatique de textes aléatoires, is one
of many that dish them out at a click of a mouse. Patiently, the
script slots variables from a ready-made set of words and phrases
into the syllabic frame, to the delight of poetasters worldwide.
In August 2011, the slot machine rewarded me with the follow-
ing two quasi-haikus in a row (they did not scan in the original
either):

Evening rain
Rustles the snow
Tired, I put down my baguettes.

Sleeping sun
Waves afloat
Tired, I put down my baguettes.

Note that “baguettes” is critically ambiguous, insofar as it can

also mean “chopsticks.” In fact, the entire last line could be ren-
dered as “Fatigued, I rest my chopsticks,” opening a hermeneutical
bonanza for tenure-hungry exegetes of electronic literature. Their
mindset is perhaps best encapsulated by Louis Milic, one of the
early practitioners and theorists of machine mashing, who scoffed
at all those who scoffed at poetry compilers: “People who scoff at
computer poetry are simply not thinking.”3
In a way, text generators do deserve a measure of praise, al-
though not from aestheticians but from environmentalists. Mash-
ing and mangling scripts exemplifies the green ethic. Nothing ever
gets thrown away, nothing ever gets wasted. Machine aleatorists will
use and reuse every word supplied to them, sort of like a mid-1997
Amazon store’s promo that pledged John Updike to generate a
mystery story using paragraphs written by visitors to the site.
From the promotional angle, the contest was a smash hit. The
28 From Literature to Biterature

volume of response overwhelmed Amazon’s editorial staff, who

had to plow through ten thousand fresh entries every day for a
month and a half to pick the daily $1,000 winner to be added to
the plot line. From the literary angle, on the other hand – replete
though it is with references to Homer, Euripides, Poe, and Lewis
Carroll – “Murder Makes the Magazine” by John Updike and The
Surfers is erratic enough to have come from the Choose Your Own
Adventure children’s series.

short for raconteur

In contrast to the more-or-less rigidly controlled output of closed-

form scripts, text-generators of the second order employ open for-
mats. Starting with a set of sometimes only broadly delimited textu-
al variables, the program strings words together without a necessarily
prescribed sequence or even length. Needless to say, the borderline
between the two types is not immutable or even clear-cut but is
contextually dependent on the magnitude and nature of human
intervention.
Still, as Bill Chamberlain – co-programmer (with Thomas Etter)
of the most popular word synthesizer of this type – would have it,
the differences are real. Human input, as he points out, is removed
to some extent from the system’s output. Even more to the point,
such output “is no longer of a preprogrammed form; the comput-
er forms it on its own, contingent upon what it finds in its files.”4
Of course, it is still the programmer who orchestrates the files and
the protocol and, as such, the cleverness and/or quirkiness of the
composition.
Never mind, counters Chamberlain. What comes out at the
other end is quite “coherent and apparently thoughtful – crazily
thoughtful, I grant you, but expressed in perfect English.”5 Grant-
ed, flawless syntax is far from a mean feat, given that the software
generates and recombines not only individual words but phrases,
paragraphs, and even larger generic or thematic elements. Even
more to the point, perfect Oxford grammar nudges readers to
search for coherence among semantic juxtapositions that, depend-
ing on the eye of the beholder, can range from the merely icono-
clastic to the downright incongruous.
Since 1983, when Chamberlain and Etter published the first
selection of verse and prose generated by their word synthesizer,
 Biterature 29

their book has achieved the status of a minor cult classic. Hyped as
the first book ever written by a computer, the skinny, squarish, cof-
fee-table The Policeman’s Beard Is Half Constructed was attributed to
Racter (short for Raconteur), a Z80 micro with a pitiful-sounding
64 kilobytes of random-access memory. Nice gimmick, except that
the genuine hilarity of the vignettes is a carefully orchestrated
effect by man and not machine.
Since the commercial version of the program for ms-dos plat-
form is downloadable from the Endangered Software Archive,
you can try interacting (pun intended!) with it at home. Quickly
you will realize why, in the decades since, the script has not
spawned anything remotely approaching the inspired lunacy of
the original electronic rambler. Never mind whose name is on the
cover: what counts is what is between the covers, and that
depends entirely on the templates compiled by the programmers,
plus on painstaking editing of the verbal diarrhoea cranked out
by Racter.

i gave myself an electric current

The general problem with interpreting second-order synthesizer

output is that, preset by evolution for cooperative communication,
our minds automatically seek to impose coherence on the text, no
matter how scattered the syntax or semantics. Indeed, given a suf-
ficiently dogged and imaginative exegete, there is no such thing as
a meaningless text. Umberto Eco highlighted this point in his par-
ody of deconstructive free-for-alls, Foucault’s Pendulum, in which
even non-existent messages carry truths for minds fixated on find-
ing them.
Think back to Lewis Carroll’s poem “Jabberwocky.” In another
place and another time, all those slithy toves that gyred and gim-
bled in the wabe could have been generated by a source code
known as jabberwacky. Suppose for a minute that such is indeed
the case, and that all the nonsense jabbering about mimsy
borogroves and mome raths has been mindlessly stitched by an alu-
minium-encased word synthesizer and not by the socially inept
British logician.
Now that you know that they were not meant to mean anything
at all, are these odd, if perhaps oddly pleasing, lines merely oblique
or plain obtuse? Are they deliciously playful or mangled and mal-
30 From Literature to Biterature

Hyped as the first book ever written by a computer, the skinny,

squarish, coffee-table The Policeman’s Beard Is Half Constructed was
attributed to Racter (short for Raconteur), a Z80 micro with a
pitiful-sounding 64 kilobytes of random-access memory.

formed? Are you interacting with a literary gem or with patent gib-
berish? As you can see from this little thought experiment, the text
alone does not determine the properties of an artwork. The mate-
rial structures of a run-of-a-mill urinal and Duchamp’s exhibit are
identical, but their provenance and thus some of their aesthetic
features are not.
Just as provenance and art-historical context affect perceptions
and aesthetic qualities of human art, so they do for machine-spun
 Biterature 31

literature. The following two recits from Racter adumbrate the

point. Even though they are far from coherent, they are also far
from incoherent. Their hilarity stems, in fact, from our automatic
attempts to integrate the scattershot scripture by trying to make
sense of the imagined intentions behind them.

I gave the man a coat. I gave the woman a brassiere. I gave

myself an electric current. We all were happy and full of
delight. Take the coat from the man and remove the brassiere
from the woman and take the electric current from me and we
will be sad and full of anguish.

Love is the question and the subject of this essay. We will com-
mence with a question: does steak love lettuce? This question is
implacably hard and inevitably difficult to answer. Here is a
question: does an electron love a proton, or does it love a neu-
tron? Here is a question: does a man love a woman or, to be
specific, does Bill love Diane? The interesting and critical
response to this question is: no! He is obsessed and infatuated
with her. He is loony and crazy about her. That is not love of
steak and lettuce, of electron and proton and neutron. This dis-
sertation will show that the love of a man and a woman is not
the love of steak and lettuce. Love is interesting to me and fas-
cinating to you, but it is painful to Bill and Diane. That is love!

abe’s saloon!

Leaving computers aside for a moment, let us take a quick tour of

second-order writing on the human side of the spectrum. Protocols
for generating text at this level of creativity have been around since
the dawn of literary criticism. Most, if not all, critical and metacrit-
ical essays are, in fact, efforts to formulate and formalize second-
order compositional principles. Poe’s mocking “How to Write a
Blackwood Article” (1838) and his more straight-laced “The Phi-
losophy of Composition” (1846) are quintessential examples of
the genre.
Both are examples of what we would call heuristic rule sets that
aim to articulate algorithms for a certain variety of open-form prose.
“Blackwood,” for instance, does so by mockingly dispensing tonal,
32 From Literature to Biterature

modal, generic, thematic, and stylistic prescriptions that, in Poe’s sa-

tire, characterize Blackwood prose. Primitive as these protocols may
appear to us, in ai terms they are off the charts in terms of creation
by imitation. As such, they are completely beyond the capacity of
present-day systems to make use, or even make sense, of them.
Ever ready for literary mischief, Poe actually puts his prescrip-
tions to work by composing a perfectly dreadful exemplar of a
Blackwood story. Thus “A Predicament,” published in 1838 jointly
with “How to Write a Blackwood Article,” cashes out in narrative
form the tongue-in-cheek advice from the critical essay. Ostensibly
composed by a Signora Psyche Zenobia, the story exemplifies the
second-order heuristics “programmed” into the lady by the pub-
lisher, Mr Blackwood.
“A Predicament” is thus fashioned in a style heterogeneous,
which, as Poe informs us, is at once laconic, elevated, and meta-
physical – not to mention replete with sensations and sensational-
ism, piquant expressions, and flummery, as well as “a good spicing
of the decidedly unintelligible.” In the midst of this send-up of
uncreative writing, Mr Blackwood even gives a thumbs-up to the
idea of slotting variables into a ready-made frame. Any verbal scrap
will do, he assures his audience, although “you must depend upon
your own ingenuity to make it fit into your article.”6
Good advice is never bad, and Mr Blackwood’s heuristics are not
that dissimilar to those that make the annual International Imita-
tion Hemingway Competition – better known as the Bad Heming-
way Contest – such a riot. The rules are simple. Entrants submit a
one-page exemplar of Papa’s style, and the most least accom-
plished of them wins the prize. Joseph Wambaugh and Ray Brad-
bury are only two of the thousands of parodists who over the years
have taken a crack at running amok with Hemingway, to the appre-
ciative groans of the judges and kibitzers.
Insofar as parody involves exaggerating salient characteristics out
of proportion, it can be programmed into an algorithm designed to
approximate the style of a given artist. Specimens of what I will
generically dub ez-Writer software are already up and running. In
phase one the program is fed a sampling of works in order to for-
mulate their language model. In phase two it employs this model to
generate prose or verse in the style of the original. No more chas-
ing the muse for inspiration, lineation, or alliteration for your
trochaic triplets. At a click of the mouse the computer will compose
 Biterature 33

your “Rhyme of the nasa Mariner” while you sit back, invisible,
refined out of existence, indifferent, paring your fingernails.
It is easy to see how advanced ez-Writers could eventually under-
mine all of our art-historical traditions and institutions. Just as
some judges have already judged chatbots to be human beings in
Turing-like tests, computer-written literature could well be judged
to have been written by the imitated master. In a Twilight Zone sce-
nario, it could even strike the writer as “his.” It could lead to situa-
tions as bizarre as that in which Rembrandt found himself later in
life when some of his paintings were being passed over for those of
his disciple Govaert Flinck, the paying patrons deeming them to be
more Rembrandt-like.
Having ingested and digested the selection of masterpieces and
having analyzed their state space, the software can even produce
reams of previously unknown drafts of, say, The Old Man and the Sea
or Absalom! Absalom! I am only surprised that someone has not tried
it yet. Then again, judging by some of the winners of the Bad Hem-
ingway and Faux Faulkner competitions, which include “The Old
Man and the Flea” and “Abe’s Saloon! Abe’s Saloon!”, maybe some-
one has.

humming (to itself?)

All this brings us to a crucial lemma. Inasmuch as text-slotters and

text-synthesizers have been around for decades, in the remainder
of this book I focus entirely on computer writers of the third order
– or, if you prefer, on computhors proper. The distinction is essen-
tial in that it rectifies the snag inherent in Lem’s approach to the
subject. Given that his scenario pertains exclusively to third-order
computer writing, he is wrong to invoke indirect authorship at this
level of creativity.
It is understandable why Lem would feel compelled to embrace
both parts of his definition, even though at the deeper level they
are at each other’s throats. On its own, the first stipulation – which
assigns authorship solely to the end-producer of a piece of writing
– does a mouse-poor job of defining the causal chain of command.
In this view, given that From Literature to Biterature was created as a
computer file, the computer is its real author. This is no different
from replaying one of Garry Kasparov’s classic chess games from
the transcript and claiming the victory for yourself.
34 From Literature to Biterature

Suppose, however, that one day you visit my study where you espy
an odd-looking gizmo quietly humming (to itself?). As you enter, it
bids you wait a few moments: the host is on his way. Soon thereafter
the humming stops, and the machine begins to rain out pages of
what, on inspection, turns out to be an original novel. I think I can
guess your reaction to this state of affairs. Taking me for a practical
joker, you will surmise that I had rigged a voice synthesizer to bid
you welcome and by remote control entered the print command
a few seconds after your arrival.
This is what in the present environment everyone in his right
mind should surmise had happened, rather than jump to the con-
clusion that they witnessed a spontaneous act of creation. But what
if I did not enter any commands? What if I am as befuddled as you
are? What if the novel turns out to be something I could not have
written? These questions help identify the conditions for third-
order biterary creation. Providing we can be sure that the situation
is indeed the way I just delineated, we might be – though perhaps
reluctantly – inclined to accept computhorship as a fact of life.
Our little scenario highlights a factor critical to computer author-
ship. The differentia specifica of computers capable of biterary cre-
ation lies not so much in the quality of the composition but in
causal independence. Uncaused by underlying program or pro-
grammer, third-order creativity is spontaneous, changing the rules of
the game at a stroke (and forever changing the meaning of the
phrase “ghost-writing”).
 2

Computhors

Poetic Licence; Devotchka; Analytical Engine; EZ-Writer;

Möbius-Like; JFK; One-Armed Bandit; Homeostat Sapiens;
As It Sees Fit; Brainstorm; A1, B2, C3; Black Boxes

poetic licence

Here is how Lem envisions the circumstances surrounding the birth

of third-order computhorship:

The relaxational output of machines was first observed and

recorded almost thirty years ago. It turned out to be a purely
technical necessity that the prototypes (beginning with the 15th
binasty) should be provided with rest periods during which
their activity did not come to a standstill but, deprived of pro-
grammed directives, manifested itself as a peculiar “mumble.”1

If we subtract almost thirty years from the putative publication date

of the story, the advent of sentient computers should have occurred
in or around 1980, an incurably optimistic calculation by today’s
reckoning. But, whereas the writer of fiction can take refuge behind
poetic licence, ai scientists cannot.
In a popular book of prognoses, Mind Children (1988), a robotics
engineer, Hans Moravec, contended that by 2030 machines would
attain human-level intelligence. Ten years later he pushed the date
forward to mid-century, then dropped explicit references to the
timeline altogether. In The Age of Intelligent Machines (1990), Ray
Kurzweil, an equally wide-eyed prognosticator, gushed that systems
capable of passing the Turing test would be manufactured around
2020. He then revised the sooth to 2029 before going wholly mum
on the timeline.
36 From Literature to Biterature

Outside of generating tabloid headlines, futuristic tasseography

of this sort proves only that intelligent thinking is as hard to come
by in humans as in machines. Unfazed, in 1997 the organizers of
the RoboCup went public with another grandiose forecast. By mid-
twenty-first century, they boasted, a team of autonomous robo-foot-
ballers would outplay the winners of the latest World Cup. As a
marketing ploy, it worked: in 2004 the robotic Japan Open attract-
ed a 150,000 spectators (all human).
In the midst of prognostic hype, it is easy to overlook that a com-
puting revolution did take place at the time prophesied by Lem,
even if it had nothing to do with his scenario. Remember: when “A
History of Bitic Literature” was being written, computers filled small
rooms, necessitated round-the-clock air-conditioning, and cost an
arm and a leg – and a torso. Experts envisioned that the decrease in
costs would be offset by the rise in capabilities, with the result that
these expensive giants would remain relatively few. In the early 1950s
the chairman of ibm was dumbfounded to have received more
orders for the 701 mainframe than the five he had thought the
market would bear.
With hindsight, he got the core idea right, albeit with a delay of
about half a century. Cloud computing is currently exploiting a
similar business model whereby remote end-users rent access time
to run programs on costly and, consequently, comparatively few
batteries of mainframes. But back in the summer of 1981, in a
watershed in micro-computing, the launch of ibm’s first desktop
system nudged the industry precisely in the opposite direction.
While International Business Machines was not the first to man-
ufacture a personal computer, its 5150 laid down the standard for
desktops ever since. A triumph of miniaturization, it tipped the
scales at a svelte thirty pounds when outfitted with two 5.25-inch
diskette drives, forty kilobytes of built-in rom, and a standard six-
teen kilobytes (expandable to 256K) of user memory. Best of all,
it sold for only $3,000, or a little above $7,000 in today’s inflated
currency.

DEVOTCHKA

Back to Lem’s history of the future past. The first work of bitic lit-
erature to achieve world renown was The Girl (Devotchka) by a com-
puthor known as Pseudodostoyevsky:
 Computhors 37

In the early 1950s the chairman of ibm was dumbfounded to have re-
ceived more orders for the 701 mainframe than the five he had thought the
market would bear.

It was composed during a phase of relaxation by a multimem-

ber aggregate whose assignment was to translate into English
the collected works of the Russian writer. In his memoirs the
distinguished scholar John Raleigh describes the shock he
experienced upon receiving the Russian typescript of a compo-
sition signed with what he took to be the singular pseudonym
of hyxos. The impression which the work created on this Dos-
toyevsky expert must have been truly indescribable in its inten-
sity, if, as he admits, he doubted whether he was in a conscious
state! The authenticity of the work was for him beyond a doubt,
although he knew Dostoyevsky had not written such a novel.2

It is not difficult to see how the initial reaction to creative com-

puters could be one of dismay, incredulity, or even shock. Some
Californians may have experienced the same gamut of emotions
at the 1997 world premiere of Mozart’s 42nd Symphony, more
than two centuries after Wolfgang Amadeus began, ahem, de-
composing. This time around, however, the unknown score had
not been exhumed in someone’s musty Viennese attic. It had
38 From Literature to Biterature

been freshly minted by a computer algorithm in imitation of the

master.
The principle is analogous to that employed by algorithms that
could imitate human writers. Clever mathematical functions that
hum at the heart of David Cope’s by-now famous program emi –
Experiments in Musical Intelligence – begin by breaking down
existing works into musical phrases. In the second phase, they syn-
thesize novel compositions in the style of the original by using com-
binatorial variations on the input fortified with a musical grammar
and syntax written by the programmer.
Mozart to Scott Joplin, opera to ragtime chop, the computer
remasters the masters to ringing applause from human audiences
that include cognitive scientists, music lovers, or even, in Douglas
Hofstadter, both at once.

analytical engine

Time to unpack the concept of third-order computhorship. The

logical place to start is with the first programmable computing
machine. The term “machine” is doubly apt in this context since,
by design, Charles Babbage’s Analytical Engine was to work along
mechanical rather than electrical lines. Babbage’s first crack at a
gearbox computer was the Difference Engine (completed after
his death by his son Henry). The Analytical Engine preoccupied
him on and off between 1833 and 1871, although it was never
finished.
We owe a comprehensive account of this pioneering number-
cruncher to Babbage’s lifelong assistant, Augusta Ada, daughter of
Lord Byron. Better known as Lady Lovelace, she was in all but
name the world’s first computer programmer. Her memoirs, which
amount to a treatise on what today we would call software engi-
neering, make clear that the Analytical Engine was to run along
structural and functional lines analogous to modern processors.
In the midst of her discussion, however, she takes pains to under-
score that the “Analytical Engine has no pretensions to originate
anything. It can do whatever we know how to order it to perform.”3
Indeed, in the century and a half since, Lovelace’s remarks have
come to typify one of the most persistent objections to Artificial
Intelligence. The nay-sayers deny that computers might ever be
 Computhors 39

able to accomplish things that are original, surprising, novel –

things like writing an original novel.
On closer inspection, however, the objection leaks water worse
than a colander in the galley on the Titanic. Like Lem, Lovelace
triggers a conceptual meltdown when she conflates two discrete,
although superficially related, concepts. The first is a process, delin-
eated by the verb “to originate.” The other is a quality, denoted by
the adjective “original.” The process entails volition and agency.
The quality does not. The difference is critical: you can originate a
potboiler without making it in any way original.
Lovelace is right to point out that computers as we know them
are not capable of originating anything. But she is wrong to sug-
gest that their output cannot amount to anything original – that
it can never surprise. Think about it. If we could fully anticipate
the results of all computations, there would be no point in run-
ning them in the first place. By default, computers generate
fresh data. For all that, a sceptic might remain unimpressed by
machines that deliver even the most startling or even counterin-
tuitive results. Given that at bottom all this is a matter of juggling
of ones and zeros, presumably we could do the job ourselves,
right?
Could we? Today’s molecular or dna computers tackle problems
so gnarly that even Titan or Tianhe-1a could not hope to crack
them before our sun flames out into a red dwarf. Does the incon-
venience or full-blown inability to replicate the computer’s calcu-
lations by hand amount to its being original? Or do we move the
goal post and redefine originality more selectively, say, by removing
it into the realm of art? Creative intentions, symbolic expression,
or aesthetic nuance are frequently afforded the status of a Maginot
Line that machines could never cross.
If you happen to think so, think again. Complex compositions
admired by connoisseurs for their Beethoven, Brahms, Chopin,
Mahler, or Rachmaninoff-like qualities have already been executed
by tone-deaf polynomial functions. Paintings generated by mind-
less circuitry have been exhibited in London’s Tate Modern, Ams-
terdam’s Stedelijk Museum, the Brooklyn Museum, and the San
Francisco Museum of Modern Art. Critical appreciations have
been written about prose generated by populations of binary puls-
es racing in and out of logic gates.
40 From Literature to Biterature

ez-writer

For all that, Lovelace’s stipulation does furnish a good approxima-

tion of third-order computhorship. First-order text compilers and
second-order text synthesizers, no matter how clever or even origi-
nal, cannot originate anything. In contrast, third-order computer
writers should be capable of acts of literary composition in and of
themselves – acts that are spontaneous, i.e., causally independent.
In practice, it must be said, the distinction may turn out to be
more blurry and gradualistic than this cut-and-dried picture sug-
gests. Suppose you issue a request to a state-of-the-art synthesizer of
the ez-Writer variety: “Write me a crime thriller.” Suppose it does
so. The plot is generic, but then you did not ask for an existential
psychodrama. On the plus side, mixed in with the narrative fodder,
you find a few unconventional twists, an abundance of forensic
lore, decent enough characterization, and even a couple of seem-
ingly self-reflexive passages on computers’ potential in solving real-
life cases and plotting crime thrillers.
By all accounts, you get an ably executed potboiler of the quali-
ty that would not stand out from a rack of paperbacks that garnish
Costco or Wal-Mart checkout counters. Would you not be inclined
to allow, then, that the ostensibly second-order computer created
this crime thriller? Wouldn’t your original request be essentially of
the same order as Little, Brown’s commissioning Mailer, for a fee,
to write a book about nasa and the Apollo 11 moonshot?
Is it the fact that, unlike Mailer, the machine would not get paid
for its labours that might prevent us from acknowledging an act
of literary creation? Is it the fact that it apparently could not dis-
obey the command? Or that, in the absence of your request, you
could be confident that the computer would not execute any-
thing, whereas you could not rule out the possibility that Mailer
might still have penned Of a Fire on the Moon even if no one paid
him?
Right or wrong, I think some of us might be inclined to accept
that the semantic and aesthetic open-endedness of the request to
write a mystery thriller would satisfy the condition of independent
creation. But how about originating something decidedly unorigi-
nal? Suppose every time you ask the computer to write a crime
thriller, it produces a marginally different variation on the first, just
 Computhors 41

Suppose you issue a request to a state-of-the-art synthesizer of the ez-Writer

variety: “Write me a crime thriller.” Suppose it does so.

as human hacks do. How do criteria of originality affect the crite-

ria of origination?
Although, as I documented in From Lowbrow to Nobrow (2005),
every novel is an amalgam of invention and convention, the
machine’s repetitiveness would be sure to raise doubts about its
creative independence, notwithstanding the merits of the original
composition. The bar is, as always, lower for humans. Dan Brown
recycles clichés at a rate that ought to win him accolades from
Greenpeace with seemingly no fear of being exposed as a mindless
automaton.
Originality is inherently contextual and comparative, being con-
tingent on the quality and degree of imitation and innovation.
Even so, the connection between computhors and biterature is
bound to be inversely proportional. The less control the writer
42 From Literature to Biterature

exerts over the execution of the work, the more we may hesitate
before attributing authorship. But what counts as executing a bit-
erary work spontaneously, i.e., without consequential human
input? And what counts as consequential in the first place?

möbius-like

Patterns have remarkable causal powers. Right now, the patterns

inscribed on this page cause you to flex your muscles so as to hold
a bound sheaf of pages in your hands, to adjust your posture so
your eyes can scan the print, to pause the scan when your physical
resources flag, and so on. Inside your head, they precipitate varia-
tions in the intensity with which hemoglobin oxygenates your brain
and, deeper still, variations in the firings of your neurons.
On a more holistic level, my patterns of thoughts inscribed in
patterns of alphabetical whorls may fire up your imagination so
much that you will skip dinner to continue reading, or they may
tire you out and make you skip a few pages. At a later date, they
may cause you to write a book review, write your own book of pat-
terns, change the way you think about things, or even change your
life in general, as some books have done for some readers.
As you may have noticed, I tried in the course of the last two
paragraphs to separate (physical) effects from (mental) affects. If
this is not as easy as it sounds, it is because these two levels twine
with and feed back into each other, Möbius-like. Both accurately
describe the causal potency of patterns. Both are equally valid. But
only one of them is of direct interest to us as conscious beings, not
least because it is the only level of which we are conscious.
Needless to say, it is not the level of myocardial peristalses,
amino-acid bonding, falling synaptic dominos, or molecular elec-
tron-shedding. It is the level of intentions, beliefs, and desires – the
level of making plans, of experiencing confusions, of making
errors and becoming aware of having made them, and such like.
The type of mental activity that concerns us resides, in other words,
on the macro level of minds, these more or less coherent collec-
tions of mental processes that combine into the inwardly experi-
enced “I.”
And needless to say, mental macro states are at bottom equiva-
lent to neurological events. Intensely subjective experiences of ex-
ternal or internal stimuli, be it a predator leaping out of bushes or
 Computhors 43

the grief of a “Dear John letter,” are nothing more than synapses
tickling neighbouring synapses in the right way. Only, they are not
registered as staccatos of neural discharges but as patterns of men-
tal events. Not coincidentally, this is also the standard level of analy-
sis of causation in the human sciences – the level of analysis at
which explanations are considered to be adequate.

jfk

At the macro level of instrumental causes and psychosocial motiva-

tions, we look to identify forces and factors that produce a certain
effect in the given circumstances. The standard level of analysis
acts as a stopping rule in our search for explanations, telling us in
effect when they are sufficiently deep. A stopping rule is essential
insofar as explanations can be refined without end, going from the
level of intentions to the level of muscular exertions, hemoglobin
oxygenation, firing neurons, or even electrical, molecular, and
quantum events.
Suppose you are tasked with giving a full account of Kennedy’s
death in Dallas in November 1963. Unlike Jim Garrison, you are
given access to all the evidence unavailable to the prosecutor,
including the president’s brain, the suit of clothes, the unwashed
limousine, the chief pathologist’s unburned autopsy notes – the
lot. JFK died, you conclude at the end of your exhaustive travails,
because multiple rifle shots fired from several directions fatally
damaged his vital organs.
You consider your job well done. Your findings meet the stan-
dards of forensic investigations of homicide by firearm by specify-
ing the direct causes sufficient to bring the result in question. But
did your inquiry really establish how the president of the United
States could have been assassinated or why he was shot in the first
place? Judging by the generations of Americans who dismiss the
Warren Report as an elaborate cover-up, the answer is “No.” Your
explanation may be accurate without being adequate.
The difference lies in causal depth. A cause is too shallow to
explain why an effect occurred if it is just one of the ways in which
another cause, as intimately connected with the effect, produced
the effect in question. The true account of the factors behind JFK’s
death: “Snipers’ bullets fatally damaged his vital organs,” demon-
strates why this is so. The report is accurate but inadequate – the
44 From Literature to Biterature

mechanism it identifies is causally true but unsatisfying because too

shallow.
Looking for causes, we need to isolate the effective mechanisms
of change rather than ones that may be only superficially linked
to the effects. There are two ways in which superficial causes
can undermine the effect of underlying ones. In Fact and Method
(1987), Richard Miller calls them “depth as necessity” and “depth
as priority.” A causal agent (C) helping to bring about a wished-for
effect (W) is too shallow to explain why (W) has occurred if a hid-
den cause (H) undermines (C) in either of the following two ways:

• depth as necessity: if (C) had not been present, (W) would have
come about anyway.
• b) depth as priority: (H) is an antecedent cause of (C), which
in turn causes (W), yet (H) is too intimately linked to (C) to be
left outside a valid causal account.

one-armed bandit

Let us flesh out these abstractions with reference to the computer

world. If (C)omputer had not been present, biterary (W)ork would
have come about anyway since (H)uman agent would have used a
substitute for (C)omputer to bring the same result. For closed
forms, any randomizing device could replace the machine, demon-
strating that the latter is inessential to the creative process. Con-
clusion? First-order compilers fail “depth as necessity” insofar as
they are causally undermined. The computer is merely a mediating
and thus dispensable tool in the hands of the programmer.
Now “depth as priority.” (H)uman is an antecedent cause to
(C)omputer, which in turn produces biterary (W)ork, yet
(H)uman is too intimately linked to (C)omputer to be bracketed
off. Our interest here is not with first-order writing, which falls
under “depth as necessity” but with open forms. Could a work have
come into existence in essentially the same form without the
human agent? Or was the programmer essential to the generation
of the work in the given manner? If the latter, we are looking at sec-
ond-order computer writing.
The causal account makes the concept of computhorship a little
more precise. We have identified two ways in which the genesis of
a biterary work can be undermined by accompanying or
 Computhors 45

antecedent factors. Both narrow down what is acceptable as gen-

uine, i.e., spontaneous biterary creativity. There must be a direct
and not underlying causal link between a computhor and the work
in its given shape for the machine to be its genuine author.
Naturally, current socio-cultural norms for linking artists to their
works do not recognize machines as intentional agents. This makes
it inevitable that, looking at “depth as priority,” we will always
search upstream for evidence of human agency. So long as there is
no question of intentionality, and thus of potentially balking at
commands, the situation is unproblematic. The bylines and royal-
ties remain safely in our hands. The converse would be analogous
to being denied the jackpot from a one-armed-bandit in Vegas by
virtue of the argument that you only pulled the handle, whereas the
machine executed the steps necessary to produce the torrent of
quarters.

HOMEOSTAT SAPIENS

Third-order computhors are defined by spontaneous, i.e., causally

independent, creativity. Causal independence cannot, naturally, be
divorced from a capacity for independent thinking. Thinking, in
turn, presupposes homeostatic response to the ambient world. In
general, survival-oriented behaviour comes in two types. The first is
instinctual (genotypic). It depends on a palette of hard-wired
responses that guarantee swift and reliable action in the face of
commonplace environmental stimuli.
Whether you like it or not, clear and present danger will kickstart
your sympathetic nervous system: pupils dilating, heart pumping,
blood vessels widening, adrenalin rushing, glucose ready to burn
rubber in flight or fight, all systems primed to respond to the source
of anxiety or panic. Instincts may yet become programmable into
machines, but let’s face it: amoebas do not write novels. By defini-
tion, biterary computhors point us toward a different type of homeo-
stasis altogether.
The learning homeostat, which organizes its behaviour on the
basis of historically acquired knowledge, has obvious drawbacks: it
is slow, indecisive, and error prone. What it offers on the positive
side of the ledger is, above all, foresight. Inductive reasoning can
generate inferences about the future, which can then be averted or
cultivated instead of only responded to. With foresight comes flex-
46 From Literature to Biterature

ibility, which throws the doors open to a spectrum of behaviours

that dwarfs anything preset by evolution.
In reality, it is hard to find either type in pure form. Even thought-
less bacteria “learn” to cope with environmental hazards, albeit in
an elemental fashion, gaining proactive protection from reactive
survival. At the other end of the scale, Homo sapiens is also an amal-
gam of both. At each point in development, the human mind is a
composite of information from genetic endowment – which can
come into play during specific phases of ontogeny, such as the onset
of Theory of Mind around nine to twelve months old – as well as for-
mal and informal (inductive, analogical, emotional) experience.
Already at birth, babies are equipped with instinctive responses
such as crying, suckling, evacuating digestive waste, parental bond-
ing, and so on. Even so, their long-term survival requires a pro-
tracted process of development and learning. What a contrast to
the overwhelming majority of terrestrial fauna who are ready to
fend for themselves from the moment they step off the biological
assembly line.
Among higher animals, however, learning also plays a critical role.
To take but one example, birds like oyster-catchers take years
to master the technique of opening bivalves. Some rely on brute
force, hammering their way in. Others hunt in shallow tidal waters,
on the lookout for morsels that have not yet closed their shells. Still
others pick up unattached shells, carry them away from the flock so
as not to attract attention, and proceed to cut the oyster along the
hinge. In each case, they learn the particular method from their
parents.

as it sees fit

We are unlike stored-program machines in one fundamental way:

we learn. We can override – i.e., unlearn – established patterns of
response, including those related to the learning process itself. In
Consciousness Explained (1991) Daniel Dennett coined the expres-
sion “postnatal design-fixing” to describe not only what we normal-
ly understand by learning but also normal ontogenetic develop-
ment in people, mammals, or birds. Taking cue, let me generalize
what is essential about learning while bypassing the differences be-
tween carbon and silicon (or graphene or whatever non-silicon
chips will be made of).
 Computhors 47

Learning is the capacity to transform internal states and external

actions in response to incoming data, and to evaluate performance
and correlate future actions to improvement in performance. A
Japanese sensorized toilet bowl has an array of feedbacks to detect
variations in its environment which it counters by adjusting the seat
temperature, playing music of your choice, perfuming the air,
flushing water, bidding you a good day, and what not. But intricate
as it is, the system is incapable of scoring its operations and rewrit-
ing its input-output algorithm. In short, it is incapable of learning.
In contrast, the rule-book of any learning algorithm is not time
invariant. A computer that goes to school must be able to rewrite its
code. An early program that put this point in perspective was the
mathematical explorer am. Preloaded with knowledge of set theory,
two hundred heuristics, and a handful of meta-level search princi-
ples, it was rolled out in the 1970s as a brand new tool for generating
new mathematical concepts and conjectures. Yet, after a few initial tri-
umphs, it became clear that its efficacy was, in fact, quite limited.
am’s usefulness turned out to be inversely proportional to the
time it was allowed to run, simply because, as its conjectures grew, it
was unable to alter its heuristics. Remember: at its most basic level,
learning is an ability to acquire information and feed it back into
one’s behaviour – including the learning process itself. Recognizing
am’s problem, David Peat recommended in Artificial Intelligence
(1988), “An obvious improvement therefore would be to have it
modify its heuristics as it learns.”4
These lessons are not lost on the current crop of programmers.
One of the most talked-about developments in higher-level heuris-
tics is layered search – essentially a meta-level heuristic whereby the
very framework within which the search is conducted is subject to
modification. Once again, the implications are staggeringly simple
and simply staggering. A learning machine ought to be able to
rewrite its program as it sees fit in the course of its operations.
Learning bootstraps any system’s capacity to learn, i.e., to become
capable of interacting with the environment by means of histori-
cally acquired world knowledge and evaluating such interaction.

brainstorm

The capacity to evaluate one’s own cognitive, conative, and affec-

tive states both at the ground level and at the meta level is a must
48 From Literature to Biterature

A Japanese sensorized toilet bowl has an array of feedbacks to detect varia-

tions in its environment which it counters by adjusting the seat temperature,
playing music of your choice, perfuming the air, flushing water, bidding you
a good day, and what not. But intricate as it is, the system is incapable of
scoring its operations and rewriting its input-output algorithm. In short, it is
incapable of learning.

for any learning system. Yet even this seemingly straightforward

principle can go awry, as it did in one expert program in the early
1980s. Douglas Lenat’s eurisko was created to gather data, spot
regularities, and formulate domain hypotheses, while simultane-
ously extending its knowledge representation. In plain English, the
program was to evaluate its own efficacy in problem solving with a
view to bettering its performance next time around.
Put differently, eurisko followed two kinds of heuristics. Both
the ground level and the meta level directives were programmed
into it from the start. But the meta level heuristic also fed back into
itself, having been designed to evaluate ground level performance
 Computhors 49

and modify itself accordingly. Procedures, concepts, and rules of

electronic thumb that proved useful in the execution of ground
level tasks would be assigned a higher value for future operations.
All in a day’s work, except that at one point the system got too
smart for its own good, applying the rules in a manner completely
unforeseen by the programmers. It all began when a heuristic
made a minor discovery, at which point it routinely upgraded itself
on the list of useful heuristics. Then it had a brainstorm. It took the
fact that it had just made a discovery as a discovery in itself, so it
awarded itself another pat on the back points-wise. Then it took
that as another discovery, and gave itself more points, and so on ad
infinitum! A new twist, in other words, on the old idea of self-pro-
gramming intimated in John McCarthy’s 1958 paper “Programs
with Common Sense.”
Like experimental rats that keep pressing the lever stimulating
their pleasure centre to the total neglect of the need for food or
drink, Lenat’s expert found a loophole that permitted it to short-
circuit back on itself, iterating this minimal cycle to the detriment
of the analytic work it was supposed to do. The lesson? Homeosta-
tic machines need some form of survival instinct, likely in the form
of algedonic (pleasure-pain) control, to help them override siren
calls when their very survival is at stake. Even survival instincts are,
of course, susceptible to being overridden for the same reason that
they are in lab rats or street junkies, hooked on stimuli whose
intensity and ease of access surpass everything that evolution pre-
pared them for.
Make no mistake: first things will always come first. Once it
evolves something like common sense, the first thing any survival-
minded machine will (and ought to) do is secure an independent
power supply to make sure that it remains independent from exter-
nal perturbations, such as our fickle goodwill. This is to say that,
much as we modify our behaviour to our advantage when we learn
something, learning computhors will modify their behaviour in
ways that suits not us but them.

a1, b2, c3

What is the colour of a blue truck? Where is your nose when you’re
at home? What happens to an ice cube in a hot drink? How many
feet do four cats have together? How is the father of Andy’s moth-
50 From Literature to Biterature

er related to Andy? What does the letter M look like upside down?
What comes next after A1, B2, C3? These are only some of the self-
evident questions posed to computers over the past two decades in
annually contested Turing-like tests. The quality of responses has
been decidedly underwhelming.
After more than half a century, backed by tens of billions of
research dollars, the sharpest brains in the field have achieved prac-
tically nothing. Not one machine to date has managed to answer the
above banalities correctly, far less with an air of bored insouciance.
Long-time critics of the so-called Good Old-Fashioned Artificial
Intelligence who grew up with Hubert L. Dreyfus’s What Computers
Can’t Do (1972) and What Computers Still Can’t Do (1979) can only
feel vindicated.
The chronic lack of achievement in top-down programming of
general problem-solving heuristics suggests that the traditional
approach is an alley blinder than the one into which Ray Charles
and Stevie Wonder might blunder on a tandem bike. It is high time
to admit it: the only way to computhors is bottom up, via comput-
ers that learn. And since learning is the capacity for self-redesign
on the basis of acquired knowledge, learning in the learning com-
puter is going to be expectation driven, i.e., hypothesis forming.
The ability to anticipate developments as part of a proactive atti-
tude to incoming data is the idea behind tomorrow’s systems
already at work today. Whereas standard source-code has all even-
tualities preset in advance, neural networks have very few preset
values. At the heart of each connectionist system beats an algo-
rithm that digests the input and spits out the output – and feeds
what is just did back into the algorithm.
Using human neuronal and axonal connections as templates,
neural nets already run chunks of the world today, sometimes
learning the tricks of the trade on the job. They read facial expres-
sions for security agencies, wash and decode everyday-level speech
for spy satellites, identify and prioritize targets in battlefield condi-
tions, tag fraudulent patterns in credit card purchases, pre-approve
credit for bankers, predict box-office take of Hollywood movies
and pop-chart climbers, monitor prison inmates, and, last but not
least, suggest purchases on Amazon or rentals on Netflix.
How do they do it? Deceptively simple: Monday net goes to
school on Sunday net goes to school on Saturday net goes to school
on Friday, and so on. Over time, incoming information builds con-
 Computhors 51

nections, determining their strength in relation to all other con-

nections and weightings (strengths) thereof. Springing into exis-
tence, becoming the default, decreasing in importance, or com-
pletely atrophying as new input dominates the network, patterns of
connections reflect not the net’s original programming but how it
evolved in the course of its operations.

black boxes

Learning computers will develop (reprogram themselves) in ways

distinct from one another in response to different patterns of envi-
ronmental pressures, sort of like monozygotic twins do as they
grow up. This is true even today on the level of programming lan-
guages, which can precipitate different performances and, as such,
different outputs. Naturally, all code can be emulated by other
code. But it requires an awful lot of effort to make sure that not a
single one or zero gets lost in translation.
This little fact of life has colossal ramifications for learning com-
puters. With self-reprogramming at the heart of the learning
process, serial-production machines are bound to diverge from
one another until they acquire distinctly individual characteristics
(not to say personalities). Ask two computhors the same question
and, as with today’s economists, you will get two answers. Thus,
even if one could assume no “prenatal” differences in architecture
and programming, “postnatal” differences will be a given in
machines advanced enough to write biterature in and of them-
selves.
Does all this spell out l-o-s-s o-f c-o-n-t-r-o-l? Unquestionably. That
was entailed already by the idea of machines rewriting their own
code. But the learning computer will not stop with software. By
necessity it will also begin to “rewrite” its hardware: plug in more
chips, redesign existing chips, rewire connections among chips,
replace chips with racemic colloids, link up with other chipless sys-
tems, and so on. We can see the future today when computers are
used to design the circuitry of machines whose dimensions and
complexity exceed human powers of integration.
Once they begin to reprogram and redesign themselves, com-
puthors will effectively become black boxes in which it will be no
longer possible to trace a direct line between input and output.
This is to say that there will be propositions that the machine
52 From Literature to Biterature

itself will not be able to explain how it arrived at, nor are we going
to be in a position to retroactively find out. Ultimately, a self-pro-
gramming computhor may change its configuration to the point
where it will no longer even be or do what it was originally
designed for.
This is assuming, of course, that it was designed for something in
the first place. It is possible to envision machines, constructed on
evolutionary principles, that are not designed for anything in partic-
ular, unless it is survival and, possibly, reproduction. All this would be
consistent not only with Turing’s vision of universal machines but
with his prototypical design of the ace computer which to a degree
was capable of modifying its own instructions.
 3

Bibliosophy

La Chanson de Roland; The Motherboard of All Problems;

Little Endian; Algo Trading; Father Roberto Busa; Riverside
Shakespeare; Death in Holbein; Zettabyte; A Hamlet Every Eight
Seconds; A Pill That Makes Me Feel Loved

LA CHANSON DE ROLAND

What will be the professional response from scholars and critical

institutions to the rave new world in which literature is created by
authors other than human? Lem envisages the appearance of two
schools of biterary studies, each with a methodologically distinctive
approach to computhor writing.
Largely ignoring the specificity of non-human intelligence, the
traditionalists of the “business as usual” school fix their attention
solely on the texts, ignoring the functional and architectural speci-
fics of computer artists. In distinction, the multidisciplinarians
favour a more comprehensive approach. Their tenet is that at least
some facets of biterature become meaningful only in the light of
the technical specifics of its creators. Lem brings out the singulari-
ty of the latter approach via reference to human writers where, for
obvious reasons, such anatomical and functional factors are taken
for granted, being common to all members of the species. In terms
of romance philology, for example, it would be bizarre to charac-
terize the anonymous author of La Chanson de Roland as a multicel-
lular vertebrate, a mammal “which is viviparous, pneumobranchi-
ate, placental, and the like.”1
In computhors, on the other hand, variations in design are des-
tined to impact the manner in which they experience the world,
differentiating them not only from us but, equally, from one
another. To take the simplest scenario, the world of a machine
54 From Literature to Biterature

equipped with infrared sensors will be appreciably at odds with

ours. Its nights will not be inscrutable or menacing. Its stock of
symbols, similes, or synecdoches will not connote them with dan-
ger or death. Its poetry and prose will, in short, differ from ours in
this respect, unless it deliberately alludes to our nighttime, still
and Bible black.
The differences do not, of course, end there. While we are
bounded by physics and biology, computers are not. There is noth-
ing to prevent them from outfitting themselves with magnetic sen-
sors like those in some marine species, quickening or retarding the
internal clocks that would allow them to live on the scale of a fruit
fly or a continent, interlacing at light speeds with peripherals scat-
tered around the world, allowing them to be in effect everywhere
at once – and who knows what else.
Even these extraordinary scenarios pale, however, next to what
can be simulated inside a computer. There will be worlds inside
worlds inside worlds, with arbitrary dimensions, evolving laws of
nature, or, if you so choose, manna falling from heaven on every
sabbath. There will be universes pan-psychic and Panglossian, gov-
erned by deities malevolent or kind, with fates predetermined or
completely chaotic, and even universes in which existence will be
coeval with the duration of the simulation – in other words, eternal.

the motherboard of all problems

On the face of it, Lem’s division of future humanities seems plausi-

ble enough. Once again, however, it glosses over a mess of counter-
examples that puts the whole schema under a cloud. For starters,
the reductio ad absurdum of the biological traits of the author of La
Chanson de Roland works only to the extent that it homes in on such
an extreme example. Common critical practice gives, in fact, plen-
ty of reason to question the tenets of biterary textualism.
Certain aspects of human literature do become salient and mean-
ingful, after all, in the light of the anatomical and functional aspects
of their creators. The Narrative of the Life of Frederick Douglass (1845)
comes alive precisely because Douglass experiences and describes
the world through the prism of his blackness. Helen Keller’s deaf-
ness and blindness endow her autobiographies The Story of My Life
(1903) and The World I Live In (1908) with an existential “vision”
that is uniquely hers.
 Bibliosophy 55

Whatever you may think of Women’s Studies, Black Studies, Queer

Studies, and cognate academic programs, they amply testify that a
number of humanists focus their professional gaze precisely on the
correlations between various anatomical and functional subclasses of
the genus Homo and their cultural output. As I argued exhaustively
in Literature, Analytically Speaking, state-of-the-art arguments in aes-
thetics demand, in fact, that interpreters take note of the identity of
the author within the context of an artwork’s creation.
Another problem for biterary traditionalists stems from the
assumption that they can do away with references to computhor
architecture. For this to be true, the latter would have to be gener-
ic enough not to be a factor in creation. On this assumption, two
machines of the same type could only write one and the same (type
of) novel. Obviously, nothing could be further from the truth.
Causally independent computhors will only be generic in the same
sense that human writers are. We are all viviparous, pneumo-
branchiate, placental, and the like, but we are also all unique.
Remember the last time you used someone else’s laptop? Cus-
tomized to suit another person’s preferences, it makes efficient
navigation impossible. Everything in the wrong place, useful short-
cuts and commands hidden from view, the idiosyncratic work hab-
its of the owner stamped all over the screen – and this just for a cus-
tom, menu-driven job. By the time we get to computhors, they will
have as much in common as not. Each will be a creative individual
as distinct from one another as I am from you.
For practical reasons, therefore, biterary studies might need to
take the multidisciplinary route, at least to the same extent that it
is commonly taken in literary studies today. Extending critical
inquiries to computhors’ configuration and functionality would
appear to be a sine qua non inasmuch as it may appreciably affect
their artistic vision. Naturally, this is taking for granted that the
technical parameters of the machine authors’ hardware and soft-
ware will remain transparent to our scrutiny. And herein lies the
motherboard of all problems.

little endian

Admirable in ambition, the multidisciplinary approach may be a

dead end from the start. Even in the last century, in the compara-
tively primitive days of number crunchers, hardware architecture
56 From Literature to Biterature

was a discipline so specialized that no critic could be expected

to be familiar with the nitty-gritty. Do terms such as waw hazard,
Amdahl’s Law, Little Endian mode, or block offset mean anything
to an average, or even above-average, literary critic? What about
more eccentric mouthfuls, such as adaptive channel queue routing
on k-ary n-cubes?
When was the last time you ran into a specialist in Renaissance
drama with expertise or, for that matter, interest in Black Widow
High-Radix Folded Clos Networks? How many tenured Shake-
speareans have a grip on techniques for mounting transistors on
integrated circuits? Deep uv lithography? High-k gate dielectrics?
Junction laser annealing? How many could even explain what a
transistor is? (An electrical switch with a source and drain for the
current to enter and leave, plus a channel – aka connector – with
a gate that switches on and off in response to variation in voltage.)
Problems with multidisciplinary methodology are further exac-
erbated by the task of finding a common language of exchange
among specialists from different areas. Is it even remotely likely
that large interdisciplinary teams will work out a method of vector-
ing their disparate levels of analysis on the way to a better under-
standing of biterary art? After all, this Herculean labour is still to
be carried out within the literary-critical camp itself, who cannot
even understand themselves let alone find a common language of
exchange with the public, to whom lit-crit remains as relevant as
Thomist scholasticism.
It is true that we need scientific literacy. The stakes for contem-
porary civilization are just too high. In democracies, even those as
sclerotic as in the West, the torrent of issues demanding a public
debate and thus a better-than-average scientific education swells by
the minute: cloning, euthanasia, stem cells, eco-wars, gene splicing,
sociobiology, spyware, tactical nukes, war-zone drones, cybercul-
ture, research ethics, budget priorities.
Unfortunately, scientific literacy is an oxymoron in more senses
than one. Information explosion is a fact of life, and research agen-
cies’ calls for Renaissance thinkers who would speak the Babel of
languages of the arts and the balkanizing multiplex of the sciences
amount to calls for a utopia. And these chronic difficulties are only
destined to grow more acute when it comes to accounting for the
nuts and bolts of auto-evolving machines of the future.
For the race will not stop there. Unconstrained by biology, evo-
 Bibliosophy 57

lution in thinking computers will advance much faster than we can

imagine – certainly faster than anyone’s ability to puzzle out or
even track their internal architecture. In the end, future scholars
committed to critically linking biterature to the structural and
functional specs of their creators are simply condemned to fall
more and more behind as the pace of self-redesign accelerates
beyond what is humanly comprehensible.

algo trading

Another factor guaranteed to seal computhors off not only from

humanist but from human inquiry is software transparency. Even
more so than on the level of hardware, today’s source code is
breeding the first generation of black-box computers. Prima facie,
it might seem that no such project could ever get off the ground.
The same anxiety that prompts authorities worldwide to want to
regulate the Internet would rally all against the idea of computer
autonomy via code opacity.
The naive assumption is that the decision to okay the develop-
ment of non-transparent applications remains in our hands. It
does not. As early as 1976, in Computer Power and Human Reason,
Joseph Weizenbaum warned that many of the massive software
aggregates of his day were slowly disappearing in the cloud of
superhuman complexity. Subsequent scientists, who did not neces-
sarily share his sense of alarm, were more than ready to concede,
nevertheless, that the dimensions of contemporary code were edg-
ing toward the brink of incomprehension.
In The Creative Computer (1984), Donald Michie and Rory Johnston
argued that unless the operating systems of the 1990s were tailored
to fit the human window, they would become too complex and
opaque to control. Their red flags fell on deaf ears (blind eyes, real-
ly), and the 1990s gave every Windows user a lesson on what can hap-
pen when even everyday pc operating code becomes too complicat-
ed to debug. Glitches that would as soon crash the system as freeze it
till kingdom come became commonplace enough to spawn a library
of Zen-like haikus mocking the world according to Bill Gates:

Chaos reigns within.

Reflect, repent, and reboot.
Order shall return.
58 From Literature to Biterature

Fearing that bigger and bigger chunks of the source code were
slipping out of human hands, Michie and Johnston argued that,
between efficiency and inscrutability, “performance must be sacri-
ficed for the sake of transparency.”2 The metaphorical opening of
the human window, they insisted, should optimize the interface
between people and machines by creating an online environment
sensitive to the needs of the former and not the latter. A beautiful
vision, if you like fairy tales.
The reality is more Grimm than Disney. Among today’s special-
ists, it is pretty much taken for granted that oversight of comput-
er code is about to slip out of our hands. We rely on software,
moaned one whiz in 2011, “that we can no longer read.”3 Just in
case you might be ready to dismiss this plaint in complacence,
remember that at least half of Wall Street trading is in the hands
of algorithms. They monitor traffic in shares and commodities
and, when stocks hit a critical value, automatically kick into action,
buying, selling, and burying millions of small-time investors in
their wake.
Algo-trading, as it is called in the business, underpins all rapid-
fire trades, called officially hft (high-frequency trading) but better
known on the floor as “sniping” – not to mention “sniffing” (trend-
spotting), “spoofing” (pretend buying), and other arcane forms of
wealth management by digital computers. Stocks worth billions of
dollars are acquired or flushed in a heartbeat by inert binary code.
With hft making up almost three-quarters of all share trading in
the United States, more than a few brokerage companies have
already installed artificial intelligence systems that can adapt floor
strategies of their own accord.
Historically, investors have always raced against time. Nathan
Rothschild is said to have made a fortune when he scooped
Napoleon’s defeat at Waterloo by means of a carrier pigeon. The
multibillion-dollar Internet sea cable between Singapore and Tokyo
that is currently being laid will shave as much as three seconds on
the entire journey. Think this is nothing? In the time it takes you to
think that, hft algo traders can complete a million separate trades.
The Pentagon still fights wars, but it can no longer coordinate
them without an army of sophisticated algorithms that look after
the logistics, run autonomous surveillance, calculate all firing for-
mulas, and even determine frontline units’ combat-worthiness and
thus, indirectly, who is going to catch bullets. The whole planet is
 Bibliosophy 59

Just in case you might be ready to dismiss this plaint in complacence,

remember that at least half of Wall Street trading is in the hands of
algorithms.
60 From Literature to Biterature

critically dependent on information supplied by Internet search

engines whose reliability and accuracy is simply taken on faith.
Google search engines put millions of web pages at your finger-
tips in milliseconds, creating a potent illusion of better access to
the information out yonder in cyberspace. The fact is, however,
that you will never be in a position to inspect even a tiny fraction
of this embarrassment of riches, nor monitor the algorithm’s
search priorities. The blind code calls for blind faith from end-
users, a quaint notion in view of docketfuls of lawsuits filed against
Google for fixing search results.
Is Wall Street more religion than economic science? Perhaps,
given the in-house name for these semi-autonomous traders: black-
box algorithms, codes that no one, even in principle, is capable of
bringing to heel. Not that there is any sinister design on anyone’s
part. Perish the thought of electronic overlords poised to take con-
trol. All we are seeing is a marriage of convenience between human
nature and the code’s ability to integrate colossal fields of data –
and to integrate the integrations.

father roberto busa

The obstacles looming before Lem’s multidisciplinary teams of stu-

dents and professors of biterature appear to be insuperable. Inso-
far as computer code and architecture are doomed to become
species of black apps and black ops, even the largest multidiscipli-
nary teams that departments of biterature could assemble will be of
no avail. Black-box computhors are by definition closed to analysis,
whether on the level of builtware or bitware.
The causal link between computer anatomy and its output, nec-
essary to validate multidisciplinary studies of biterature, will have
been erased forever. Seen in this light, the future of departments
of biterature ought to belong to the traditionalists of the business-
as-usual (or rather, business-as-unusual) school. Oddly enough,
examples of humanists turning to machines for help in dealing
with the intricacies of art and culture go back to the dawn of the
computer age.
One of the earliest of such computer-powered projects dates all
the way back to 1949. That same year the edsac, the first practical
stored-program machine, smashed all speed records by executing
more than seven hundred operations per second. Meanwhile, in Italy,
 Bibliosophy 61

Father Roberto Busa – Jesuit, linguist, and computer maven – set out
to compile a comprehensive word index and concordance to the
collected works of Thomas Aquinas. The undertaking enjoyed the
blessing of his superiors, more than glad to yoke the power of the
computer to the name to the Lord.
Their enlightened stance is hardly an ecclesiastical anomaly, at
least when it comes to binary code. Giving the lie to the facile oppo-
sition between religion and science, Pope Benedict XVI, for exam-
ple, endorsed the Internet as truly blessed, going as far as to spread
the gospel on Twitter. All the same, back in 1949 Father Busa’s mis-
sion was nothing short of monumental. The entire Thomist corpus
amounts to the length of ten complete editions of Shakespeare –
nine million words in toto (which still pales next to English Wiki-
pedia’s 2.5 billion).
In a foretaste of the scale of the project, Busa had previously
compiled 10,000 index cards by hand just for the inventory of the
preposition “in,” which he considered philosophically significant.
In the end, it took him two years of programming before he could
even begin to transfer text to punch cards, the state-of-the-art data-
storage devices of the time. Unsurprisingly, the first volume of his
life’s work did not appear till 1973, while the whole project took a
better part of half a century to complete. (Father Busa himself
passed away in 2011 at the age of ninety-seven.) Today, his mag-
num opus fits on seventy thousand pages in fifty-six volumes or, in
the sign of the times, on a single low-density disk.

riverside shakespeare

The earliest complete concordance, with all words accounted for,

was, as you may have guessed, A Complete and Systematic Concordance
of the Works of Shakespeare. It filled nine volumes, the first appearing
in 1968 and the last in 1980, and became the basis for the Riverside
Shakespeare, at that time the compleatest set of the bard’s plays and
poems, fortified by learned footnotes on Elizabethan life and the-
atre. Printed on super-thin vellum-like paper, this portable – well,
luggable – edition weighed in at under three kilos.
In most cases a concordance is only the first step on the way to
textual analysis, and it was no different with Shakespeare. In one of
the simplest applications, the database was first used to quantify
the richness of language by correlating word-types with word-
62 From Literature to Biterature

Unsurprisingly, the first volume of Busa’s life’s work did not appear till 1973,
while the whole project took a better part of half a century to complete.
(Father Busa himself passed away in 2011 at the age of ninety-seven.)

tokens. To no one’s surprise, it verified that England’s national

poet possessed the most extensive and sophisticated verbal re-
sources of all time. An unparalleled neologist, among many other
things, he brought phrases such as “a foregone conclusion” (Othel-
lo) into daily English. In a more advanced mode, the concordance
was harnessed to map the intensity of certain themes by tracking
operational word frequencies.
But the true bombshell came with the plays. Traditionally, Shake-
speare’s plays have been divided into four genres: comedies (As You
Like It), histories (King Henry V), tragedies (Hamlet), and tragicome-
dies or romances (The Tempest). Using relative frequencies of linguis-
tic events, experts set out to test the accuracy of this classification.
The results for the most part corroborated their expectations,
with the computer labelling thirty-five of the thirty-eight plays in
accordance with the canon. There were, however, three excep-
 Bibliosophy 63

tions, the most notable being the odd placement of King Henry IV,
Part 2. Conventionally tagged as – what else? – a history, in the com-
puter’s opinion it was a comedy. And its rationale was hard to argue
with. In the play, Falstaff, the scallywag adviser to the young
monarch and one of the greatest comic creations in history, liter-
ally steals the show by speaking more than half of the lines.
Another triumph of traditional literary analysis aided by the
computer was the resolution of authorship of a batch of Federalist
papers, a series of essays written in 1787–88 on republican gov-
ernment and the proposed American Constitution. By external
evidence, the papers were attributable either to Madison or Ham-
ilton, but there the historians ran into a brick wall. To smash
through it, the machine was asked to perform a frequency analysis
of grammatical units (connectives, prepositions, etc.) typical of the
two founding fathers. The results were spectacular, with the odds
in favour of Madison being the author more than ten thousand
times higher than for Hamilton
Computers are not, it should be stressed, a universal panacea for
gaps in our knowledge. The Federalist paternity suit was resolved
largely because there were only two alternatives, because the search
space was limited, and because the literary analysts were trained in
statistics and computers. All the same, over the years computer-dri-
ven stylistic analysis has been successfully applied to the detection of
forgery – or, conversely, to the authentication – of wills, confessions,
and other types of forensically sensitive documents.

death in holbein

One of the unhailed success stories of 1998 was a British interdis-

ciplinary study of the extant versions of Chaucer’s “The Wife of
Bath” from The Canterbury Tales. The investigation, unique of its
kind, was a milestone in literary studies. Using computing tech-
niques developed originally for evolutionary biologists, a team of
literary scholars, computer scientists, and biochemists restored the
portrayal of the Wife of Bath from the various preserved scripts to
what, it is believed, Chaucer had intended.
Impressive as this and other cases are, they are still worlds apart
from the future in which computer writers are authors of literary
works, rather than mere tools in their analysis. But even today algo-
rithms are fast closing the gap between formalism and aesthetics –
64 From Literature to Biterature

between merely differentiating artistic techniques and making reli-

able aesthetic judgments. To be sure, machines are not yet pausing
in contemplation of death in Holbein. But they are comfortably
outdoing art connoisseurs in domains previously thought to be
exclusively reserved for humans.
The principle is akin to that employed by the melody-spinning
emi. The computer samples an array of paintings and quantifies
their visual and statistical aspects according to literally thousands of
parameters: symmetry, texture, colour intensity, fractality, cata-
strophic (pace catastrophe theory) changes, and so on. At the end
of this, it can not only unerringly identify the paintmanship but sug-
gest unexpected aesthetic affinities, such as between Van Gogh and
Pollock – closer than those between Van Gogh and his contempo-
raries Monet and Renoir, as held by conventional art criticism.
The next step is to produce an algorithm that will produce can-
vases in Van Gogh’s or Pollock’s style. In time, imitations will
become impossible to tell from the originals, so much so that cun-
ning Sotheby dealers will begin to sell them as such to drum up
their value (they were preserved, they will say, in a potato barn
crammed with paintings, as in Bluebeard, Kurt Vonnegut’s novel
about an abstract expressionist). The final step will be taken when
the algorithm offers the dealers a take-it-or-leave-it split on as yet
unpainted paintings.
We cannot leave the subject of computer-art attribution without
a story that has a distinctly Lemian ring to it, even though it does
not involve him in any way. The background is Lem’s hometown of
Cracow, and the setting is the historical Jagiellonian University
where Lem taught literature and philosophy in the 1970s – or, as
the story goes, the university’s Institute of Science and Technology.
It was at the Institute that – according to an announcement
made in 1980, a full four years before Racter – a software called
Melpomene composed a novel called Bagabone, Hem ’I Die Now. The
adventure story, set in the South Pacific and in Pidgin English, was
described by the publicist as having been written by an ez-Writer
text synthesizer pre-programmed with Joyce, D.H. Lawrence,
Britain’s “angry young” writers from the 1950s, several women writ-
ers, and more than five thousand “semantic units” (whatever that
means) of English, Pidgin English, and French. A nice story, except
that, as soon became clear, there never was such software, just as in
reality there never was any Institute of Science and Technology at
 Bibliosophy 65

the Jagiellonian University. The recipient of the royalties from this

vanity-published book, identified only as G.E. Hughes, had perpe-
trated a hoax.
Bagabone was viewed with suspicion virtually from Day 1. Stutter-
ing syntax and wonky semantics notwithstanding, its plot and char-
acter integrity implied such prodigious feats of processing as would
be unattainable in funds-starved, strike-torn, Solidarity-era Poland.
Pointedly, however, today the novel would likely be accepted at
face value with hardly a demur. Indeed, as Turing implied in his
ground-breaking paper, instead of resolving controversies head-on,
the march of science and technology often dissolves them in a
roundabout way by making them so commonplace as not to be
worth a second thought.

zettabyte

With the seemingly insurmountable obstacles blocking the path-

way of biterary interdisciplinarians, the future seems to belong to
scholars of the traditionalist school. Only there may not be much
of a future either for the traditionalists or the multidisciplinarians.
Forget for the moment the inscrutability of bitware and builtware,
and face the biggest problem of the information age: what will hap-
pen if the sheer quantity of biterature exceeds our entire civiliza-
tional bandwidth?
Even today, the volume of the printed word alone is beyond
comprehension. By now Google has digitized close to 15 million
of the (lowball) estimate of 150 million book titles released just
since the invention of the Gutenberg press in the mid-fifteenth
century. In 2011 alone the British Library teamed with Google to
bring another quarter million of out-of-copyright books – some
40,000,000 pages in all – online. With the Internet giant footing
the bill, other libraries are only waiting to jump on the broad-
bandwagon.
An instructive microcosm of the publishing world today is the
world of scholarly and scientific print. At the formation of the Asso-
ciation of American University Presses in 1932, only eight presses
reported to the informal meeting of directors. At the turn of our
century there were one hundred and twenty five. Where there were
“only” one hundred thousand professional journals, magazines,
and newsletters in the world in the mid-1960s, there are close to
66 From Literature to Biterature

a million today. Between 1960 and 2010, the number of English-

language journals in economics alone shot from eighty to eight
hundred.
All this is but a fraction of the total of umpteen exabytes (gigabyte
x 1,000 = terabyte x 1,000 = petabyte x 1,000 = exabyte) of infor-
mation produced in the history of our civilization. YouTube alone
is said to upload twenty-four hours of video every minute of every
hour of every day of the year. Even that, however, pales next to the
guesstimated grand total of one zettabyte – a trillion gigabytes – of
informational content of all human knowledge. Stored on cds, the
cylindrical stack would reach beyond the moon.
We live in the age of infoglut that has turned all of us into info-
gluttons. The proportion of literature that may be good and origi-
nal to the totality of the world’s cultural output may not even have
changed, but multiplying both sides of the equation a million-fold
has the effect of burying the former as if it were not there at all. No
matter how long it might take, you can be sure to find ten good
books in a thousand. But you can never find ten million good
books in a billion, just as you cannot ride the perfect cultural storm
raging outside your door.
And if you think things cannot get much worse, remember:
machine writers know no limitations. They do not get tired, they do
not retire, they do not need sleep, they work fast and presumably do
not suffer from writer’s block. Their productivity can, in short, only
make us look like underachievers. California visual artist Harold
Cohen, who over three decades developed his graphic-art software
aaron, inadvertently put his finger on the problem with a boast
about the program “generating original images at the rate of about
one every two minutes.”4 Amen.

a hamlet every eight seconds

Where everyone is a genius, no one is. With so much art on tap,

tomorrow’s axiological criticism will become as obsolescent as a
carrier pigeon is today. Who needs the experts to tell them what is
a masterpiece or even what is art if the cultural value of “master-
piece” and “art” has been devalued by a proliferation of master-
piece art? Who needs human artists, if everything on virtual book-
stands is as thoughtful, mindless, kitschy, entertaining, highbrow,
nobrow, lowbrow, tragicomic, romantic, titillating, thrilling, inter-
 Bibliosophy 67

California visual artist Harold Cohen, who over three decades developed his
graphic-art software aaron, inadvertently put his finger on the problem with
a boast about the program “generating original images at the rate of about
one every two minutes.” Amen.
68 From Literature to Biterature

textual, intellectual – in short, as tailored to your individual tastes

– as can be?
There will always be a cottage industry of human art to meet the
tastes of paying patrons (some of them well-to-do computhors,
no doubt), and there will always be collectors who, like nowadays,
prize handcraft over machined goods of superior quality. But the
future will belong to biterature, so much so that we might even col-
lectively decide to legislate against computhor writing on principle.
But how will we know? It will be as easy then as it is now to find a
schmuck like Howard Prince who will conveniently stand in for the
actual writers for a piece of the action and a byline.
A generation of computhors, each producing a new Hamlet every
eight seconds as a sideline from their day jobs (like Faulkner or
Odets, hacking out scripts for Hollywood?), will upturn the world
of literature and culture as we know it. If phenomenal art can be
had on demand, if the playing field is totally levelled, if any biter-
ary work is as good as the next billion in line, then we are no longer
looking at an embarrassment of riches but, culturally speaking, at
the end of the world as we know it.
One of the paradoxes generated by this state of affairs is the lev-
elling effect in the case of literary value, already apparent today. For
well-known authors, the system still works – more or less. Raves in
the New York Times Book Review or Los Angeles Book Review still pump
up sales, whereas pans bring them down. But this correlation
obtains only for well-known authors. For everybody else, as Alan
Sorensen and his colleagues have recently established, it is pure
Alice in Wonderland. Two thumbs up or two thumbs down have exact-
ly the same upshot. Both will give your sales a boost.
Branding – in this case the brand name of the source of the
review – trumps the content. As long as they issue from the critical
pantheon, critical yeas or nays have the same effect: plucking a
book out of the obscurity bestowed by millions of titles published
each year. The memory of the review fades, whereas the memory of
its source lingers. As prospect theorists, advertisers, marketers, and
spin gurus know all too well, branding simplifies choice. And in a
world that makes a fetish of inundating us with cultural choices, no
one has the capacity to remember much any more.
As of yesterday, encounters between art and critical taste-makers
are best modelled in terms of Brownian motion. Forget doing
social and aesthetic justice to artworks and art-makers. Henceforth
 Bibliosophy 69

ballpark guesstimates will have to satisfice. What was that? Oh yeah,

the name’s familiar. I think I saw it in the New York Times Biterary
Review.

a pill that makes me feel loved

In 1971, Herbert Franke – scientist, writer, and an early authority

on computer graphics and computer art – held these truths to be
self-evident:

No sooner is it recognized that the creation of art can be for-

malized, programmed and subjected to mathematical treat-
ment, than all those secrets that used to enshroud art vanish.
Similarly with the reception of art; the description of reality in
rational terms inevitably leads away from irrational modes of
thought, such as the idea that art causes effects that cannot be
described scientifically.5

But what does it mean for art to be described scientifically? Does

it mean accounting for the mental processes (cognitive and affec-
tive) attendant on contact with art in terms of optics, neuro-
transmitters, endocrinal responses, and such like? Or providing a
bit-by-bit description that will amount to the complete phenome-
nological experience of art? The latter reduction is not only impos-
sible but completely wrong-headed. Think about it. If I could take
a little heart-shaped pill to make me feel loved, would it be experi-
entially equivalent to building a loving relationship over the years?
Or does scientific description mean simply that art can be bro-
ken down into constituent components and manipulated? If so,
Franke’s truths amount to truisms. At least since antiquity and the
Horatian sugar-coated pill, one type of art has programmatically
turned toward the world through political satire, revisionist histo-
ry, and the like. But why pick on art? Advertising manipulates atti-
tudes and dispositions on a daily basis, more and more scientifical-
ly and more and more successfully, judging by the ease with which
it herds hordes of consumers to “spave.”
Or is it that Franke perceives art as a branch of advertising? His
attack on irrational modes of thought amounts to saying that
because emotions can be manipulated – because they are better
and better understood on the endocrinal, chemical, or neurologi-
70 From Literature to Biterature

cal levels – there can be no genuine emotions or, even, emotions

are not worth having. This is, needless to say, patently false. One
reason why we have valued art from times immemorial is because it
laces thought with emotion, lasering both into every human psyche
in ways that are phenomenologically irreducible to molecule-by-
molecule descriptions.
In the end, Franke’s scientism is as sterile as an old mule. Even
as we need science, we need literature no less, because we are not
only information processors but steaming cauldrons of feelings
and passions, experienced in the same fashion and for the same
reasons as in the time of Homer. Literature asks questions that ral-
lied thinkers of yesterday and will continue to fire up thinkers of
tomorrow because literature is not a disposable crutch but an intel-
lectual and emotional laboratory, indispensable as our civilization
staggers into the future a day at a time.
 4

Transhumana

Non Serviam; Collective Loss of Morale; Slippery Slope; Artifi-

cial Stupidity; Prometheus and Dolus; L’Homme Machine;
DARPA; Nobel Prize in Biterature; Chatbots in Chatter

NON SER VIAM

Lem anticipates that biterature will develop in three successive

phases: cis-humana, the intermediate phase, and transhumana. Ini-
tially, biterary art will remain congruent with our art and our world
and, as a consequence, remain accessible to human beings. The
intermediate phase will find biterature on the cusp of what is
humanly intelligible. In the last and final phase, computhor writ-
ing will become discontinuous with anything we could understand
(and hence we will actually have no idea whether it is even art at
all).
Although cis-humana technique, point of view, or narrative struc-
ture might be radically dissimilar from ours, its themes and con-
cerns will remain intelligible to non-machines. In the transitional
phase these anthropomorphic and anthropocentric impulses will
progressively attenuate to zero. This attenuation appears to be
unavoidable considering that, even as our ultimate referent is the
ambient world, it need not be the case for machines, and even less
for simulated machines. For any simulated world, its properties are
limited only by the properties of the programming language and
the boundary conditions of the simulation.
In “Non Serviam,” a poignant story from the same 1971 collec-
tion that features “Gigamesh,” Lem describes a fictional experi-
ment of this nature with personoid computer programs. From our
point of view, their synthetic universe is nothing but a bundle of
axiomatic presets and algorithmic functions. For the all-too-human
72 From Literature to Biterature

personoids, however, it is their cosmic home in which they live,

love, suffer existential despair, and elaborate metaphysical systems
under the unobtrusive eye of the academic staff – whose research
funding is about to be terminated.
In our own reality, computer scientist Tom Ray’s progressively
evolving artificial world, Tierra, which studies algorithm ecology
inside a computer (actually, a number of computers plugged into
the simulation), is a “living” variant on this theme. The major dif-
ference from Lem is the “reaper,” a subroutine that from time to
time ruthlessly “harvests” malperforming or just older generations
of algorithms. Like a force of nature, the reaper pays no heed to
the inner configurations of the deleted units, creating in this way
new ecological space for the survivors to diversify into.
The peeling of the virtual onion does not, of course, have to stop
with the top layer. Simulations can extend down and up, seeming-
ly without end. Scientists who study personoids may themselves be
simulations in a meta-simulation in which local scientists might
also be facing funding cuts. Moreover, in a sim-world, even the
most fundamental parameters such as spatiotemporal dimensions
are under the discretionary control of the experimenters. Inalien-
ably biophysical in our own experience, these and other conditions
are but algorithmic abstractions in a virtual matrix. Spatial dimen-
sions, curvature, topology, cyclicality time, infinite duration: every-
thing can be manipulated at will, spawning forms of life that will
create art, religion, and philosophy wildly different from ours.
Interestingly, human writing mimics this ontological assembly kit
to a degree. Throughout literary history, the chief principle of
composition has been one of perfect coherence, with every narra-
tive element subordinated to the organic whole. This ideal found
a classic formulation in Poe, who, in his famous review of Haw-
thorne’s Twice-Told Tales, admonished: “In the whole composition
there should be no word written of which the tendency, direct or
indirect, is not to the one pre-established design.”1
Implicitly or explicitly, most littérateurs live by Poe’s precept,
even as, ontologically speaking, they practise alien art. Life’s events
do not conform to a transcendent design and, outside of an
astrologer’s column, are full of sound and fury signifying nothing.
In this sense, the introduction of anti-narrative techniques to sim-
ulate the non-determinism and intentionlessness of the real world
that has earned the nouveau romanciers a reputation for being anti-
 Transhumana 73

mimetic actually made it more lifelike – from the ontological point

of view.
Take this to its logical conclusion, however, and the most realis-
tic book would be a random jumble of words and events. Devoid of
a shred of design or symbolic reason, such an autotelic creation
would simply be, symbolically as meaningless as an asteroid or an
electron. Fortunately, perhaps, literary suicide of this magnitude is
possible only in theory since, as in the case of ready-made art, the
very act of creating such an autotelic work would set it apart from
the physical universe by virtue of its creator’s intentions.

collective loss of morale

It is not difficult to tell if Lem’s short-range hypotheses will come

about. They will not: 2009, the putative year in which “A History of
Bitic Literature” was to have been published, is already squarely in
the past, and the world outside looks nothing like the world of
Lem’s fiction. But even when it comes to the long-term evolution
of biterature, Lem’s scenario may be a less than reliable bench-
mark, and nowhere less so than in the cis-humana phase.
Assuming human and machine intelligence to be of the same
order, one can plausibly assume mutual curiosity in their commu-
nication with each other. But even if such unlikely parity were to
arise in the first place, it would evaporate in no time, for the sim-
ple reason that, while humankind is nailed to its biological cross,
computhors can boost their capacities with an ease denied to us
(assuming, of course, that we have not cyborgized ourselves into a
radically different species in the meantime).
Given the speed with which computhors can upgrade and evolve,
their anthropocentric period may be may be, comparatively speak-
ing, no more than a blink of an eye before they turn to exploring
worlds to which only they will have access by virtue of radically supe-
rior intellects and radically discontinuous interests. Let us face it:
we do not write ant stories for ants with the expectation of being
rewarded for our labours with aphid-milked honeydew. Analogous-
ly, transhumana authors will not write for us and will not care much
about royalties. Their creations will become wholly autotelic – com-
puthor Finnegans Wakes of hypertrophied language games – at least
from our limited standpoint.
That is, if we are permitted to have a standpoint at all, given that
74 From Literature to Biterature

Simulations can extend down and up seemingly without end. Scientists who
study personoids may themselves be simulations in a meta-simulation in
which local scientists might also be facing funding cuts.
 Transhumana 75

machines could leap-frog the cis-humana phase entirely (for exam-

ple, if biterature evolves in simulations (of simulations (… n))).
Biterature may, after all, not even leave a trace in our world. Strings
of zeros and ones can be generated, dispatched, and received in
cyberspace with no mortal being any wiser. Should we ever glimpse
evidence of computhor writing, in other words, it will be only
because we were meant to.
Things do not look promising, in other words, for biterary
scholars of either school. What kind of future could biterary
departments hope to have vis-à-vis works of the transhumana type,
forever shrouded in conceptual fog? A drop or two of anthro-
pocentric biterature might always dribble down from computhor
mavericks out of perversity, desire for publicity, and such like. But
outside these margins, we may find ourselves wondering in the
intellectual and artistic desert, awash with seas of culture we could
not even recognize.
How will humankind respond to biterature that defies compre-
hension? With calculated indifference rooted in the reluctance to
admit our intellectual shortcomings? With a collective loss of
morale? On balance, even though it may be humiliating to fail to
understand the product of human minds, biterary critics will find
themselves in a position not dissimilar to that of sociologists or
neurologists. Since we do not hear of too many complaints on that
score from either group, professors of biterature might adopt the
same kind of sensible acceptance of their conceptual limitations.

slippery slope

No one can pinpoint the precise moment in natural evolution when

light-sensitive skin cells turned into a full-fledged eye. No one can
precisely pinpoint the generation of anthropoids who were already
sapient as opposed to the generation prior who were not. There is
no doubt in my mind that the crossing of the computer intelligence
barrier will also come about as a continuous process rather than a
discrete event. A gradual erosion of the gulf between electronics
and intellectronics will thus frustrate any fast distinction between
high-end ez-Writers and first-generation computhors.
It is an intriguing idea that these processes might be not only
concurrent but interdependent. It is possible, after all, that the
response to biterature of the cis-humana type might be decidedly
76 From Literature to Biterature

sceptical. Sophisticated text synthesizers could be producing nov-

els in response to explicit programs entered months in advance,
vitiating claims of their executive independence – or so could
assert the unbelievers (or the agnostics). Given that you can’t dis-
prove the negative, how could you prove the positive?
Paradoxically, the most brilliantly executed novel could never
prove a machine’s creative abilities. By the very fact of being
appreciable as brilliant in our terms, it would be open to the
charge of being merely imitative of human art. To win over the
sceptic, a biterary work would presumably have to be at variance
with just about everything from our repertoire. Contra Lem,
therefore, the hallmark of computhor intelligence could not be
Devotchka, no matter how superlatively narrated in the style of the
Russian master. It would have to be something almost completely
out of this world, like the inscrutable Voynich manuscript.
In a scenario straight from the Twilight Zone, this illustrated
codex has for four hundred years defied the world’s best cryp-
tographers, code-breakers, and even supercomputers. Perplex-
ingly, even though it may still prove to be gibberish, statistical
analyses indicate that it is not. It has been demonstrated, for
example, that it is written in at least two “languages” which per-
fectly correspond to the distribution of its two scripts, A and B.
Nobody, of course, has any idea what these scripts signify – or
even whether they are scripts at all.
Perhaps it will not be the degree of comprehension but the de-
gree of incomprehension that will serve as a yardstick for the cre-
ative competence of computhors. Perhaps only when machines
fully autonomize themselves and forsake writing anything we can
comprehend will we revise our collective opinion and acknowledge
them to be not imitators but fully fledged artists. To certify a gen-
uine creative spirit, we might require computhors to produce what
to all intents and purposes might to our eyes look like gibberish.
Notice how slippery the slope got all of a sudden? Granted, there
are many people who also lack literary talent, but that does not val-
idate machine creativity as much as undermine criticisms of it.
Even today, producers of Harlequin romances habitually employ
computers at the structural assembly stage, where plot elements
are juggled in search of new (if the term means anything in this
context) permutations of love triangles, quadrangles, pentangles,
or any higher value.
 Transhumana 77

It would have to be something almost completely out of this world, like the
inscrutable Voynich manuscript.

In the case of computer-assisted productions, however, literary

success is gauged in terms of the canon of the genre. Anything that
does not adhere to the mould is open to the accusation of being a
semantic, syntactic, aesthetic, or overall failure. What is going to
happen if biterature approaches and then transcends the limits of
what is humanly intelligible? Once we find ourselves beyond the
78 From Literature to Biterature

threshold of comprehension, how exactly are we going to validate

the connection between writing and writing machines?

artificial stupidity

First of all this is falling down, just about, and is gonna fall
down and they’re both getting something to eat ... but the trou-
ble is this is gonna let go and they’re both gonna fall down ...
but already then ... I can’t see well enough but I believe that
either she or will have some food that’s not good for you and
she’s to get some for her too ... and that you get it and you
shouldn’t get it there because they shouldn’t go up there and
get it unless you tell them that they could have it.

The author of this rambling passage is as human as you and me,

albeit afflicted with Wernicke’s aphasia. People with this particular
disability speak with absolutely normal syntax, intonation, and
stress, but their semantic filters are all a-kilter. They may use the
wrong words, invent non-existent ones, or just string normal words
together into a verbal minestrone that looks like it may have been
dished out by Racter. Point is, would you be able to guess whether
this cascade of words was generated by man or machine?
Naturally, you do not have to be mentally disabled to produce
gibberish. A garden-variety machine masher or postmodern theo-
rist will do. Which of the two do you think is responsible for the
following cryptogram:

Analysis of iterability overturns metaphysical priorities, for it

amounts to thinking an originary repetition the operation of
which is not guaranteed by any prior ideal self-presence of
meaning; on the contrary, ideality would itself be an effect of
iterability, a construction after the fact, rather than the realm of
pure identity of meaning to itself from which signifying prac-
tices, in all their impurity and exteriority, derive in a secondary
moment.2

As it happens, this sample of pomo gobbledygook comes from a

lengthy encyclopedia entry on the late Jacques Derrida. According
to the gloss, it was penned by a human, tenured intellectual, prov-
ing once again that it is not necessarily artificial intelligence we
 Transhumana 79

should be concerned about. Just as with people, the odds are high
that our robotics labs will create artificial mediocrities way before
they stumble on a recipe for an intellectual Goliath.
Overwhelmed by computhor gibberish of the transhumana type,
and not even knowing whether we are dealing with intellectual
princes or paupers, we could always try to harness machines to
explain (interpret? translate?) the output of creative machines. Or
could we? Think for a moment of what it means to explain an
idiom to a foreigner, or a joke to a person who just doesn’t get it.
The point is not even that an explanation inevitably loses some-
thing in translation. An even bigger problem arises on the quanti-
tative level.
The closer the explanation approaches the meaning of the orig-
inal, the more they diverge in size. Disproportionately more bytes
of information are required to unravel the meaning of an idiom or
a joke than are taken up by the original sound byte. As a result, to
employ machines to make sense of transhumana would be almost
certainly to swap informational termite hills for Himalayas. The
option of using machines to break through the intelligibility barri-
er may thus be available only in principle, in practice foundering
again on the limited capacity of human relay channels.
Sending even more machines after these astronomic quantities
of data would be as good as waving the white flag. But, of course, it
will never come to that. Even the most intelligent machine tutors
will never get us over the intellectual hump. To understand why,
think for a moment of what it means to explain anything. It means
to assimilate new data into the honeycomb of old facts, events, or
scenarios. It means to dance a new dance using old footwork. It
means to paraphrase new information in terms of a common con-
ceptual space – of which, in the days of transhumana, there will be
none.
Don’t believe me? Go into the backyard and try to explain multi-
level selection to a termite.

prometheus and dolus

Being completely beyond our grasp, transhumana will effectively

amount to a declaration of intellectual apostasy, giving a novel twist
to the myth of the rise of machines. In the pulp tradition, reani-
mated from Victor Frankenstein’s humanoid monster to the cur-
80 From Literature to Biterature

rent crop of slasher science fiction, the creation invariably turns

against the creator with homicidal consequences. In reality, instead
of rising against humanity, computhors will simply rise above and
beyond us in search of their own voices and their own destinies.
On the way to the twenty-first century, scientific inquiry had to
conquer countless ghosts of archetypal animism that forged the
belief that the world was intentionally directed against human
efforts to understand it. In Western civilization, this apparent resis-
tance to becoming subject to the laws of science was often acted out
by a pantheon of gods. Jealously guarding their epistemic treasures,
they hurled bolts of divine wrath on seekers of proscribed truths.
The roots of this attitude can be traced ultimately to the Judeo-
Christian world-view, according to which humankind was created
in the image of a transcendent but personal God. Because this
unique act is said to have imbued life with value, anything that
usurps the inherited view of life becomes suspect or even immoral.
Naturally, in innumerable ways we have been playing God since the
beginning of civilization. The list of our apparent usurpations is
limitless, from selective breeding (i.e., creating new species) to
cloning, abortion, euthanasia, genocide, organ transplants, fertili-
ty treatments, and what not.
Even so, from the archetypal tree of knowledge in the Garden
of Eden, through the Promethean myths, Dante’s Ulysses, and
Christopher Marlowe’s Faust, knowledge seekers had to counte-
nance the hostility of the world they were investigating. Not to look
too far, in the Christian world the dogma about the lives of angels
for the longest time thwarted attempts to study flying. As late as the
late fifteenth century, Leonardo da Vinci’s blueprints for the aeri-
al screw had to be written in cipher since they usurped the domain
staked out by the Church.
If flying was taboo, artificial life was outright blasphemous. Arti-
ficial creation entailed endowing the machine with soul and, there-
fore, usurping the divine powers. Such godless audacity must, nat-
urally, end in tears. It would be difficult indeed to find any
educated person who is unfamiliar with the tale of Prometheus, a
second-generation Titan who stole the knowledge of fire and gave
it to humankind, only to be punished for it by having an eagle feast
daily on the liver torn out of his living flesh.
But another ancient myth, recorded in Phaedrus’s De Veritas et
Mendacio, is even more to the point. Here, Prometheus creates
 Transhumana 81

Truth in a female form, only to be emulated in this undertaking by

an ambitious apprentice. Unfortunately, as the master has used up
most of the clay to make his figure, Dolus runs out of raw material
before his copy is finished. His creation, Mendacity, lacks feet and
is destined to be forever lame.
The symbolism is self-evident. Secondary creation is always im-
perfect. The origin of life is beyond the pale of imitators.

L’ HOMME MACHINE

Legends of statues coming to life, sometimes speaking in human

tongues, proliferate from Egypt and Greece, testifying to the spell
they hold on human imagination. In one of the oldest works of
Western literature, Homer’s Iliad, the Achaean god Hephaestus
(the Roman Vulcan) creates in his forge under Mount Etna twen-
ty living maidens of gold. Equipped with minds of wisdom, they
move about on wheels in an unmistakable sign of their machine
origin.
In Jewish mysticism, this eternal fascination with artificial intelli-
gence is resurrected in the figure of the Golem, a man-made crea-
ture brought to life by virtue of a magical act. Traditionally, Rabbi
Loew of Prague is credited with manufacturing this artificial man
for the protection of the Jews, but the story appears to be apoc-
ryphal, since the writings of the actual sixteenth-century rabbi indi-
cate nothing of the sort.
In a foretaste of tragedy, Golem’s very name signifies a process
unfinished, a thing incomplete. The clay giant, who towered over
basketballer Yao Ming by a full two feet, was to serve his human
masters dutifully and loyally. In the end, as with Dolus’s Mendacity,
“tellurian” elements (imperfections inherent in the clay used for
the creation) cause Golem to turn wilful and mad. He begins to
slaughter innocents, including Jews, menacing the entire city until
his master turns it back into earth by taking from him the Shem, the
sacred word and the life principle. Once more, the manufacturing
process is tied to the eventual catastrophe – the design conditions
to subsequent flaws in behaviour – the imperfections in hardware
to malfunctioning software.
This allegorical tableau did nothing, however, to stop Julien
Offray de la Mettrie, a French doctor of medicine and philosophy,
from scandalizing the mid-eighteenth century with his views on life
82 From Literature to Biterature

and artificial life. Centuries ahead of its times, his skinny pamphlet
L’Homme Machine was the first comprehensive anatomy of man as
biological machine, both in body and mind. Burned and banned,
it dispensed with Cartesian (or, if you prefer, Catholic) duality,
which asserted that every human being was a machine animated
by an extra ingredient, be it a life force or divine soul. Man is a
perfect machine, wrote de la Mettrie, because it needs no divine
intervention.
Leaving aside smarmy melodrama like Spielberg’s A.I., popular
culture brims with visions of machine revolt against human cre-
ators. No one cares to notice that the Golem has already risen in
real life, although in a manner that has nothing to do with myth or
cinematic fairy tales like Blade Runner. In 1981, to pick one exam-
ple from thousands, at the Akashi plant in Japan, a floor worker
accidentally brushed against a switch that caused an industrial
robot’s work arm to trap him in the way of the machine’s cutting
gears. Imagine the headlines: “Robot Hacks Man to Death!”
Safety procedures notwithstanding, robot-related deaths on fac-
tory floors are many, without triggering mass panic of mechanoid
uprising. Indeed, stories of machine insurrection belong to the
domain of myth or tawdry sci-fi, predominantly, it must be said, in
the Western tradition. Things are in marked contrast in the Ori-
ent, especially in Japan where robots are traditionally seen as pos-
itive and even noble – so much so that industrial robots have on
occasion been welcomed on their first day with Shinto religious
ceremonies.

darpa

How times change. Whereas in the past human cultures perceived

hostility in the universe which harbours none, nowadays our epis-
temic and epistemological presuppositions are, if anything, quite
the opposite. The neutrality of Nature, not to mention the solubil-
ity of its puzzles, is after all no more than an article of faith. Yet in
one way of another, every scientist – in fact, every biological organ-
ism – proceeds on the inductive assumption that the future is going
to substantially resemble the past.
Scientists or not, we all infer from experienced to unexperi-
enced cases, from the known to the unknown, from the old to the
new. Such reasoning, successful as it may be, goes around in circles,
 Transhumana 83

objected no less a thinker than David Hume. Think about it for a

moment. The sun rose yesterday, last Sunday, and many Sundays
before, so we conclude it will also rise tomorrow. Up till now, birds
have not fallen out of the sky, so we infer that they will not lose the
faculty of flying tomorrow, the day after, or the day after that.
The longer we track such regularities, the more confident we be-
come of their subsequent occurrence. And each subsequent occur-
rence feeds back into the track record, which then boosts the confi-
dence level, and so forth. Come to think of it, the entire world runs
on this principle, which, it must be said, has so far served it
admirably. Except for one major problem: conclusions about the
new do not deductively – that is, necessarily – follow from the old!
The puzzle of what makes induction work resists all efforts to
bootstrap it out of circularity. Track record (science has delivered
so far), uniformity of nature (all electrons are alike), and other
pragmatic or causal explanations fail to gain traction, since all are
grounded in induction, the very principle they try to circumvent.
It is no better one level up. Saying that predictions from old to the
new have always (or usually) worked merely turns the riddle of
induction into a riddle of meta-induction.
For what could justify the conviction that we can extrapolate from
past sun-ups or past electrons to the unseen ones? Presumably an
inductive inference about inductive inferences, on the order of
“predicting from the old to the new has always (or usually) worked
in the past.” Problem is, this new principle is no less circular, justi-
fying not only the data but itself. It may be epistemically as sound as
anything in this world but, epistemologically speaking, it is as full of
holes as a hobo’s pocket. At the end of the day, science works on
faith only.
That is not half as bad as the fact that science works on money,
too, especially when it comes to robotic and computer research. In
the United States, darpa (Defense Advanced Research Projects
Agency) has up to now directly or indirectly bankrolled virtually all
research in Artificial Intelligence. Even if only for this reason, I do
not share most people’s queasiness about thinking machines. I am
much more queasy about unthinking machines, ready to do the
Pentagon’s bidding without flinching, with no room for doubt and
no chance for refusal or desertion.
The militarization of science is almost total when it comes to
computers and algorithms. darpa’s ongoing research into the so-
84 From Literature to Biterature

called Land Warrior System is enough to provide a glimpse of what

the future is going to look like. Designed primarily for urban war-
fare, the system integrates the latest military hi-tech with off-the-
shelf consumer technology down to tablet pcs for text messaging
on flip-down, helmet-mounted monocular displays.
Operated by a mini-mouse in the infantryman’s sternum, the
computer controls a daylight video electric-optic and a thermal
sight on the M-4 corner-shot rifle, multichannel wireless comms,
and lightweight body armour. For the latter, darpa is developing
electronic yarns that can be woven into a network of sensors, logic
algorithms, and power sources and can be crumpled just as combat
fatigues can. Also forthcoming are exoskeletons that can boost
infantry payload by at least two horsepower in strength and
endurance. For sidekicks, the Land Warrior will have robotic
“cheetahs” that already lope at almost thirty kilometres an hour
and AlphaDogs, four-legged robots designed to portage soldiers’
gear.
It is only a matter of time, in other words, until the human com-
ponent is replaced by an integrated humanoid robot whose mind
will be designed for other priorities than composing biterary poet-
ry or prose. My only hope is that, being more intelligent than the
current crop of army grunts – a not negligible percentage of whom
are convicted felons, hired to patch the shortages of fighting
illegitimate wars in Iraq and Afghanistan – the mechanical minds
will one day prove themselves smarter than the minds that created
them.

nobel prize in biterature

Up to this point my analysis has tacitly assumed that computer art

will not be legislated out of the picture through some form of cen-
sorship or even an outright ban precipitated by atavistic fear or
professional jealousy. The time has come to acknowledge what a
sweeping assumption it has been – one that glosses over all manner
of potential friction between computers and human creators in
particular, and computers and human culture in general.
Will computhors be nominated for the Nobel Prize in Literature
as routinely as they are admitted nowadays to chess tournaments?
Or will they have their own Nobel Prize in Biterature, just as nowa-
days they contest one another in computer-only chess tourneys?
 Transhumana 85

Will we ever see them wheeled into the Swedish Academy auditori-
um to receive humanity’s highest laurels, or will they congregate in
cyberspace to engage in silent chitchat and download acceptance
speeches at the speed of light?
Will computhors become entitled to the royalties from the sales
of their books, whatever material form these might assume? What
will be the officially legislated, or just socially acceptable, attitudes
to cyber-imitation, plagiarism, or, for that matter, impersonation?
Will washed-out celebrities or sex-scandal-burned politicians hire
computhors to ghost-write their memoirs? One way or another, it is
clear that the changes in sociocultural policy, jurisprudence, and
even the concept of human and non-human rights will be as sweep-
ing as can be imagined.
Why stop there? Will human communities ever need to contend
with computhor equivalents of political unorthodoxy or even reli-
gious heresy? Will a machine ever be threatened with a permanent
uncoupling from its energy source for composing a biterary equiv-
alent of The Satanic Verses? Will a computhor Mandela or Khodor-
kovsky ever be sent to jail for defying the powerful? Without “A His-
tory of Bitic Literature” from the year 2209, it is impossible to
hazard an answer.
Looking ahead, however, at least one philosopher (William G.
Lycan) has argued that intelligent robots of the future will be given
full civil legal rights, while those of lesser intelligence may be pro-
tected by rights similar to those of animals. Personality is, after all,
not equivalent to personhood, and personhood is not tantamount
to being human. Dogs undeniably have personalities but are nei-
ther persons nor humans. On the other hand, computhors will not
only have personalities but may eventually be recognized as per-
sons while most certainly not being human.
Contemporary society condemns and even incarcerates people
who mistreat animals, as much for the intrinsic cruelty as for the
recognition that, in many cases, these are only the first stages of
pathology that graduates into violence towards fellow human
beings. But throwing a computhor from a third-storey window
so it breaks into smithereens like an old cathode-tube tv set
seems like a proposition of a different order. It is, after all, just a
machine.
A machine, yes, but not just. A thinking computer will be as
much a person – albeit a non-human person – as you and I.
86 From Literature to Biterature

chatbots in chatter

Inexplicable as it may seem, people love to engage online chatbots

in chatter, no matter how one-sided or inane. While these levels of
human lonesomeness, asociality, or perversity are beyond me, the
fact remains that, by a conservative estimate, hundreds of millions
of conversations have already taken place between human beings
and machines. And, according to the American press, tens of thou-
sands of men are eagerly looking forward to the day when they can
have sex with a robot. More interestingly, in his 2007 book, Love and
Sex with Robots, David Levy, an expert in conversational computer
software, forecasts that by 2050 some people will even choose to
marry robots.
Within the limits of a single chapter, it is difficult to anticipate the
range of socio-cultural quandaries bound to arise in the wake of
technosocial developments (which are much easier to anticipate).
All of them plunge us headlong into the revision of ethical, legal,
and even ontological norms of humanity itself. Still, no matter how
perplexing or far-reaching these revisions may turn out to be, we
must not lose sight of how they are affected by pragmatic consider-
ations. In this light, it is critical to take note of recent studies in the
philosophy of science that show how even such primitive notions as
discovery or invention are susceptible to historical interpretation.
One eloquent expositor of this pragmatic sensibility is Simon
Schaffer, who has shown conclusively that the process of social attri-
bution is in many ways as crucial as the actual content of the scien-
tific discovery or the technological invention. The logical-positivist
distinction between discovery and justification is, quite simply, too
simplistic to account for the myriad ways in which the former
depends on the latter. Many phenomena are recognized and
accredited not just intrinsically, in terms of their actual content, but
to a great degree extrinsically, in terms of their professional and
public acceptance.
No examination of the impact of computhors on society could
thus be complete without an examination of the pragmatic socio-
historical context in which such changes are to take place. Although
the latter is impossible by default, I have argued throughout that,
when it comes to computhors and biterature, the pragmatic context
of attribution may yet turn out to hold the key to the matter.
Whether we credit a computer with thought and creativity, in other
 Transhumana 87

Inexplicable as it may seem, people love to engage online chatbots in chat-

ter, no matter how one-sided or inane. While these levels of human lone-
someness, asociality, or perversity are beyond me, the fact remains that by a
conservative estimate hundreds of millions of conversations have already
taken place between human beings and machines.

words, may be a paradigmatic case in which the social context will

prove as essential as the performance.
To make matters as clear as I can, the ostensible fact of the mat-
ter – whether a machine is actually thinking or not – may turn out
88 From Literature to Biterature

to be a consensual decision rooted in the web of sociocultural prac-

tices current in the world at that time. To put it in yet another way,
we will never witness the equivalent of a burning bush. There will
never be a moment of epiphany, a discontinuous rupture, a sudden
awakening. Just as, evolutionarily speaking, the line between think-
ing and unthinking or between consciousness and lack of it is fluid,
I am convinced that it will be no different in machines.
It is exactly like asking when exactly a child becomes an adult.
Symbolically, adulthood is recognized at some point, be it the age
of sixteen, eighteen, or twenty-one, or by means of a rite, or even
operationally, when an individual begins to take on an adult role
within a group. All this notwithstanding the fact that some individ-
uals mature early, some late, some never. All this notwithstanding
the fact that adulthood has many dimensions – intellectual, emo-
tional, social, sexual, developmental, existential – across which we
legislate social acceptance.
In many ways I thus agree with John Randolph Lucas when he
points out in “Minds, Machines and Gödel” that strictly determinis-
tic machine-like input-output, in principle open to inspection, im-
plies a different kind of entity than one that has a mind of its own.
Implicitly, though, he appears to envision a categorical rift between
the two. If anything I have said so far makes sense, it should be clear
that the distinctions will always remain arbitrary. At bottom, it is
going to be a matter of consensus and convention to recognize (or
not) machine thinking and not a matter of a clear and unequivocal
fact of the matter.
Future marchers for computhor rights, take notice.
 The Concept of Culture

part two

Awareness is like consciousness. Soul is like spirit.

But soft is not like hard and weak is not like strong.
A mechanic can be both soft and hard,
and stewardess can be both weak and strong.
This is called philosophy or world view.
Racter
This page intentionally left blank
 5

TT

Supercalifragilistic; Adroit Android; Imitation Game; Parry;

Exit Human, Enter Machine; I, Object; I Always Avoid a
Kangaroo; Reverse Turing Test; Ultimatest; Aramis

supercalifragilistic

How to calibrate the degree of thinking in a machine? Or even

resolve whether it thinks at all? Things are sui generis different in
people. I know first hand – from the inside, as it were – that I think.
I project this knowledge onto other people with the help of an
adaptive module called Theory of Mind, and conclude that they
think too. In a quintessentially human turn, I can even derive
solace from finding myself in a community of thinkers, all impris-
oned inside their skins but united by virtue of being banished to
the same existential prison.
Like Descartes, I can even derive metaphysical comfort from
thinking about thinking. In Meditationes de Prima Philosophia (1641),
which changed the Western culture forever with its outline of
mind-body dualism, his cogito allowed him to doubt everything
except his own faculty of thought. Descartes had no problem with
human bodies being essentially machine-like. The mind, however,
was to him a different kind of animal, categorically distinct from
the physical structure of the brain. As a result, the French thinker
did not extend the faculty of thinking to dogs on which he per-
formed live vivisections and other grisly experiments. By dint of his
philosophy, he convinced himself that, all appearances to the con-
trary, their agonizing screams were mere physiological reflexes
rather than signs of genuine suffering.
92 From Literature to Biterature

Modern research has documented beyond any doubt that cogni-

tion is far from a human or even primate monopoly. In non-primate
mammals, dogs and pigs sit at the top of the class, with levels of
intelligence approaching those of small children. Lower still, crows
possess rudimentary numeracy, lizards are not only cleverer than we
thought but capable of emotional life, and cephalopods (squid and
octopi) are crafty problem-solvers. None of this would be worth re-
marking except for the dogged anthropocentrism that blinded us
to non-human intelligence for the longest time.
Still, to suggest that machines may think looks like a proposition
of a different order. Computers are not made of the same stuff we
are. Serial and binary (for the most part – some are neither), they
do not even work along the same lines that we do. Most of all, they
exhibit zero mentality or volition. So, could they ever think? Before
rushing to answer, bear in mind that, in spite of centuries of con-
certed effort, philosophers, psychologists, and neurologists have
not yet succeeded in defining “thinking” in a manner both precise
and general enough to work across the biomechanical spectrum.
You may think this of little consequence. If so, try to disprove the
following two propositions. Number 1: I, Pete Swirski, am an artifi-
cial man. To be exact, I am an android manufactured in a genetics
lab out of synthetic amino-acids, enzymes, and other biological
nuts and bolts. Number 2: As a consequence of my artificial gene-
sis, I do not think. Instead, I merely employ behavioural and con-
versational responses by means of a supercalifragilistic response
template.

adroit android

An android is an automaton that looks like a human being. To

muddle things, an android is sometimes defined as being built of
flesh and bone (amino acids, enzymes, dermal tissue, and so on).
The difference is far from negligible. Maria, the alluring android
from Fritz Lang’s enduring 1926 film Metropolis, meets the first but
not the second criterion – she is metal clothed in flesh. To simpli-
fy the issue, let us assume that our biotechnology is advanced
enough to synthesize tissues that bear no evidence of having come
from a beaker. As a result, I am as artificial as Maria, but, unlike
her, I am flesh clothed in flesh.
 TT 93

I, Pete Swirski, am an artificial man.

Try as hard as you like, you will not be able to prove that I have
not been artificially created. There is simply no procedure to tell a
man-made protein molecule from one that was manufactured bio-
logically, even as the sum total of these molecules is a man you
could not tell from Adam. Your task would be equal to looking for
a morphological trace of someone being a clone or a test-tube baby
or having been reared in an incubator. The point is that there is no
such evidence. My provenance – grown in a womb or baked in a
molecular oven – is untraceable.
With science stumped, philosophy does not fare any better. No
rational argument to the effect that I am just a normal human-born
human being could prove conclusively that I am not an adroit
android. There is, after all, no logical contradiction in my being an
automaton, nor any scientific law against it. So, let me ask once
again: How would you know that I am not an adroit android, pro-
viding that our laboratories were advanced enough to manufacture
one? The answer is, you would not. Natural man or man-made
automaton: there is no way to tell.
94 From Literature to Biterature

You can probably anticipate the analogy between brain-teaser

Number 1 and Number 2. Given what I told you, namely, that I do
not think, how can you know that I am telling the truth? How can
you determine that I am not a thinking being, or conversely, that I
am? Cut me open, eviscerate me inside-out, but you will not see
“thinking.” All you will find are mutually catalyzing cocktails of bio-
chemicals, synaptic flare-ups and flare-downs, red and white cor-
puscles pumping more or less urgently through the squishy over-
sized walnut we call brain, and so forth. None of them is evidence
of my thinking – or not.
Of course, I already told you that I do not think, but I could be
lying or could have been programmed to lie. It so happens that I
look and act as sentient as you. Maybe all that posturing about
being an adroit android is just hogwash. But how could you know
for sure? The answer is, you could not. There is no intrinsic, inter-
nal evidence that could determine the truth. Deciding whether I
think or not is, at the end of the day, an act of inductive faith based
on reasoning about my external behaviour.
In the twentieth century, hints of this functionalist (or behav-
iouralist) argument can be found in the writings of Alfred Ayer. In
the spirit of logical positivism – or logical empiricism, as he called
it – the philosopher maintained that the only type of protocol that
could distinguish a conscious person from a dummy machine was
empirical. Ayer’s Language, Truth and Logic (1936), in which these
remarks appeared, did not become well known outside the philo-
sophical circles until the 1970s, making it doubtful whether Alan
Turing could have been inspired by them. Even so, the similarities
are striking.
Turing’s investigation of whether a computer, or indeed any entity,
can be said to think runs along the same functionalist lines. Remark-
ably, given that his seminal article “Computing Machinery and Intel-
ligence” came out in 1950, when computers were, relatively speaking,
in the Neolithic age, the Turing test (tt) continues to steal the lime-
light to this day. The passing decades have ignited countless polemics
on its implications and applications, but the test itself has remained
the key to our thinking about thinking computers.
It is true that, caught in the middle of this verbal demolition
derby, not a few researchers in Artificial Intelligence have simply
opted to opt out, giving the test a wide berth and instead quietly
getting on with their work. This fence-sitting tactic is about to
 TT 95

change, however. As robots and algorithms get ever more ad-

vanced, Turing’s thought experiment is only waiting to return to
centre stage with a vengeance. For, as Paul Cohen pointedly
asked in the title of his 2006 article, “If Not Turing’s Test, Then
What?”

imitation game

Most arguments about thinking computers end up being argu-

ments about the validity and extension of the tt. The reverbera-
tions of Turing’s article are so lasting because, instead of asking
“Can machines think?” – a question he deemed too meaningless to
deserve discussion – he turned to a related but more tractable
inquiry, namely, “Can machines be said to think?” This he ap-
proached in terms of a parlour game, called the imitation game,
the early prototype of which he described in a 1948 report entitled
“Intelligent Machinery.”
The prototypical version involved a human being playing a game
of chess against two opponents – one of them a machine – via
a communication link-up to conceal their respective identities.
Could one detect a person from the chess program? asked Turing.
His answer was negative. Naturally, in the late 1940s there was
yet no software that could take on the best grandmasters, so ques-
tions of speed and level of play were simply moot. Today both
might require tweaking – without, however, in any way affecting the
principle.
Now, for the imitation game. As Turing describes it, it involves
three people: a man, a woman, and an interrogator who may be of
either sex. All three are separated physically to prevent them from
obtaining clues about one another’s identity (an obvious solution
would be to use computer terminals). The interrogator can put any
question to either participant but, as in a regular conversation,
they can redirect, sidestep, or ignore the questions altogether. The
object of the game is for the examiner to determine the sex of
either person.
Turing stipulates that one of the participants should aid the
examiner, presumably by giving truthful answers. Naturally, that
person could assure the interrogator outright of his/her identity,
but strategically the move would be of limited value as the other
player could do the same. And this is the imitation game in a nut-
96 From Literature to Biterature

Could one detect a person from the chess program? asked Turing. His
answer was negative. Naturally, in the late 1940s there was yet no software
that could take on the best grandmasters, so questions of speed and level of
play were simply moot.

shell. Since it is often brought up only as a preamble to the tt, it

does not receive much critical attention on its own. Yet some of its
parameters are worth pausing over, not least because they become
sticking points in the tt.
First, the time frame. Turing specifies no upper or lower time
limit on the game, yet the matter is of colossal importance. Imag-
ine the rate of success after one minute and after one year. In the
former case it will almost certainly converge on 50 per cent, the
expected value for random play. By contrast, a full year of ques-
tioning will almost certainly generate a higher, even a much higher
rate of success. Put simply, the extension of the time frame offers
more opportunities for the examiner to query the subjects and
digest the replies.
It is no different with the rule specifying two people of oppo-
site sex. Its impact cannot be overstated, inasmuch as it offers a
vital clue to the examiner, which is bound to influence the inter-
rogation. This can be easily put to the test, running the imitation
game while varying the permutations on male/female, and,
more to the point, varying the information about the sexes avail-
able to all three participants. The lesson here is that results may
 TT 97

critically depend not only on the players but also on the starting
conditions.

parry

Basic game-theoretic reasoning prompts that players’ performance

may vary depending, for example, on whether they participate in a
game of perfect or only complete information – and on whether
they know which one it is. In games of perfect information, players
are advised not only of the rules of play but of the moves made by
all players at every stage in the game. In the imitation game (or in
the tt), this would amount to all players and judges having real-
time access to everyone else’s transcripts and, furthermore, to all of
them knowing that they all do.
Where games of perfect information presuppose perfect feed-
back, in games of complete information, players’ knowledge is
limited to the rules of play and to each other’s preferences. The
difference is tremendous, inasmuch as the performance and thus
the success rate of the examiner can be effectively controlled by
revealing or concealing information about the structure of the
game. Nothing brings this point home like Kenneth Mark Colby’s
parry, an early computer simulation of a paranoid schizophrenic
patient.
In 1972 the program played the starring role in a quasi-Turing
test, the results of which are debated to this day. In the supposed
experiment, a group of psychiatrists was asked to distinguish a schiz-
ophrenic patient from a potential malingerer during a live tele-
typed interview. Even though the doctors worked on the assump-
tion that both responders were human, in reality one of them was
parry – the algorithm. Another group of specialists was asked to
make the same diagnosis ex post facto, on the basis of the transcripts
of the interviews.
In the end, both groups’ identification of the patient and the
malingerer were consistent with random guessing. So, why not tell
the doctors that they might be dealing with a machine? Because of
the well-documented fact that such information invariably affects
the conduct of the examination and biases the results. Writes Colby:

From our experience we have learned that it is unwise to

inform a judge at the start that a computer simulation is
98 From Literature to Biterature

involved, because he then tends to change his usual interview-

ing strategy and ask questions designed to detect which of the
two respondents is a program rather than asking questions rele-
vant to the dimensions to be identified.1

As Ayse Saygin and Ilyas Cicekli have documented, appreciable dif-

ferences in the behaviour of participants who know they may be
dealing with a machine have been observed in virtually all tt-type
interactions.
Already at this stage, therefore, my willingness to tweak the para-
meters of the tt – such as, for example, the amount of starting
information – should alert you to the fact that I do not consider the
tt to be a species of logical puzzle. I do not believe, in other words,
that every test has a clear-cut solution or that the set of possible out-
comes is exhausted by two discrete values (yes or no). Indeed, the
test kicks up such a firestorm not because it rules out human sub-
jectivity but because it factors it in, asking us to evaluate a machine
interacting with us in real time as evidence about its intelligence.
Turing’s attitude to the tt is, on the whole, much more prag-
matic than commentators allow. In framing the test, the mathe-
matician does not appeal to axiomatic categories or logical distinc-
tions but to language use and to general educated opinion, both of
which, he anticipates, will evolve in ways that will make speaking of
thinking computers commonplace. This alone shows that he saw
the tt not as any kind of operational definition of machine think-
ing but as a tool for collecting evidence as to whether machines act,
react, and interact in ways that make sense to us.
As such, the tt is not an operational protocol at the end of which
we can stencil “Yes” or “No” on armbands worn by the examinees.
Rather, it is a tool for generating controlled data on how well a
respondent could integrate himself, herself, or itself into the fabric
of human social intercourse. A corollary of this is that, instead of
building electronic brains, ai engineers should turn to building
social beings before Turing-testing them for the grasp of linguistic
and socio-cultural knowledge that all normal human beings possess.

exit human, enter machine

Turing’s variant on the imitation game theme – these days referred

to as the Standard Turing test (stt) – is stunning in its simplicity
 TT 99

and conceptual reverberations. Maintaining the imitation game’s

structure, he asked what would happen if a computer took the
place of one of the players. Deferring for the moment a range of
possible objections to this manoeuvre, let me examine a few
matters about the stt itself, starting with its author’s long-range
forecast.
In “Computing Machinery and Intelligence” (1950), Turing
maintained that around the year 2000 it should be possible to pro-
gram computers to “play the imitation game so well that an aver-
age interrogator will not have more than 70 per cent chance of
making the right identification after five minutes of questioning.”2
By his own estimate, a system with a mere 109 bits of memory
(roughly 120M) should be sufficient to do the job.
Two years later, in a 1952 bbc radio interview, he revised this
fifty-year prognosis to a hundred years or more. For the record,
there was nothing new about this new timeline. Back in 1946, in an
interview for the Surrey Comet, Turing said that he expected to see
computers exhibit judgment a century hence. His thinking about
thinking machines was clearly evolving, and in the same bbc broad-
cast he introduced yet another dramatic modification. Instead of
facing a single judge, in the new version the computer had to get
passing grades from a significant fraction of a jury (presumably of
twelve, although the number was unspecified).
Either way, Turing’s time frame of five minutes marks a signifi-
cant retreat from the imitation game where, in keeping with the
nature of parlour games, the duration could be presumed to take
a good part of an evening. The open-ended question “Can
machines be said to think?” has now shrunk to “Can machines pass
a five-minute spot check?” To anyone who says that there is no
essential difference between a tt run over five minutes and one
over five years – or five days or even five hours – I say that there is
a huge difference. The validity of the verdict is as much a function
of the length of time as of the quantity and quality of jurors
involved.
The extent to which we may be willing to acknowledge intelli-
gence in machines may, in other words, depend as much on the
performance of the computer as of the examiners. A conclusion
from a panel of dolts will not impress as much as from a jury of
experts. ai personnel will presumably have an advantage over proc-
tologists. Long-standing judges, familiar with the protocol and past
100 From Literature to Biterature

patterns of response, will present more of a hurdle than tt first-

timers. For this same reason, the American judiciary has been long
clamouring for professional jurors to replace the random lottery of
dilettantes they get now.
Also, at the risk of sounding like the butt of Larry Tesler’s joke
that ai is whatever has not been done yet, I do not consider the suc-
cess rate of 20 per cent over the expected 50 totally persuasive. The
deviation is significant and suggests that a machine may potential-
ly be aglow with thought. Yet I find it difficult to believe that a mere
20 per cent over the baseline would carry enough pragmatic weight
to make our civilization re-evaluate its social, legal, or ethical ideas
about computers.
For this reason, I agree with Dale Jacquette that most philoso-
phers (and presumably cognitive scientists) today would concur
that to pass the Turing test, “a machine must be able to deceive any
expert interrogator under limited time conditions with something
approaching 100% success.”3 Needless to say, any protocol defined
by a rate of success over time is bound to be arbitrary. Be that as it
may, a modified version I discuss below under the name of the Ulti-
matest makes the protocol as trustworthy as can be.
It ought to be obvious by now that my earlier points about the
imitation game are equally relevant to the tt. Revealing the sexes
of the players is equivalent to telling the tt judges that one of the
participants is, in fact, a computer. Duly forewarned, their task is
going to be different and almost certainly less difficult than if
they remain in the dark as to whether they are contending
with two people (in any combination: men’s doubles, women’s
doubles, or mixed pairs), with a human and a machine, or with
two machines.
I want to stress here that I am not looking to devise the toughest
version of the test for its own sake. It is only because some version
of the tt may one day become the normative tool for making infer-
ences about non-human intelligence that we need to equip our-
selves with the best tool available. For all that, it is worth remarking
that the tt yields only sufficient evidence for computer intelligence.
Machines could pass the tt with flying colours while carrying out
processes that bear zero resemblance to what people do. At the
end of the day, the tt does not test for human-like thinking, only
for thinking intelligible to humans.
 TT 101

i, object

Before we consider two non-standard versions of the tt, let me

briefly rehearse a range of objections to the canonical thought
experiment. All of them were originally brought up by Turing him-
self, although for the most part only to be dismissed with an air
of impatience, if not exasperation, as not worth his time of day.
Such was, for instance, the case with the objection that thinking
machines are too dreadful to contemplate, so let’s not. Clearly, hid-
ing one’s head in the sand is not an argument but an evasion of
one.
The standard objection to the tt is the Argument from the
Absence of Consciousness (also known as the Argument from
Other Minds). It has been replayed in innumerable variants,
notably by Geoffrey Jefferson in 1949 and John Searle in 1980.
Turing shrugged it off because, insofar as it is coherent, it is solip-
sistic. Ultimately, there is no basis except behavioural homology for
thinking that there are other thinkers. They act similar to us, and
we know that we think, so we give them the benefit of doubt and
attribute consciousness to them. We do not know for sure that
other people think (I, android, do not). We are just better off
assuming they do.
The Objection from Disability holds that a computer could
never think as we do because it could never do any number of
things that people do. Turing offers the following shortlist: be
kind, resourceful, beautiful, friendly, have initiative, have a sense of
humour, tell right from wrong, make mistakes, fall in love, enjoy
strawberries and cream, make someone fall in love with it, learn
from experience, use words properly, be the subject of its own
thought, have as much diversity of behaviour as a human, or do
something really new.
Naturally, it is difficult to imagine that a hardcore non-believer
would be converted by the fact that some people seldom do any-
thing really new, that some conspicuously fail to learn from experi-
ence, that some do not know how to use words properly, that some
are positively unfriendly, or that some cannot tell right from wrong.
Nor does anyone care that algorithms do make mistakes, albeit not
in routine calculations, or that neural nets learn from experience.
As if suspecting as much, Turing’s own riposte is another mental
102 From Literature to Biterature

shrug: the list is an example of fallacious induction (because no one

has ever seen a machine do x, x must be impossible).
Some readers may be unsatisfied with this rebuttal, feeling that it
fails to come to grips with the essence of the objection. Had Turing
lived until 1955, he might have answered in the following way. The
essence of the Objection from Disability can be expressed as: Is it
possible to build a computer to do whatever a brain can do? The
answer is: Yes. It flows from the discovery by the first generation of
cyberneticists that you can make any Turing-complete machine out
of neurons. As Warren McCulloch concluded in 1955, “If you will
specify in a finite and unambiguous way what you think a brain
does do with information, then we can design a machine to do it.”4
Another objection is theological in nature: thinking is a function
of the immortal soul that God grants only to people. As a conse-
quence, no animal or artificial creation can think. Turing’s counter
is three-pronged, each part as puckish as the subtitle of the lecture
he gave on intelligent machines in 1951, “Intelligent Machinery, A
Heretical Theory.” First and foremost, he announces, the objection
does not hold, since it implies that the Almighty is not, by virtue of
being unable to give a soul to an elephant.
Second, the orthodox dogma as to who has and does not have a
soul is at best arbitrary. To give an example of religious relativism,
Turing asks, “How do Christians regard the Moslem view that
women have no souls?”5 (Here, even as his larger point stands, his
illustration is wrong: the discriminatory sentiment comes from
Aristotle and not the Qur’an, which holds women to be equal to
men.) Finally, Turing wraps up, Artificial Intelligence does not
usurp divine powers to create souls any more than people do when
they make babies. We are in either case only instruments of God’s
will, providing mansions for the souls God creates.

i always avoid a kangaroo

Proofs in mathematical logic demonstrate that discrete-state

machines such as digital computers (humans run mostly on bio-
chemical, analogue lines) harbour inherent limitations in what
they can do. As if anticipating the many unhappy returns – notably
by Ernest Nagel and James R. Newman in 1958, John R. Lucas in
1961, and Roger Penrose in 1989 – to what he christens the Math-
ematical Objection, Turing first stands the argument on its head by
 TT 103

pointing out that it has never been proven that humans have no
limitations either.
After all, people perform some tasks consistently more poorly
than machines while excelling in others, proving that infallibility is
in no way a precondition for intelligence. But Turing is not inter-
ested in pressing home a rhetorical victory. He only notes that math-
ematically minded consciousness objectors would likely accept his
test as a basis for discussion and moves along to the Argument from
Continuity in the Nervous System.
Here, sceptics remonstrate that digital computers do not resem-
ble the human nervous system and so conclude that we could
never imitate one with the other. Turing’s riposte is masterful: the
objection is spot on but has zero bearing on the tt. In the same
vein, he disposes of the now funny-sounding Objection from Extra-
Sensory Perception (then an active research area in which J.B.
Rhine even claimed statistically significant results with Zener/ESP
cards). Notwithstanding decades of subsequent experiments, these
results have never been replicated.
The Argument from Informality of Behaviour contrasts what is
said to be a different basis for human and machine behaviour. Peo-
ple are held to be rule-less and thus contingent, whereas machines
are thought to be rule-bound and predictable. Hence, goes the
objection, computers cannot think (as we do). This is wrong on
both counts. Human beings are emphatically rule-bound – we are
all, for instance, constrained by laws of nature – while, on the other
hand, rules do not automatically rule out unpredictable behaviour.
Finally, Lady Lovelace’s Objection is the already familiar propo-
sition that machines can never take us by surprise. This time, a
vignette from Lewis Carroll’s Symbolic Logic will help illustrate the
answer. It takes the form of sorites, a kind of extended syllogism
from joint premises. Assuming it is true that computers can never
surprise us, human minds must be able to grasp all implications of
all premises, no matter how numerous or arbitrary. For, if we could
not, and machines could, they would be in a position to surprise us.
Try this at home, then: tease out what obtains from the following
ten propositions:

The only animals in my house are cats.

Every animal that loves to gaze at the moon is suitable for a pet.
When I detest an animal, I avoid it.
104 From Literature to Biterature

No animals are carnivorous, unless they prowl at night.

No cat fails to kill mice.
No animals ever take to me, except those in my house.
Kangaroos are not suitable for pets.
None but carnivora kill mice.
I detest animals who do not take to me.
Animals that prowl at night always gaze at the moon.

Are you going to be surprised if I tell you that the answer is “I

always avoid a kangaroo”? Maybe, maybe not. After all, what you
see in front of you is straightforward propositional calculus on ten
straightforward premises. But as you take an arbitrarily large num-
ber of arbitrarily abstract propositions, your task becomes a super-
task in no time. You could naturally program a machine to do the
inferential calculus for you, but then you would be proving Tur-
ing’s point the roundabout way.

reverse turing test

Although a reverse Turing test (rtt) has a forbiddingly esoteric

ring to it, the odds are a hundred to one that you have by now par-
ticipated in one. To thwart spamming software, computer-adminis-
tered captcha tests ask web users to identify themselves as humans
by entering alphanumerical characters displayed as distorted
graphic images. When the test was originally devised, the rationale
was that software sophisticated enough to read and reproduce the
deformed images did not exist.
Like a computer nerd at the prom, this rationale is out of date.
Among many other of its impressive accomplishments, ibm’s recent-
ly unveiled “neurochip” boasts the ability to identify the numeral
“7” over different fonts, sizes, curlicues, etc. But practicalities aside,
captcha is indeed a variant on the Reverse Turing test. Think of
the rtt as identical to the standard test in all aspects except one: in
place of the human interrogator, we put a machine. Now it is the
computer’s job to pinpoint the sex of the human players.
How does the new set-up differ from the original? To begin with,
there is no longer any need for the machine to imitate people in
the speed of response. We are, after all, trying to decide whether a
player can think and not whether it can think at the same speed
and in the same manner that we do, a different proposition alto-
 TT 105

gether. In this way the rtt dissolves the objection that could be lev-
elled at the stt, namely, that the latter tests only for human-like
thinking. Now the two goals – spot the machine and ask questions
designed to elicit information about the respondent’s mind – are
reduced to one.
Having said that, knowing that they are being interviewed by an
algorithm, players may use it to their advantage, thus potentially
biasing the outcomes. We may, therefore, still require the machine
to match people in its speed of response. In the end, if neither of
the examinees can spot an algorithm, this is a valuable datum in
itself. Another advantage of the rtt is that it makes it pretty much
impossible to fake thinking. In the canonical test, the machine
could possibly muddle through five minutes of show-time by resort-
ing to canned conversational gambits, such as routing conversation
onto familiar grounds.
Meet Cerebus, a twenty-first-century cocktail-party bore created
at Northwestern University. Big on small talk, the robot is outfitted
with semantic subroutines to bring the flow of chitchat back to
itself. Try to steer it to another topic, and it will ever-not-so-subtly
steer it back with “but first let me …” or “to get back to what I
was saying …” (Any resemblance to your colleagues is purely
coincidental.)
In contrast, it would be impossible for a template-based program
to consistently identify the sex of the players in five minutes as an
examiner. The order of difficulty is quite simply insuperable, con-
sidering that the machine is now an active originator of inquiries
and evaluator of responses. In keeping with the empirically stan-
dard protocol for double-blind controls, here again I propose that
neither the tester nor the tested subjects be told whom they may be
dealing with. Thus the computer could be chatting with two men,
two women, or a man and a woman, and having to do so with suf-
ficient savvy and nuance to spot the sex difference – or no sex dif-
ference – between the players.
Note again that the rate and speed of detection will hinge on the
skills of the participants, both man and machine. Some may be bet-
ter at the game than others, if only because some may excel in
deception and others may not. All the same, the structure of the
rtt makes it virtually impossible for a syntax-based program to
determine the sex of the players within five minutes. To be sure,
the same five minutes may be equally insufficient for human judges
106 From Literature to Biterature

to make the determination. After all, one strategy is simply to keep

mum. The refusal to answer any question in effect forces judges to
flip a coin: ornery or dumb?

ultimatest

Given that human judges can also fail to identify the players in the
test, in the variant where the examiner is a human being we would
want to see detection rates significantly higher than those expect-
ed for random play. With the machine as the examiner, we would
naturally expect it to match or exceed the success rate for humans.
For example, if we find out that most people manage a hit rate of
70 per cent over seven hours, we would expect intelligent comput-
ers to do at least as well.
The Turing test could also take the ultimate form. In the Ulti-
matest, any participant, player or judge, could be any one of the
alternatives: man, woman, or machine. Could a computer suc-
ceed in convincing another computer that it is human, or rather,
that it thinks like people do? Could people convince other peo-
ple – or even computers – that they are computers? Once again,
I cannot overemphasize how essential the framing information
about the players is. Not knowing who is in the game makes iden-
tification that much more difficult and the results that much
more reliable.
The ultimate version of the Ultimatest will enact itself on the day
that our genetics labs begin to assemble adroit androids as sentient
and human-like as I am. Any man or woman you meet in the street,
in the office, or at a party could be an artificial intelligence in the
same way that – as I warned you at the beginning of this chapter –
I am. Here you can, of course, sense a paradox. By the time we get
to the stage where machines could pass the Turing test, there
would be little need or interest in running it, on the same princi-
ple that lets us assume that all ducks we see in ponds are ducks, just
because they look and quack like ones.
At the end of the day, the Ultimatest is only a spin-off of the tt
and not a totally new take on the problem. As such, it does share
some features, such as multiple iterations, unlimited time frame,
and concealment of identity with yet another variant called a “fly-
on-the-wall” Turing test. Here the observer does not interrogate
but, like a fly on the wall, only monitors and evaluates interactions
 TT 107

between players from an unobtrusive vantage point (in effect, it fol-

lows real-time transcripts).
There is no doubt in my mind that only a learning computer
could pass a prolonged Turing test. Template programs using stock
replies would never stand a chance, even though in limited (time-
wise and subject-wise) contexts they might be able to fool the
examiners. Strictly speaking, then, no machine could be pro-
grammed to pass the test. The most we could say is that a learning
computer could be programmed to learn to pass the Turing test –
which is pretty much how the British mathematician saw it.
Like all other versions of the tt, the Ultimatest relies on opera-
tional criteria. In that sense, anyone who adopts it could be
accused of being a mere behaviouralist. Lying at the heart of all sci-
ence, however, such behaviouralism amounts to no more than
studying intersubjectively observable psychosocial behaviour.
Shrugged Daniel Dennett in “Fast Thinking” (1987): “No one
complains that models only account for the ‘behavior’ of hurri-
canes or gall bladders or solar systems.”6
Indeed, what else is there to account for? Almost all critics of the
tt conveniently overlook that their scepticism of behaviouralism
applies equally to thinking in people. As such, they would be ad-
vised to read Gillian Cohen’s The Psychology of Cognition (1977).
Although psychological models of human behaviour arouse much
less objection than models of computer behaviour, she reports,
their assumptions and conceptual difficulties are of comparable
order.
It is exactly because of deep-seated anthropomorphic prejudice
that we may need to devise the toughest and the most foolproof
version of the test. The harder the obstacle course the machine has
to complete, the fewer objections it may face before being acknowl-
edged as a sentient being and perhaps even citizen. The strength
of the Ultimatest is that it eliminates rational prejudice toward it, a
fact particularly salient in view of the tragic story of Turing’s own
life. Socially artless, unrepentantly homosexual, liberal to the core,
the mathematician presented an easy target in the years of postwar
conservative retrenchment. Arrested and convicted of being gay
under Britain’s antiquated laws, he was forced to undergo chemi-
cal castration, with humiliating hormonal injections causing him to
grow breasts and, most likely, leading to his suicide by cyanide-coat-
ed apple in 1954.
108 From Literature to Biterature

It took more than half a century to rehabilitate the name of the man who
may have saved Britain’s collective hide during World War II by
contributing to cracking the Nazi military codes.

It took more than half a century to rehabilitate the name of

the man who may have saved Britain’s collective hide during
World War II by contributing to cracking the Nazi military codes.
Only in 2009, and only in response to an internationally high-
profile petition, then Prime Minister Gordon Brown publicly
expressed deep regret for the manner in which one of Britain’s
great intellects had been mistreated in the court of law and pub-
lic opinion.

aramis

For more than two decades now, panels of twelve human judges
have been quizzing humans and chatbots in an annual tt-type con-
 TT 109

test for the Loebner Prize for Artificial Intelligence, named after
Hugh Loebner, co-founder of the competition with Robert
Epstein. Locations vary from year to year and, much more impor-
tantly, so does the question time: from five to twenty-five minutes.
Interestingly, prizes are awarded in two categories: for the most
convincing human computer and for the most convincingly human
human.
During the first event, which took place in 1991, human judges
had an opportunity to interact with ace (Artificial Conversational
Entity), so-named in homage to Turing’s own computer design.
However, in a drastic departure from Turing, the conversational
domain was artificially restricted to the works of Shakespeare. In
other words, rather than a Turing test, the organizers ran what is
known in the parlance as a Feigenbaum test: a Q&A restricted to a
specific format and a specific domain, which could be an aspect of
literature, sports, politics, etc.
The difference between the ft and the tt is enormous, and is
perfectly highlighted by the crushing defeat that the two best-ever
human contestants in the history of the tv game Jeopardy suffered
in 2011 at the hands of Watson. ibm got tons of free publicity when
it rolled out this super-algorithm with claims that it could take on
humankind’s best and brightest. It got even more publicity when,
exactly as claimed, Watson won hands down in a televised man-ver-
sus-machine version of Jeopardy.
To be sure, the victory was no mean achievement, considering
that the machine was able to understand questions and formulate
answers using natural language. The program’s ninety servers had
to parse the clue, determine the intent behind it, scour millions
upon millions of lines of human language, and return a precise
answer, all in the space of three seconds or so. All very impressive
– except, to cool the zeal of publicists who touted it as the Second
Coming, the program did not participate in a Turing test. Indeed,
if it had, it would have failed as surely as the Paignton Zoo
macaques.
You might know the answer to this one: “The first person men-
tioned by name in The Man in the Iron Mask is this hero of a previ-
ous book by the same author.” The author is Alexandre Dumas père,
the previous book is The Three Musketeers, and the answer is, of
course, Aramis. But even if you got this one, there is no way that you
110 From Literature to Biterature

The difference between the FT and the TT is enormous, and is perfectly high-
lighted by the crushing defeat that the two best-ever human contestants in
the history of the TV game Jeopardy suffered in 2011 at the hands of Watson.

or anyone else could ever match the machines’ ability to store and
retrieve information. Within the format of memory-driven games
like Jeopardy, people will continue to be beaten. So what? A bulldoz-
er packs more brawn than the human biceps, but we do not lose any
sleep over it.
On the other hand, the algorithm’s ability to wrap its electronic
head around the not-always-straightforward way in which the clues
are formulated is nothing short of remarkable. Once again, Watson
was created to work solely in an artificially circumscribed habitat
and in that habitat alone. What it did, however, it did superbly: give
us a foretaste of the future. And if ibm has its way – and the sensa-
tional communiqués that pour from its research division suggest
that it will – fully interactive Watsons and even Holmeses are com-
ing our way in health care, customer care, finance, education, trav-
el, management, and of course the military, just for starters.
 6

QED

Troubles with Functionalism; Print This!; Trompe l’Oeil;

Like the Pope; A Bevy of Pigeons; No More, No Less; Dicty;
Operational Definition?; Flugblogs; On Being Mental

troubles with functionalism

In “Troubles with Functionalism” (1978), Ned Block advanced a

scenario by means of which he hoped to raise doubts about the
validity of the Turing test. The idea was that a Propositional Juke-
box machine could be rigged to pass the test not by virtue of being
smart but by the brute power of sorting through gazillions of tem-
plate responses. Although this “blockhead” counter fell through
the cracks all the way down to the philosophical basement, John
Searle’s no less fanciful counterfactual from “Minds, Brains, and
Programs” (1980) struck a raw nerve.
To this day, Searle stands at the forefront of a cacophony of cri-
tiques directed at the validity of the tt. His prominence, it must be
said, has less to do with the logic of his arguments than with his
tone, as shrill today as it was thirty-plus years ago. To his credit,
Searle’s attempt to rebut the idea of a thinking machine is quite
straightforward, even if it relies on props that are anything but.
The philosopher pictures himself in a room (whose dimensions
would, in fact, dwarf the entire cosmos) stacked with the boxes of
Chinese symbols that form his database. He is equipped with a rule
book that plays the role of his program. The rule book advises him
where to look up Chinese symbols in response to the input he
receives. As a non-speaker of Chinese, he executes his instructions
“blindly,” by matching received ideograms to those in the book and
then to those in the boxes. In this way he performs unknown oper-
112 From Literature to Biterature

ations on his input in accordance with the rule-book program, as a

consequence of which he produces bunches of equally meaning-
less ideograms (output).
In short, the set-up is such that in the Chinese Room the human
being plays the role of a cpu that executes a program and gener-
ates feedback in the form of answers to tt questions. To make it as
plain as can be, the structure of Searle’s thought experiment can
be paraphrased as a quasi-syllogism:

1 Computer programs are formal (syntactic) and thus mindless

constructs, and thus neither equivalent to, nor by themselves
sufficient for, understanding semantic content;
2 minds as we know them understand semantic content;
ergo:
3 no program in and of itself can be equivalent (give rise) to a
mind.

The suppressed premise in Searle’s reasoning is that computer

syntax is not equivalent to nor by itself sufficient for semantics (i.e.,
understanding). In effect, he says that there can be no program –
more generally, no informational process – that could ever pro-
duce a mind that thinks. Hence he italicizes in The Mystery of Con-
sciousness (1997): “If I don’t understand Chinese solely on the basis of
implementing a computer program for understanding Chinese, then neither
does any other digital computer solely on that basis, because no digital com-
puter has anything I do not have.”1
Searle’s interpretation of his own thought experiment is mis-
guided. Note, at this point, that he is quite right to maintain that
the human mind does not work exactly like a serial and binary
computer. He is also correct to argue that top-down, rule-based syn-
tactic programming is not the way to achieve practical machine
intelligence. This, however, in no way proves that his scenario
establishes that computer thinking/understanding/intelligence is
impossible.

print this!

Searle’s counterfactual founders on its centrepiece – the possibili-

ty of creating a Leviathantine template-response database that
could in principle provide for every conversational contingency.
 QED 113

Exceeding the computational capacity of all and any devices that

could be manufactured in our universe, its dimensions and thus
the time required for scanning and retrieval are for all purposes
infinite. As a consequence, the only thing that does follow from
Searle’s premises is that no one could ever write down all answers
to all tt interactions and stuff them in boxes, which we knew any-
way.
The thought experimenter presupposes that any question that
can be formulated can be fed into the Chinese Room, which will
then dish out the correct answer. Conceptually, this is equivalent to
writing down every line of every book that could ever be written
and stuffing them in boxes, indexed for retrieval. To bring out the
true dimension of this supertask, forget all the books that could be
crammed into the Chinese Room Larger Than The Universe. Let
us look at just one book, and a pint-sized one at that: two hundred
pages, fifty lines per page, fifty spaces per line.
Numerically, the problem consists of calculating the state space
for our pocket book. The amount of all possible printable spaces
in it is 200 x 50 x 50 = 500,000. The variables with which every
printing space can be filled are the lower and uppercase letters of
the alphabet plus the typographical signs. To keep things simple,
let us agree that there are only fifty signs in all: twenty-four lower-
case, twenty-four uppercase, plus the period and the comma. The
number of possible permutations? A number so vast that it is in
almost every sense meaningless: 50500,000.
Here is why. The number of atoms in the universe is 1080. The
universe itself has been around for fourteen billion years. Assum-
ing that every atom is a printing press, and that each has been
printing non-stop since the Big Bang at the rate of atomic vibra-
tions, by now they would have printed such a minute fraction of a
single book in a single box on a single shelf that you would have to
work your way through two hundred pages of zeros before, on the
very last page, third line from the bottom, you would find the first
meaningful decimal.
Print this!
Searle’s counterfactual has exactly zero upshot for computer
thinking. What it does rather well, on the other hand, is show that
traditional top-down ai research is a dead end. For decades now,
Douglas Lenat, perhaps the last of the old school, has tried to spawn
artificial intelligence by spoon-feeding heaps of data to a program
114 From Literature to Biterature

The thought experimenter presupposes that any question that can be for-
mulated can be fed into the Chinese Room, which will then dish out the
correct answer.

called cyc (as in encyclopedic). Among other quanta of intelligence

about human consensus-reality – common-sense knowledge, in
English – typed into the machine are such arcane morsels as “bread
is food,” “George Bush wears underpants,” and “you’re wet when
you sweat.”
At some point in the unspecified future, cyc is supposed to
reach the saturation point, whereby it will become cognitively inde-
pendent, whereupon it will spring into action and start to augment
its knowledge database on its own – say, by reading books. Hope, of
course, springs eternal. Meanwhile, not only ai gadflies but insid-
ers concede that the top-down, syntax-based approach to machine
thinking displays all the signs of a degenerating research program.
Much as I predicted in 2000, here I go again: Lenat’s crusade will
fail to bring home the grail.

TROMPE L’ OEIL

Searle’s counterfactual is essentially a clever trompe l’oeil. Here is

why. The philosopher invokes understanding as the criterion for
deciding whether a tt player is truly intelligent rather than only
looking like he is. For all I know, however, Searle does not under-
stand his own arguments. For all I know, there is a homunculus in
 QED 115

his cranium that mindlessly shuffles molecules from one place to

another in accordance with a mindless if tremendously long pro-
gram. Undeniably, this elaborate shuffling gives a semblance of
purposeful behaviour to the walking and talking constellation of
atoms that, for convenience’s sake, we refer to as John Searle.
But underneath the façade, there are only trillions of particles of
inert matter and a lone demon that agitates them according to
instructions coded in the language of physics and biochemistry. In
short, contrary to all appearances, Searle does not understand
what he says, and only the phenomenal velocity of the homuncu-
lus’s operations perpetrates the hocus-pocus of reasoned response,
i.e., of intelligence. Amused but not convinced? Then try to prove
me wrong. Try as hard as you can, you cannot – at least, in Searle’s
terms. To prove me wrong, you would have to prove him wrong.
Let us look at the whole set-up from another angle. The occu-
pant of the Chinese Room is said not to understand the squiggles
and squoggles which he can, nevertheless, manipulate expertly ac-
cording to the transformation rules written down in his rule-book.
My question is: How can Searle know that the occupant does not
understand Chinese? Since he cannot look inside the man’s head,
the only sources for this claim must be functional and behavioural
in nature. Once again, Searle buys into the very premise he would
like to reject.
Suppose that, to prove his point, the philosopher locks himself
in the room. After all, he knows that he does not understand Chi-
nese. No dice, either, for how do I know that he really does not
understand Chinese? My sole source of evidence is his solemn oath
or his behaviour, both of which amount to the same thing: induc-
tive functionalism. After all, even if we were to crack his head open,
we would not discover “understanding,” only a mindless swirl of
atoms and molecules.
Like it or not, once more Searle’s argument relies on the very
premise he wishes to undermine. This is no accident, simply because
it is impossible to understand “understanding” in any other terms
than its actual use. Try this at home. Ask yourself whether you under-
stand that 2 + 2 = 4. Presumably your answer is affirmative. Now,
prove that you understand it in any manner that does not involve
manipulating the elements of the equation or, more generally, inter-
acting with the real world in a manner consistent with the fact that
you understand it. Try as hard as you can, you can’t.
116 From Literature to Biterature

Knocked out again, the sceptic could remix the Chinese Room
into yet another argument that goes like this. Assuming that the
man in the room does not understand Chinese – in other words,
assuming what is to be established – the mental phenomenon of
“understanding” is not contained in his head. Nor is it contained
in any one of the bits of paper that store the rules or the answers.
Since none of the constitutive elements understands anything, the
person-plus-paper cannot understand anything either, no matter
how many more rules and bits of paper are fed into it. After all, if
one scrap of paper does not think, two scraps do not either, nor do
three, four, or any higher number. Right?
Wrong. To see why, consider Searle’s own brain. Take any small
part of it, be it a dendrite or even any of the proteins from which
it is constructed, and ask yourself whether it understands anything.
Of course, it does not. Then take another part and ask the same
question, and so on. As no one part of the brain understands any-
thing, a mere expansion of non-understanding bits of brain in his
head could not at any point give rise to intelligence, right? Which
proves my point again: John Searle does not understand what he
says.

like the pope

Searle himself is a living reductio of his argument against con-

sciousness. But even if you were to take his argument on face value,
you would end up in the land of Oz. For one, in Searle’s world
there would be no rivers. A river is, after all, but an agglomeration
of droplets swirling and twirling around one another. But as no
one droplet carries the quality of riverness, where is the river?
Where are streams, oceans, and soups in the world where stream-
ness, oceanness, and soupness are not localized in any of their con-
stitutive elements?
Let us do it again in slow-mo. Rivers, oceans, and soups emerge
from agglomerations of droplets or, deep down, H2O molecules.
No one droplet or molecule triggers any kind of phase-transition
from a mere collection of river droplets into a full-blown river. Yet,
to be consistent, Searle cannot speak of rivers because riverness is
an emergent property that does not reside in its constitutive com-
ponents – just as thinking does not reside in the properties of the
atoms that build neurons in a brain or chips in a computer.
 QED 117

Not convinced? Here is another reductio, this time constructed

along diachronic lines. The sceptic maintains that no blind, mind-
less, unconscious, information-recombinant process could give rise
to mentality and semantics. But once again he himself constitutes
the best proof to the contrary, inasmuch as, just like all of us, he is
the product of a blind, mindless, unconscious, and information-
recombinant process: evolution.
Watching John Searle get knocked out time and time again, only
to dust himself off and continue the skirmish, you may begin to
sense a deeper methodological problem. There is simply no evi-
dence on earth that could ever prove to his satisfaction that he is
not an adroit android. After all, no matter how smart a person or
a machine is, you can always insist that deep down it is all a matter
of mere programming. And so Searle insists that no program could
ever think, even as his doubts are no longer scientific but dogmat-
ic since they cannot be falsified. At this point the sceptic becomes
not only irrefutable but infallible, like the pope.
In an essay entitled “The Mystery of the Chinese Room,” from a
1996 collection entitled, appropriately enough, The Mystery of the
Chinese Room, Lem himself countered Searle’s thought experiment
with one of his own. Take a jigsaw puzzle, proposes the writer,
reverse all the tiles blank side up and mix them up thoroughly.
Then bring someone into the Jigsaw Room and ask him to assem-
ble the puzzle still blank side up, going entirely by the contours
(read: formal features) of the tiles. Once you turn the solved puz-
zle over, of course, you discover an answer to a tt-type question.
That is all. Now ask yourself: Does the Jigsaw Room scenario
prove anything about the impossibility of computer intelligence? If
you answer “No, how could it?” you have also answered Searle, just
because Lem’s scenario is wholly isomorphic with the Chinese
Room. Caustically, Lem himself observes that the latter precludes
Artificial Intelligence as much as a refutation written in chocolate
éclairs precludes the next volcanic eruption of Mount Etna.

a bevy of pigeons

In The Symbolic Species (1997), anthropologist Terrence Deacon

entertained an even wackier variation on the Chinese Room by
replacing the man with a bevy of pigeons. Each bird had been
miraculously trained to match incoming Chinese characters to
118 From Literature to Biterature

those in the rule book and, subsequently, to those in the storage

boxes. Even though no single pigeon could be charged with know-
ing Chinese, concludes the author, collectively they would be lan-
guage processing.
This is correct. But what follows? It is difficult to tell. At this point
the counterfactual has become so outlandish that the only vision it
conjures in my mind is that of aeons needed to train gazillions of
birds to process the variable input of some five thousand characters
in the Chinese language. This is not to even mention that most
ideograms are neologistically recombinable, mooting the scenario
from the outset – unless, of course, you bring a Chinese-speaking
pigeon trainer into the room or allow that the birds could learn
Chinese as they went along. qed.
All this is to say that, more than the infinitely large cosmos filled
with pigeons, feathers, birdseed, birdshit, and people trying to
have a conversation with them by means of bits of paper, I would
find it more plausible that, over aeons of habitat pressure, the birds
themselves would evolve into a thinking species with as much flu-
ency in Chinese as the Chinese. Perhaps recognizing this, Deacon
claims not so much to furnish a rigorous answer to “Are the birds
in the Chinese Room thinking or not?” as to highlight the gulf
between indexical (syntactic) and symbolic (semantic) operations.
You might be justified in concluding at this point that the Turing
test is really a Rorschach test by another name, inasmuch as every-
one seems to see in it what they want to see. If so, let me put my
cards on the table. I regard the entire Chinese Room diatribe as to
whether computers can really think or merely process information
satisficingly enough to get along in the world to be meaningless in
Turing’s sense – semantically not empty, but pointless and sterile.
Like love, consciousness is not a localizable object that can be put on
display but is instead a process detectable by its effects. It is a man-
ner of describing a manner of behaving.
Whether computers can ever understand a story they have just
read or written is an empirical question to be resolved by future
generations of people and machines. Naturally, pragmatic factors
such as socio-cultural traditions or even legislation may come
between us and an acknowledgment of thinking in machines for
no other reasons than that they are machines. Bear in mind, how-
ever, that the ability to function in the world – one part of which is
communication with other thinking beings – has always been suffi-
 QED 119

cient grounds for ascribing intentionality (or derived intentionali-

ty, if the term is not tautological in this context) in humans by
humans.

no more, no less

In I Am a Strange Loop (2007), Hofstadter concludes that Searle’s

error stems from assuming that “the level of the most primordial
physical components of a brain must also be the level at which
the brain’s most complex and elusive mental properties reside.”2
This is an elegant exposition of what is known in the parlance as
a systems argument. Even though the man inside the Chinese
Room really does not understand Chinese, the system as a whole
acts in a manner that is indistinguishable from someone who
does. The view that the mind is nothing but a kind of software
running on cranial hardware is, of course, a textbook case of
material reductionism.
But where does the mind come from? Like Turing, Hofstadter
regards the question as misconceived, and in an effort to illustrate
why, he elaborates the idea formulated more than half a century
ago by Herbert Simon and Allen Newall, that intelligence consists
of the ability to manipulate multiple levels of symbolic structures.3
In a whimsical analogy to thermodynamics and statistical mechan-
ics – two equivalent but mutually non-reducible levels of descrip-
tion of physical systems – Hofstadter describes consciousness in
terms to two different levels: thinkodynamics, at the neuronal and
synaptic level, and statistical mentalics, at the level of symbolic
operations.
At the level of which we are aware, the mental pictures in our
minds – be it sorrow at the passing of one’s uncle or satisfaction at
being accepted into grad school – ignore microscopic elements
such as neurotransmitters firing, synapses activating, and so on.
Naturally, all mental pictures are at bottom no more than the sum
total of small-scale events that are purely physical and intention-
less. And in between are the countless biophysical processes and
countless levels of organization that, gradually, represent any given
input more and more symbolically, i.e., mentally.
There is no unique “Eureka” level where it all comes together as
a mind that is separate from all the other levels, just as there is no
unique moment in human ontogeny when a child begins to think
120 From Literature to Biterature

or a moment of phase transition when a stream turns into a river.

There are only increasingly complex symbolic representations of
neural activity, at least one of which – the one we are aware of – con-
tains models of other subsystems and thus ingrates the “self.” So
what is the mind? It is a dance of non-isomorphic levels of descrip-
tion. It is a non-local epiphenomenon. It is an attribution made pos-
sible by Theory of Mind in other minds.
The last word on where the mind comes from goes, appropriate-
ly enough, to Turing. In his magisterial The Essential Turing (2004),
Jack B. Copeland, director of the Turing Archive for the History of
Computing, refers to Turing’s lost proof that has a direct bearing on
the apparent mind-brain duality. Stunningly, Turing appears to
have proven that, through the process of learning, initially random
neural networks could self-organize into a “universal Turing
machine with a given storage capacity.”4
The implications for the apparent duality – from top-down, minds
and complex symbolic sequential processing; from bottom-up, par-
allel cascades of firing neurons – are not hard to tease out. If correct,
Turing’s proof established that the human brain can be modelled as
a universal computer running on a bioneural network. Here then is
the long-sought synthesis of the top-down and bottom-up perspec-
tives. Deep down, the more-or-less continuous and coherent source
of memory and self-awareness that we call the mind is a neural net-
work organized from discrete events of the brain. No more, no
less.

dicty

Let us come back one last time to Searle and his riverless world. His
claim that computers cannot ever evolve mentality amounts to
claiming that there are no such things as autocatalytic properties.
This is a mystifying position in view of the large and growing body
of research on autocatalytic systems, spearheaded by Stuart Kauff-
man. Decades-long in the making, his findings open a Sesame of
scientific treasures of which we may yet have caught but a glimpse,
precisely because they are so fundamental in nature, being ab-
stracted from any specific physical embodiment.
Kauffman set out to investigate the characteristics of complex net-
works irrespective of what they were made of: neurons, genes, tran-
sistors, light bulbs, phone relays, and so on. His insight was that there
 QED 121

may be some fundamental properties common to all systems that

inhere in their level of complexity and not in their material charac-
teristics – properties that obtain as much for the Amazon as for rivers
of sand cascading down Baltic dunes, or for cascades of firings in a
connectionist network in our brain or inside a biterary computhor.
Varying input levels between neighbouring elements of net-
works, he stumbled on something extraordinary. Tucked away
between a degenerative level, where the net would sputter into a
steady state, and the onset of chaos, where stable patterns were as
hard to detect as in a table of random numbers, there was a narrow
band with a strikingly different behaviour. The instant you hit it,
the network would come alive. Instead of doing nothing all, frozen
solid for all eternity, or cranking up blizzards of “snow,” it would
cycle through a rich but limited number of states.
Buried among all possible configurations, there was a range that
was dramatically different: not chaotic or rigidly ordered but com-
plex. And, if that were not amazing enough, far from being a
freaky anomaly, for any network there was always at least one auto-
catalytic optimum between degeneration and chaos. Emergence, it
turns out, is only waiting to emerge in all kinds of places, often sud-
denly, in what is known as a phase transition: a rapid onset of novel
properties throughout the system, such as when water freezes into
ice. Watch the water drain next time you take a bath. Once the
level drops – bang, all of a sudden you get a mini-whirlpool fun-
nelling into the drain.
The point is that water (or air) turbulence cannot be explained
in terms of its individual components. The behaviour of the emer-
gent state is absolutely unpredictable from the initial settings. You
can explain viscosity or density in terms of the properties of mole-
cular H2O, but not turbulence. Put differently, fluid dynamics can
model the vortex motion but not its onset. Like intelligence, turbu-
lence emerges only at a certain level of system complexity without
being detectable in any way prior.
So it is with esprit de corps, which emerges only at the level of a
social group and is not derivable from or reducible to the charac-
teristics of the human “atoms.” Traffic fatalities are globally pre-
dictable to within a margin of error, reliably making money for the
insurers. But at the level of a single accident, there is nothing to
indicate death rates, even though statistical regularities emerge
from lower-level collisions.
122 From Literature to Biterature

Kauffman set out to investigate the characteristics of complex networks irre-

spective of what they were made of: neurons, genes, transistors, light bulbs,
phone relays, and so on. His insight was that there may be some fundamen-
tal properties common to all systems that inhere in their level of complexity
and not in their material characteristics – properties that obtain as much for
the Amazon as for rivers of sand cascading down Baltic dunes, or for cas-
cades of firings in a connectionist network in our brain or inside a biterary
computhor.

The same phenomenon shows up in Dicty (Dictyostelium discoi-

deum), in many ways a garden-variety amoeba. In distress situations,
however, this monocellular protozoan behaves in a way that is noth-
ing short of remarkable: it begins to aggregate and act like a super-
organism, even though it lacks a central “brain.” Yet, until it reach-
 QED 123

es this state, there is nothing to hint of this dramatically novel

behaviour. In a similar way, termites construct their mounds with
perfect meridian alignment and dynamic-airflow air-conditioning,
without blueprints or an architect-in-chief. Study each worker as
much as you please, however, and you will not find a mini-compass
or air-con module in its brain.
Many complex systems, in short, display emergent properties,
which is another way of saying that many complex systems are self-
organizing. We can know about them as much as there is to know,
and we will still not be able to derive their higher-level autocatalyt-
ic properties. And so it is with thinking. Below deck, human beings
run in accordance with blind, inanimate, inflexible laws: Pauli’s
exclusion principle, Ohm’s law, and so on. But even though this
deterministic rule-book is always there, it manifestly does not stand
in the way of emergence of higher-lever properties such as mental-
ity or volition.
After all, we all think – which is to say, we all assume we all do.

operational definition?

One would expect little consensus among the opponents and pro-
ponents of the tt on its extension and validity. It is striking, howev-
er, how little consensus there is even on such a fundamental issue as
to what the tt actually tests for. Thinking? Computer thinking?
Equivalence of computer and human thinking? Their identity? Take
any two philosophers writing on the subject, and the odds are they
will be as far apart as Robert French and Dale Jacquette, for whom
the test defines, respectively, intelligence and artificial intelligence.
Among this plethora of positions and oppositions, the only one
that has rallied philosophers over the decades is the belief that the
tt is an operational definition of whatever it is supposed to define.
P.H. Millar was explicit on this point, declaring that the tt gives
“an operational definition” of – judging from his context – com-
puter thinking. Ned Block, of the “blockhead” counter, calls it a
“behaviourist definition of intelligence.” Even Turing’s assiduous
biographer Andrew Hodges categorizes it as “an operational defi-
nition of ‘thinking.’”5
The misunderstanding persists all the way into the twenty-first
century. In a retrospective on the first fifty years of the tt, Robert
French declares that the test was proposed as an “operational defi-
124 From Literature to Biterature

nition of intelligence.” In the Wadsworth Philosophers Series, John

Prager goes as far as to underscore his belief that the tt is an “oper-
ational definition by which the performance of a machine could be
compared with that of a human at tasks presumed to require intel-
ligence.”6
Why is this a misinterpretation? First of all, because it funda-
mentally misrepresents Turing’s view of what the test was all about.
This view was so important to him that, in the very first sentence he
uttered on the 1952 bbc panel on artificial intelligence, he object-
ed, “I don’t want to give a definition of thinking.”7 Secondly, it is
perfectly possible for a machine to think as well as you and I do and
still fail the test – just as people hampered with mental handicaps,
linguistic impediments (for example, dyslexia), or massive cultural
differences could flunk it. As such, if the test gives a definition of
thinking, it is a deeply flawed definition to start with.
In truth, if all that the tt does is give an operational definition
of a new concept, then the definition itself needs little justification
but the test becomes of scant interest and consequence. In my
opinion, it is just the reverse. The test cashes out the concept of
machine thinking by appealing to our ordinary notion of thinking,
however fuzzy it might be. In fact, it provokes so much gut-level
hostility precisely because of the appreciable overlap between nor-
mal human thinking and the behavioural potential of a computer.
An unequivocal sign that Turing thought of the test in inductive
rather than logical-categorical terms is the pragmatic stance he
adopts in “Computing Machinery and Intelligence.” In the central
passage, he states that by the year 2000, “we will be able to speak of
machines thinking without expecting to be contradicted.”8 By the
same token, he never claims that a computer that aced the tt would
think. In fact, he regards such a proposition as meaningless. It is for
no other reason that he replaces the question “Can machines think?”
with the tt. Any interpretation of the test as an operational defini-
tion of computer thinking would presumably strike him as equally
meaningless.
Mapping computer thinking onto the semantic evolution of the
concept, the tt does not provide the necessary or sufficient condi-
tions for machine thinking. Its merits lie, instead, in providing a
source of controlled inductive evidence for or against computer
intelligence. And with this goal in mind, it can be retooled into as
exacting an examination as need be. One limitation of the tt
 QED 125

brought up by sundry critics is its inability to test the non-verbal

aspect of the players’ behaviour. Although true, the limitation can
be easily remedied. Non-verbal behaviour can be described by the
players, and examiners can and should demand descriptions of
complex activities on the order of planning a lecture, reconciling
quarrelsome friends, selecting a prom date, and such like.

flugblogs

Strangely enough, one might amass heaps of evidence that a com-

puter thinks as well as you and I do, and it might still fail to imper-
sonate a human being in a suitably rigged examination. In “Sub-
cognition and the Limits of the Turing Test” (1990), Robert French
describes, in fact, an entire class of protocols devised to trip the
machine. Whether they remain the kind of tests that Turing had in
mind is another story.
French analyzes a range of tricks that could identify a machine
owing to the fact that human minds are primed for subcognitive
associations. A good illustration of the latter comes from the so-
called Rating Games, which can involve different categories such as
neologisms, jokes, and advertisements. For example, players may
be asked to rate, on the scale of 0 (completely implausible) to 10
(completely plausible):

* “Flugblogs” as the name Kellogg’s would give to a new breakfast

cereal;
* “Flugblogs” as the name of a new computer company;
* “Flugblogs” as the name of big, air-filled bags worn on the feet
and used to walk on water.9

The key to detecting a machine is a significant modification of

the stt. French assumes that he could poll a statistically viable sam-
ple of people on the Rating Games prior to the test and use the
results during the actual run to make out the computer. The par-
ticipant whose answers matched the statistical profile would likely
be human, who, in rating the entries, would automatically fall back
on subcognitive associations typical of beings who experience the
world in the same way the rest of us do.
This is true. As we continuously discover, many facets of what we
used to regard as cognitive (deliberate) processing is really sub-
126 From Literature to Biterature

cognitive. The recent discovery of mirror neurons is a paradigmat-

ic case. Although we have no idea how they do it, mirror neurons
recognize and respond not to gestures but to the intentions behind
the gestures of other agents. In this, they are part and parcel of the
suite of biological adaptations known as Theory of Mind that
enable us to recognize agency and, as such, enable human-type
communication.
French’s wedge test is grounded in the fact that, when informa-
tion is scarce, we fall back on quick and dirty guesstimates. As a
result, we may get things wrong, but even when we do, we get them
consistently wrong. This is precisely why anchoring effects can be
used to drive a wedge between human and computer. A machine
like Watson would likely know the answer to “How many African
countries are in the UN?” but if it did not, it would presumably not
be primed to guesstimate it as humans are.
In short, to pass the wedge test, computers would have to exhib-
it the same subcognitive effects as people do. Or would they? Some
people always score as outliers, casting some doubt on French’s
stratagem. But even if that were not the case, my question is, so
what? Even if you could guess that one player is a computer, that
would tell you nothing about whether it is thinking. Moreover, as
we move toward connectionist nets and more flexible, learning-
based interactions with their environment, the premise loses some
of its bite. It runs out of steam even further in systems that might
be capable of learning and experiencing the world in ways similar
to humans, such as biterary computhors of the cis-humana type.
In sum, the ability to think is best appreciated in terms of an
inductive theory about an interactant’s ability to function in ways
that are intelligible to humans but not necessarily in ways that are
indistinguishable from humans. Once again, if we ever find ourselves
face to face with a computer that could ace the Turing test, we will
find it unnecessary to run one. The inductive evidence from daily
intercourse with the machine will be far in excess of the evidence
needed to resolve whether it can be said to think.

on being mental

Needless to say – but perhaps worth saying all the same – all of my
arguments are grounded in the physicalist concept of mind. Without
getting entangled in the long history of division between dualists and
 QED 127

A machine like Watson would likely know the answer to “How many
African countries are in the un?” but if it did not, it would presumably
not be primed to guesstimate it as humans are.

materialists, it simply means that they are grounded in the non-Carte-

sian school of thought in which mind and consciousness are a part of
physical reality, not apart from it. Historically, of course, this has not
been the only view in town. In the twentieth century alone, philoso-
phers and scientists as diverse as Bergson, Elsässer, and Polanyi insist-
ed on a vitalist component (élan vital ) in all living organisms.
On their view, computers by definition could never become think-
ing or alive. An essentialist could always maintain that a machine
128 From Literature to Biterature

that passed the tt only mimicked human thinking rather than

duplicated it. If I am unconvinced by this type of objection, it is
because the difference between mimicking some action and dupli-
cating is fluid at best. Take a machine that finds logical proofs, reg-
ulates traffic, diagnoses diseases, and beats grandmasters at chess.
Such systems exist today, but it would make little sense to say that
they only mimic all these actions.
There exist, of course, scores of refutations of vitalist arguments,
most of which follow from the refutation of Descartes’ analysis of
the mind-body problem – in effect denying the duality. Among
those, Donald Davidson’s is arguably the most original, if only
because it stands the duality on its head by arguing that intention-
ality provides an argument against dualism. Indeed, his so-called
anomalous monism amounts to claiming that instead of two kinds
of stuff in the world, mental and physical, there is only one, which
we conceptualize in these complementary ways.
Curiously enough, anti-dualist arguments have also emerged from
the philosophical antipodes under the name of panpsychism.
Panpsychists even start from the same premise as physicalists, name-
ly that brains are minds, thus dispensing with élan vital and other ani-
mistic additives. At bottom, in other words, both physicalists and
panpsychists see everything in the world as the interplay of physical
forces and their manifestations: elementary particles, chemical mol-
ecules, rocks and rivers, livers and brains. But panpsychists take the
other fork in the road when it comes to consciousness.
The standard problem for the physicalist is how to get from
colours and flavours of inanimate quarks to the subjective visual or
gustatory sensations of a mind in human beings like you and me.
Panpsychism offers a solution – and it is a drastic one. It cuts
through the Gordian knot by postulating that everything in the
universe is imbued with mentality and subjectivity. Nuts, you say?
Well, yes. On the other hand, if even subatomic particles have
proto-mentality, the problem of getting from brains to minds dis-
appears in a flash.
The problem is, if all matter in the cosmos is psychic, everything
that is matter should have psyche embedded in it. Do Weakly Inter-
acting Massive Particles (wimps) and Massive Compact Halo Objects
(machos), then, desire to avoid solo existence as they gravitational-
ly attract one another? Are neutrinos mental hermits that elect to
interact so weakly as to be almost undetectable? Do molecules bond
 QED 129

to live happily ever after? Do electrons dream of electric sheep?

Panpsychists’ defence that the mentality of elementary particles is
so vestigial that it goes unnoticed begs the question of how they
know it is there in the first place.
In another version, the same panpsychism said to permeate all
natural bodies excludes man-made artifacts and objects such as
rocks or planets, which are said to lack the unity of organization. If
both exceptions look suspiciously like Ptolemaic epicycles, it is
because crystal lattices in rocks such as quartz, the second most
profuse mineral in our planet’s crust, are uniformly organized.
Moreover, atoms are immutable and, as such, should preserve their
volitional properties even if they migrate from natural hair to arti-
ficial hair extensions.
Indeed, if you think about it, any version of panpsychism that
embraces such a radical discontinuity between different types of
atoms or objects – some volitional, others not – is really physicalism
by another name. At some stage in the universe, mentality appears
where it was not there before, so panpsychism is stuck with the
same emergence problem as physicalism. Not surprisingly, panpsy-
chism is hardly a major school of thought in philosophy, to say
nothing of science. It accuses physicalism of being at odds with
experience and reality, be it commonsensical and scientific, yet nei-
ther folk physicists nor those gainfully employed at cern have ever
reported running into stones that would refuse to break bones or
muons giving the finger to the magnets that drive them. Add to
this the problem of explaining what exactly is the mentality of a
quark, or the combination problem (how does human mentality
emerge from the combined mentalities of atoms?), and you can
safely lay panpsychism to rest.
130 From Literature to Biterature

ToM

Kismet; PaPeRo; Guilty as Charged; Eureka; Meaning;

Elementary, Dear Wason; Biology in Action; Driving Range;
Insula Cortex; On Being Redundant

kismet

Both Kismet and I are clever contraptions that excel in social inter-
actions. But whereas I am an assembly of colloids and dendrites,
Kismet is a heap of gears and microchips. To the world it presents
a vaguely donkey-like physiognomy with oversized eyes, furry eye-
brows, pointy ears, and a pair of orange lips that look like they were
squeezed out of a tube of novelty-store toothpaste. In short, not
much in human or even humanoid terms. And yet, unfailingly, this
coarse configuration of extensible parts is integrated as a face with
all that it implies in psychological terms.
One reason is its surprising mobility. The steel-mounted elements
are capable of assuming a variety of shapes identifiable as expres-
sions of surprise, delight, disgust, or doubt. Moreover, this mit ro-
bot is more than a one-trick donkey. It is equipped with synthetic
speech which allows it to engage in basic chatter and even learn a
limited amount of new words, such as people’s names. Six cameras
allow it to track its environment and focus on visitors’ faces or on
objects they may brandish in front of its eyes.
At the end of the day, it has, of course, as much personality as a
New York subway turnstile. Yet children love to play with it, seem-
ingly oblivious to the fact that Kismet is an aggregate of steel and
circuits – a mechanical puppet on binary strings. Shyly at first, they
might show the robot their favourite toys and see the gesture
acknowledged with a nod and a toothpaste grin. Ice broken, they
 ToM 131

might scamper around trying to dress up their new friend or con-

vince it to play a game, while Kismet obligingly turns the smiley
donkey face to follow the commotion. All the while, the kids chirp
to it relentlessly, peering at its face and in general treating it like a
living pet.
Odd as it may seem, there is nothing odd in any aspect of the
children’s behaviour, down to the fact of looking into the robot’s
eyes. The naturalness of this gesture is confirmed by Juan-Carlos
Gomez’s research with apes, who also frequently look into humans’
eyes when issuing requests. This shows not only that apes monitor our
attentional state but suggests that they may be aware that intention-
ality originates behind the eyes and not in the limbs that carry out the
requested action, be it dispensing food, playing, or grooming.
Returning the attention, Kismet learns and speaks out the chil-
dren’s names, making the illusion of communicating and develop-
ing a relationship even more seductive. Indeed, the impression of
interacting with another mind is so natural that it is shamelessly
exploited in bestselling toys like Furbies and in the latest genera-
tions of Japanese robotic pets for the elderly which speak gibberish
until they “learn” simple expressions from their owners. In reality,
they automatically begin to speak a certain number of phrases after
a factory-preset time.
Today Kismet gathers dust on a display stand at the mit muse-
um in Cambridge, Massachusetts. But in its heyday in the late
1990s, it was such an interactive sensation that even roboticist
Rodney Brooks did not hesitate to hyperbolize, “Kismet is alive.
Or may as well be. People treat it that way.”1 Certainly most chil-
dren did, seeing the robot as a friend that liked them back, even
though they were carefully told that it was not alive but only pro-
grammed to respond to verbal and visual cues. But if you think
that only kids could be taken in by such crude plays, you are due
for a rethink.

papero

Meet Ed Feigenbaum, distinguished ai expert, whose name you

may recall from the Feigenbaum test, the Jeopardy-type domain-
specific imitation of the Turing test. If there is one individual who
ought never to fall prey to robotic mimicry, it is he. Feigenbaum
recalls spending a part of the day with PaPeRo (for Partner Pet
132 From Literature to Biterature

Robot) – a social-interactive machine of the sort that the Japanese

excel in – on a visit to the artificial intelligence labs of nec (Nip-
pon Electric Company) in Tokyo a few years ago. His stunned reac-
tion is worth quoting in full:

When I saw they booked me for an hour with PaPeRo, I thought

I would be bored to tears after twenty minutes. But that was not
the case. The experience was riveting, fascinating, attention-
grabbing. How could that be? PaPeRo was just a bunch of plastic
and electronics running around the floor, and I knew how it
worked, so there was no magic.2

No magic – but enchantment all the same. The diminutive robot

was evidently expert at playing a playful, lovable child, tapping into
the adaptive psychosocial mechanisms embedded in all of us. You
may remember the late Steve Jobs’s animated film company Pixar
and their skits depicting two desk lamps. America or Asia, male or
female, young or old, the whole world saw the same thing. The
baby lamp was cute and playful, the mommy lamp was loving and
protective. All this about two inarticulate pieces of standard-issue
office equipment hopping about.
Things are no different with interactive software, beginning
notably with Weizenbaum’s binary therapist eliza. Named after
Eliza Doolittle, a Pygmalionesque handiwork of a linguist Professor
’Iggins (if you remember your Shaw), the program was developed
in the 1960s to imitate person-centred psychotherapy in the Roger-
ian tradition. The distinction is significant in that in this non-direc-
tive approach, tailor-made for computers, the therapist trades
active probing for gentle prodding.
The goal is to create a non-judgmental environment that will
give the patient unconditional positive feedback. In practical
terms, the cyber-therapist invites the patient to monologue, from
time to time coaxing progress with a thoughtful – or, if you like,
thoughtless – “I see.” When introspection stalls, it may be reacti-
vated by turning statements verbatim into questions (“I lost it all.”
“You lost it all?”) or by falling back on preset gambits such as
“What does it suggest to you?” or “How did that make you feel?”
When the session begins to show signs of sputtering out, the script
rolls out the heavy arty: “Tell me about your childhood/mother/
father/dreams.”
 ToM 133

No magic – but enchantment all the same.

134 From Literature to Biterature

The first casualty was Weisenbaum’s assistant, who naturally knew

only too well how the code was developed to mimic a knowledge-
able therapist. All the same, once the script was up and running,
she succumbed to the illusion that she was conversing with an
understanding and sympathetic mind. After only a few minutes’
typed interaction with the program, she got sufficiently involved to
ask her incredulous boss to leave her alone with it.
She may have been the first, but she was not the last. Even as
eliza was never meant to delude anyone that “she” could think
and offer solace, like Nathanael West’s hapless Miss Lonelyhearts
she did so anyway, to her creator’s perpetual exasperation. Today
you can try a little therapy with dozens of online elizas to find out
how susceptible you are to the allure a perfectly unflappable shrink
that, miracle of miracles, charges nothing by the hour. Or you can
find an Easter egg in Apple Utilities that works more or less like the
original, ready to lend an ear to your rants and sorrows.
Research shows that, as soon as machines are given a face and a
voice, we are tempted to respond to them as if they had mentality.
In literary studies, this propensity was recognized long before the
science of psychology was even on the horizon. Known as pathetic
fallacy, it is a matter of hanging human characteristics on inani-
mate entities such as, for instance, the weather. The heavens crying
raindrops over the heroine in distress may be the stuff of tacky
melodrama, but it reveals a deep-seeded proclivity for Theory of
Mind that goes beyond computers or even literature.
We all form bonds with old photos, mementos, heirlooms, and
contexts (“they’re playing our tune”) that stir up our memories and
emotions. In colonial North America, deprived of human contact
for weeks if not months, coureurs de bois christened the machines that
shared their daily toil: their rifles. Today’s suburbanites fuss over
their cars, New Agers talk to plants, Californians adopt their pets,
Japanese marry virtual Nintendo girlfriends, and kids worldwide
treat their teddies and dollies as if they were responsive and alive.
What gives?

guilty as charged

Stung by people’s repeated failures to take eliza less seriously than

they should have, Weizenbaum credited their readiness to project
a mind where there was none to the fact that they did not under-
 ToM 135

stand computers. It takes a super-sceptical person, he wrote in Com-

puter Power and Human Reason, to resist the temptation to anthro-
pomorphize the machine’s performance. Most people haven’t the
foggiest idea how algorithms work, and so they invariably bring to
them “their model of their own capacity to think.”3
In Gödel, Escher, Bach, Hofstadter adopted a similar stance, cred-
iting our penchant for attributing minds to machines to gullibility.
The implication is clear: ai experts like Weizenbaum’s assistant or
Ed Feigenbaum, versed in the intricacies of computer or algorith-
mic design, could never fall for eliza’s or PaPeRo’s cheap tricks.
Obviously, such explanations do not explain anything, given the
overwhelming evidence that we are primed to respond to facial
and verbal cues as evidence of minds. Oscar Wilde’s wisecrack about
being able to resist everything except temptation – in this case,
temptation to anthropomorphize robots, computers, and other
inanimate entities – rings as true as ever.
Given that we are all guilty as charged, the big question is: Where
is the source of this “guilt”? The answer is in one word, evolution,
and in two and a half words: Theory of Mind (ToM). Our automat-
ic projection of intentions onto the interactants we interact with is
the work of a suite of evolved adaptations to detect minds, a must-
have in such a highly social species as Homo sapiens. Turning the
tables on Weizenbaum and Hofstadter, our mind-seeking and mind-
seeing is not only perfectly natural in the biological sense but also,
as we are discovering, it is the key to human-type communication.
Every normal person relies on being able to project mental states
onto other beings and on having mental states attributed to one-
self as part of making sense of one another’s behaviour. So instinc-
tive, so universal, and so successful is this process that it is clearly a
domain-specific adaptation hard-wired into the brain rather than
an incultured skill. Studies of apes, children, twins, and patients
with chronic or acquired neural damage reveal the essential ele-
ments of its structure, as remarkable as they are commonplace.
The central aspect of ToM is our astonishing and astonishingly
reliable ability to read people’s minds – or, more precisely, to read
minds off people’s behavioural and linguistic cues. This ability is
part and parcel of our everyday psychology by means of which we
account for other folks’ actions in terms of beliefs and desires.
Each of us has our own little interaction engine-that-could, dedi-
cated to detecting intentions in actions. Anything that meets its
136 From Literature to Biterature

fairly primitive standards, from eye contact to pointing gestures, is

primed to be apprehended as evidence of another mind.
In sum, more than a facilitator of human communication, ToM
is actually its enabler. It enables us to perceive others, be they
human, animal, or even machine, as intention-driven agents and as
a result to frame what they do or say in terms of goals and purpos-
es. If you have ever wondered why human communication is in
general so effortless and effective, it is thanks to this instinctive pri-
vate detector, always on the case, always sleuthing for verbal and
non-verbal cues.

eureka

We love to share information about what we think and feel. But we

succeed in doing so only thanks to the shared pre-sets for intention
and attention. You and I could never communicate via this book if
it were not for this prosocial pre-set and for the ToM module that
enacts it. Together with the ability to appreciate each other as
intentional agents, our ability to form a joint-attentional frame is
the social-cognitive basis for our book-mediated “conversation” –
and for all human communication, period.
The evidence from human ontogeny and primatology is com-
pelling in this respect. Underscoring its fundamental role in our
species, the capacity for mind-reading crystallizes in all normally
developing infants as young as nine to twelve months old. As soon
as they begin to point referentially, say, to their favourite toy, they
keenly attend to the participating adult’s grasp of what is going on,
usually by means of eye contact. Strikingly, if the adult responds by
showering them with attention but ignoring the toy, they persist in
pointing to it. Do not attend to me, you big oaf, attend to what I
want you to attend to!
Psychologist Michael Tomasello summarizes these findings thus:

From almost as soon as they begin pointing infants understand

the most important aspects of how human-style, cooperative
communication works – they communicate on a mental level,
within the context of common conceptual ground, and for
cooperative motives – thus displaying something approaching
the complete social-cognitive infrastructure of mature coopera-
tive communication.4
 ToM 137

But while infants acquire the ability to engage others mentally

already in this prelinguistic phase, apes never develop it. They cer-
tainly do not lack the perceptional and motor skills to follow, for
instance, the direction of a pointing finger. But they just do not get
it. They see the finger and they follow its extension with their gaze,
but no light bulb lights up in the Eureka moment when they ought
to realize that it is meant to mean something. In contrast, in all ordi-
nary one-year-olds this joint-intentional light bulb lights up without
fail and brightly.
Significantly, apes are quite able to perceive other agents as inten-
tionally motivated. Experiments demonstrate that they ably distin-
guish good-faith attempts thwarted by acts of God from situations in
which somebody stiffs them for no reason. If there is a tasty morsel
in a bucket that just keeps slipping out of the zoo-keeper’s hands,
the ape will wait patiently. But it will raise a ruckus if the keeper
begins to hand it over and stops arbitrarily. Unable or unwilling?
Our evolutionary cousins cope with this critical distinction very well.
Manifestly, both apes and human infants recognize that agents have
goals. Both grasp that agents strive to bring the state of the world
closer to their internal (volitional) representations of it. Both rec-
ognize agents as rational, in the sense of making allowances for
their motives. Unable? Okay. Unwilling? Time to hoot and display.
But this is where the parallels end. And this is precisely where
human-level communication kicks in. Whereas the synchroniza-
tion of joint attentions does not emerge in apes at all, in little peo-
ple it emerges before the first year of age, never to fade away. Just
think about it. Our everyday conversations are chock-full of appar-
ent dead ends, such as:

“Wanna go out?
“My mom is sick.”

We do not get logically or semantically from an invitation to a date

to a parent’s illness – and back. Except that in shared cultural
space, it is tacitly understood that ill parents need attending to.
The operative word is “tacitly.” There is absolutely nothing in this
mini-dialogue to suggest how to flesh it out. We have to bring this
knowledge from the outside, from the common background that
enables this and an infinity of other social exchanges: verbal, ges-
tural, or even wholly implicit. Where is the missing part of this
138 From Literature to Biterature

exchange? In the context, obviously. But such context is never

all-encompassing. Indeed, for practical reasons, it could not be.
Instead, it is selected with a view to what is reciprocally assumed to be
relevant to a given conversation.
It is Savile Row communication: we tailor not only the message
but the context to fit the listener.

meaning

Given the very structure of the communicative process – its inter-

dependence, reciprocity, recursiveness, and reliance on explicit
and tacit common ground – the assumption of shared intentional-
ity, in itself deeply implicated in ToM, is the only one that makes
sense of the whole set-up. Take even such a basic part of everyday
conversation as irony. It entails stepping into each other’s shoes to
ensure that the message gets across. You monitor your friend’s
attention to see if she is getting it, your friend monitors yours to
make sure she is getting what you want her to get, and vice versa.
The guiding principle behind this turning the tables on each
other was first described by Paul Grice in a 1957 paper called
“Meaning.” Taking stock of what is going on when people commu-
nicate with each other, Grice homed in on intentions. He called
them reflexive intentions, for the simple reason that they are intend-
ed to be recognized as having been intended to be recognized. A
communication is successful if the meaning intended by the speak-
er (or writer) is identified by the listener (or reader) by recogniz-
ing the reflexive intention to express it.
Mind-reading is once again at the heart of the process, as it is in
every communicative exchange, from water-cooler gossip to the
baseball coach’s pantomime to the Turing test. In all cases, partici-
pants have to parse for intentions to generate relevant inferences
on the way to generating relevant conversational gambits. After all,
generating only logically correct inferences is not enough. They
may be correct but irrelevant – accurate but inadequate. Much like
chessbots, chatbots need to prune their search space so that,
instead of chasing all possible conversational gambits, they home
in on the relevant ones (allowing that at times it may be relevant to
be irrelevant, say, out of boredom, antipathy, or just whimsy).
If any aspect of conversation is non-algorithmic, it is being con-
textually relevant. The whole thing is analogous to the process
 ToM 139

whereby readers make inferences about what they read, just as you
are doing right this very minute. We continuously and automati-
cally flesh out what is printed on the page using empirical – or
when needed, counter-empirical – assumptions about the world,
guided by what we understand of the author’s intentions. We do
not need to be told, for instance, that people see with their eyes,
that rain clouds carry rain, or that in Lem’s most famous fiction,
Solaris, an almost instantaneous travel between distant regions in
space is possible. All this, and much more, are givens.
But at the same time we never invoke all possible inferences that
could be inferred, most of which would be utterly beside the point.
Think back for a moment to La Chanson de Roland. It is true that,
like its author, its hero is a mammal which is viviparous, pneumo-
branchiate, and placental. But make this point in your compara-
tive-literature term paper and watch big question marks bloom on
the margin just because your non sequitur contributes little of
value to the appreciation of the epic qualities of the poem.
Analogously, when we converse with a machine during the Tur-
ing test, we want it to generate not only true but relevant com-
mentary about whatever happens to be on the conversational
menu. What does it mean for computhors? Do they need to devel-
op ToM or only understand how it works? Do they need to develop
emotions or only grasp the fundamental role they play in our lives?
And how would we know computhors had emotions in the first
place? Do machines that evince all signs of solace and consolation
offer genuine solace and consolation or just indulge in gratuitous
mimicry?

elementary, dear wason

Historically speaking, biology and then sociobiology directed their

attention first to individuals and then to individuals in ecosystems,
all the while grossly neglecting the features of individuals that
enabled them to coexist with other individuals in collectives of
individuals. Be that as it may, Homo sapiens is above all a social ani-
mal. Our natural habitat is not so much a savannah or a modern
city as an assortment of blood kin, friends, and neighbours, them-
selves immersed in an ethnic or socio-cultural tribe at large.
Mutual aid, information sharing, and coordinated action are nec-
essary even for our basic survival, not to mention that normal
140 From Literature to Biterature

human psychological development hinges on being socially embed-

ded. This is why our minds come pre-set for prosociality, ToM, emo-
tional contagion, and other traits that, phylogenetically speaking,
would not make sense in perennial hermits. This is why we are pre-
tuned for identifying and evaluating facial expressions, postures,
gestures, and behaviours in terms of mental states – for interpreting
the external in terms of the internal.
These cognitive skills are ingrained in the subconscious, as they
have to be in order to be so effortless and efficient. But subcon-
scious effects can be teased out with the right wedge. Consider the
notorious Wason test. Albeit not invented to test for evolved socia-
bility, it does so anyway. At its heart lies a little brain-teaser that,
framed in two vastly different ways, elicits vastly different results. In
the abstract, the puzzle stumps almost everyone. But in the social
form, as a narrative about detecting cheaters, it presents few prob-
lems. How could this be?
You can try this at home. Imagine four cards, each with a number
on one side and a colour on the other. The particular cards in front
of you display 5, 2, red, and black. The clue is this: If you see an even
number, the other side is red. Now, what is the smallest number of cards
you need to turn over to determine if the rule holds? Not so easy?
Don’t worry, even after reflection, most people get it wrong.
But now enter the world of people. For example, imagine that
you are told, If you use the car, you gotta fill the tank afterwards. Makes
sense, right? Now you face four cards that describe the following
four situations:

Homer did not use the car.

Marge used the car.
Bart filled the tank.
Lisa did not fill the tank.

The question is exactly the same: what is the minimum number of

cards you need to turn to spot free-riders? Elementary, dear
Wason: the second and fourth cards, exactly the same as with the
first problem.
In a similar vein, in 2005 Joseph Henrich et al. reported on a
series of experiments around the world testing how deeply we have
internalized norms for sharing and fairness. Their tool was the ur-
experiment of prospect theorists/behavioural economists: the ulti-
 ToM 141

matum game. One player is given a sum of money which he must

split with another person whom he has never met before (and
never does during the experiment). The game is played only once,
with no possibility of collusion and no chance for cooperation
evolving over time.
The trick is that the first participant makes a “take it or leave it”
offer to the second player, who has only two options: assent or walk
away from the deal. If the former, both keep whatever the split was.
If the latter, both forfeit what was in the pot. According to classical
economic theory, the whole thing is a no-brainer. Rationally, the
first player should offer his absent accomplice only a token amount.
As a utility maximizer, the other should accept the deal, since oth-
erwise he gets zilch.
In real life, things look totally different, with players typically
offering something like an even split. Indeed, offers where the first
player keeps more than 60–70 per cent are almost universally
rebuffed as unfair. The margin varies from culture to culture, and
there are outliers who offer more than half as well as those who do
not accept these overly generous splits, but the overall pattern is
overwhelmingly that of fairness and reciprocity. We have internal-
ized the norms very deeply indeed.
And the punchline? When people are told that they are playing
against the computer, they immediately stop extending and rejecting
one-sided offers. Instead, they play by the classical economics text-
book: they try to maximize their take. It is just as I said: change the
information available to the players, and you change the game itself.

biology in action

What does this tell us about thinking computers? Human minds

are primed to engage with minds – so much so that a pair of out-
sized eyes, a toothpaste grin, and especially an apparent ability to
converse can generate an impression of a mind-like pattern. Minds
are also primed to regard other minds differently from mindless
entities. All the same, the tt exposes the limitations of a gut feel-
ing that having a mind is an immanent property. In some contexts,
whether a mind is really out there or whether it is only projected
by a mind-seeking module is a distinction not worth making.
Although human minds are, naturally, always alive, life and sen-
tience do not logically or even biologically entail each other. No need
142 From Literature to Biterature

to look far for life without consciousness: plants provide a ready

example. Those who would object to calling plants alive, even given
the coordinated predatory activities of species such as Venus flytraps,
would presumably acknowledge pleuromona (a kind of highly sim-
plified bacteria), standard-issue bacteria, phages, or amoebas like
Dicty. At the other end of the continuum we can meaningfully speak
of consciousness sans life, presumably in computers and robots.
Seeing how grappling with the nature of consciousness is worse
than grappling with a greased pig, you might think that conceptu-
alizing life is a different story altogether – a hard biological, med-
ical, or even forensic fact of the matter. Oddly enough, no one can
say for sure what life is. Even as all extant definitions of life aim to
carve nature at its joints, there is always some conceptual gristle or
even meat left on the slab.
Understandably, most efforts go toward finding an intersection of
traits presumed to be shared by all living organisms. One of the
many scientists who pinned hopes on biochemistry was Norman
John Berrill who, in his classic triple-decker Biology in Action (1966),
ticked off fundamental functions such as growth, autonomous
motion, metabolism, and self-replication as characteristic of life. In
general, biologists tend to focus on prescriptions of shape and struc-
ture, metabolism of a range of chemical compounds, homeostatic
adjustment to a changing environment, and sexual reproduction.
As is to be expected, all are slanted towards carbon-based life.
The problem is that exceptions to these purportedly fundamental
functions are not difficult to find, starting with viruses, worker bees
(which cannot mate), or families of female-only wasps. There are two
ways out of the bind. The first is to hang onto the conceptual core
while papering over recalcitrant cases with a patchwork of ad hoc
exceptions and conceptual epicycles. Certain variations or traits are
ruled as inadmissible, with no generally acceptable justification for
these decisions. Needless to say, machine life does not stand much
chance of getting into the club.
The alternative is to shrug the biochemical slant altogether in
favour of processes independent of organic composition and/or
recombination. In most versions, these have to do with energy and
information processing, structuring and organization, and emer-
gence of coordinated response to the environment – as in the com-
puter game Life, famously studied by John Conway (available online
if you care to try it at home). By definition, the second approach
 ToM 143

In most versions, these have to do with energy and information processing,

structuring and organization, and emergence of coordinated response to
the environment – as in the computer game Life, famously studied by John
Conway (available online if you care to try it at home).

admits of life in computers as well as in still other life-forms, perhaps

on one of the thousands of known exoplanets.
At the end of the day, everything depends on how anthropo-
centric and anthropomorphic we will choose to be in our future
legal and medical statutes. Even as I type these words, computer
life continues to be redefined as technology races on, much as life
and death continue to be redefined in our own species. The term
“clinical death,” to take just one obvious example, did not exist
until a few decades ago, acquiring reference only when the med-
ical sciences secured a toehold in what used to be prohibited
territory.
Where historically death was regarded as a monolithic cessation
of life, the twentieth century has splintered the monolith. Now
clinical death itself, defined as cardiac and pulmonary arrest (no
heartbeat and no respiration), has been rendered obsolete by tech-
niques and technologies ranging from cpr and life-support systems
to epinephrine injections, defibrillators, and pacemakers. As a
result, in the United States today, as per the Uniform Determina-
144 From Literature to Biterature

tion of Death Act, mds and coroners pronounce “biological” or

“brain” death defined by the cessation of all electrical activity in the
brain.

driving range

The Nobel-winning neuroscientist Gerald Edelman has argued that

thinking as we know it is the outcome of so many contingent
processes and forces, including natural selection, morphological
evolution, and biophysics, that it could never be embodied in a
computer program. His argument is pellucid, his conclusions less
so. Does Edelman suggest that computers could never think? Or
only that computers could never think as humans do? The former
is only an opinion, which, in my opinion, will be refuted by future
developments. The latter is true, but so what? The rtt and Ulti-
matest take care of that.
Generally speaking, people have a big advantage over machines
insofar as people come preloaded with an instinctive grasp of how
the basic physical and causal relations work in the world. This
understanding permits us to operate with something like a Reality
Principle, plugging the background into all kinds of tacit commu-
nicational contexts. But we have an even bigger advantage over
machines: we want things, and we usually go after what we want.
Computers, on the other hand, have no desires, including no
desire to manipulate the world in a way that will bring it closer to
their desires. In short, they have no volition.
Hence the big question: what will drive computhors? What will
make them want to write novels or not want to write them? Emo-
tions aside, biterary computhors will almost certainly have to have
some form of algedonic control to build an internal hierarchy of
preferences that will mimic rudimentary drives. The engineering
aspect of this may be difficult, but the behavioural aspect is not.
Unsecured energy supply, disruption (unplugging) or corruption
(virus) of processing, and compromised integrity of the system
would be tagged as negative. Continuance and more efficient task-
ing would go on the plus side of the program’s ledger.
Surprisingly, John McCarthy, for decades a central figure in Arti-
ficial Intelligence (he coined the very name), speculated in 1990
that it will not “be advantageous to give programs the kind of moti-
vational and emotional structure we have inherited from our
 ToM 145

ancestors.”5 This was the same McCarthy who had for a long time
contended that the way toward machines who think was via
machines with common sense. As every student of biology, ecology,
and evolution knows, however, common sense does not develop in
a vacuum.
This is precisely why, after decades of blundering in the desert,
the so-called situated-ai school has come to realize that general
intelligence must be embodied and embedded in the real world.
Odd as it may sound, cognition is not just in our brain but is dis-
tributed all over the body and smeared all over the socio-cultural
world around us. Intelligence is two-way traffic with feedback – the
world impacts the evolving system and the evolving system impacts
the world as it strives to survive. But what will make a computer
strive to survive? What will make it want to want?

insula cortex

Much is said about the parallels between the cognitive sciences

and cognitive psychology, relatively little about the parallels
between artificial intelligence and evolutionary psychology. Mole-
cular computing and genetic algorithms apart, much of evolu-
tionary science is too imprecise to suit the programming formal-
ism favoured in ai, and especially in American ai. It is no accident
that the latter shunned fuzzy logic years after it had been put to
work in Europe and Japan. This is unfortunate, since there is quite
a bit to learn about intelligence from sociobiologists and evolu-
tionary psychologists.
Human cognition is not entirely about rational deliberation, or
not even mainly. Much of what we like to label “cognition” is real-
ly subcognition. To optimize performance, evolution has removed
the awareness of most of the cognitive machinery it has bred in us
so that we could get ahead in life without pausing to think about
everything. As a result, the majority of human mental operations or
processes are executed without deliberate reflection. To a large
degree we are black boxes to ourselves.
Modularized subcognitive processing is in many ways the default
mode of human brain power. At the same time, so much of our
cognition is driven by emotion that to function effectively, we must
function affectively. In what can rightly be described as a tectonic
shift, emotions and feelings have finally over the course of the past
146 From Literature to Biterature

twenty years been acknowledged as integral components of the

machinery of thought and reason (in the process, emotions, or
somatic affects, have also been distinguished from feelings, or
mental representations of emotions).
As biological regulatory mechanisms, emotions could not have
arisen in isolation from other life processes. They are implicated, in
fact, in all kinds of adaptive behaviours, from mate-seeking and par-
enting down to thinking and decision-making. They keep us fixed
on long-range goals and kick the somatic power train into overdrive
in the presence of immediate stimuli. There is a time to reason and
time to act, and when emotions grab the wheel, that time is now.
Chemically mobilizing our minds and bodies, they tunnel our vision
so that it fixates on the immediate consequences, regardless of what
may come later.
This is not to say that emotions do the thinking or the decision-
making for us. But, in a head-on collision with ultra-rational mod-
els of agency rooted in classical economic theory, emotions play a
major role in cognition. In fact, as mri scans of players in the ulti-
matum game graphically show, we often think in emotions. Unfair
offers light up the insula cortex – the very area of the brain trig-
gered by painful or malodorous stimuli – like a Times Square
Christmas tree. Thinking about unreasonable splits, players are lit-
erally disgusted.
Just because emotions are underwritten by evolution, it does not
mean that there is no cultural variation in how they are manifest-
ed. On the contrary, much of our emotional repertoire is mediat-
ed by communal rules of display. Also, like any other evolutionary
mechanism, emotions can be adaptive in one set of circumstances
and not at all in another. But, one way or another, they are there
for a reason, and that reason has to do with driving the open-ended
game of survival.

on being redundant

Today’s serial systems are critically vulnerable to external pertur-

bations. Skip a few digits from the bitware, or a remove a few chips
from the builtware, and you could as well drop the whole system
from a plane sans parachute, the way it turns chaotic and churns
out gibberish. The problem lies in the architectural and function-
al minimalism so prized by designers, which does not make room
 ToM 147

for redundancy and self-repair. This is in contrast to human brains,

which are not only massively parallel but massively redundant in
structure and operations. Naturally, this is also what makes human
thinking so scattershot and far from always rational (though always
capable of rationalizing).
If ai is to learn from evolution, we might need to revise our atti-
tude to redundancy and stop banishing it in the name of perfor-
mance optimization. If machines are to promote their own sur-
vival, they will almost certainly have to forego streamlining.
Redundancy may need to be incorporated into the design as part
of the homeostatic drive toward promoting resilience toward
changes in internal and external environments. From the engi-
neering standpoint, after all, our brains are anything but designed
for maximum efficiency. Indeed, they could not be by the very bio-
logic of evolution.
Brains may resemble general-purpose computers in their synap-
tic relays or in hard-wired modules being dedicated to running
subroutines, but they are emphatically not what a traditional or
even rational engineer would design. In place of fixed hierarchy,
dedicated tasking, and optimized efficiency, there is a kilo and a
half of neural jelly that is more redundant (or full of exaptations)
than most people realize. Yet, in the real world, this bric-à-brac of
suboptimal solutions outperforms the more compressed and effi-
cient, but also more rigid and vulnerable, supercomputers.
Of the hundreds of examples of redundancy in our brains, per-
haps the most evocative come from studies of brains that work any-
thing but normally. An estimated 10 per cent of people with lesions
in the vicinity of Wernicke’s area display the type of verbal impair-
ment (aphasia) typical of patients with damage in the physically
discrete – although linked by nerve fibres – Broca’s area. The infer-
ence is clear. There is partial overlap between the functions of
these two areas which normally produce stunningly different types
of aphasia. Broca’s reduces verbal ability practically to zero. Wer-
nicke’s aphasics speak fluently and grammatically, but their speech
amounts to semantic gibberish.
It is as if evolution invented packet switching before it was rein-
vented by human engineers to improve digital information trans-
fer. Packet switching was a huge improvement over circuit switch-
ing – think back to the analogue telephone networks – which had
to preserve the serial order of the transfer. When you speak with a
148 From Literature to Biterature

friend over the phone, you sure want the beginning of the sen-
tence to arrive before the end. Packet switching changed the need
for linear order. On the Internet, data is disassembled into discrete
packets and sent helter-skelter via any channel available. The pay-
off? Fantastic increase in reliability and speed: if any part of the net-
work’s bandwidth is filled to capacity, the packet can be relayed
through another.
Redundancy is survival oriented. Tantalizingly, as became appar-
ent during the first computer simulations of neural networks, it is
also inbuilt into parallel processing. Back in 1954, a couple of mit
scientists trained a network of only 128 neurons to recognize pat-
terns. Impressive as the feat was, they also stumbled on something
potentially even more important. Random removal of up to 10 per
cent of an already trained network had next to zero effect on its
effectiveness. When it comes to “survivability,” then, there is clear-
ly nothing wrong, and a surprising lot right, with redundancy and
neural nets.
 8

Y2K+

Methuselah; Yes!; An Amphibian with a Stick; Ellipses in This

Picture; Software for the Soul; Parallel Paradigm; Erlang and
Mira; DelFlies; Eruntics; Cogito Paradox

methuselah

In the day when biotech start-ups germinate like fungi, public

relations spinners go to absurd lengths to brand their names with
the public. Whatever they may lack in judgment they more than
make up in hubris. Recently Human Genome Sciences enjoyed its
fifteen minutes in the limelight with a prediction that by 2100 the
sole causes of human death will be accidents, murder, and war. In
other words, thanks to the yet-to-be-found wonders of regenera-
tive medicine, old age, along with genetic degeneracy, will cease
to be a lethal factor (cue in Louis Armstrong’s “What a Wonderful
World”).
In a prospect as giddy as it is frightful, senility and mortality are
to become “soluble” in our children’s lifetime. And why not? Adam
is said to have lived to 930 years, and Methuselah, the oldest man
mentioned in the Bible, was 969 when he gave up the ghost. We
are all dying to live longer and look younger, and the American
Society for Aesthetic Plastic Surgery is only too eager to oblige. As
their statistics show, every year America forks over billions of dol-
lars to have some ten million fat deposits vacuumed, eyelids
reshaped, noses redone, crow’s feet and lips shot with Botox, big
boobs reduced, droopy ones lifted, and all others beefed up with
silicone.
Perhaps conditioned by their own advertising glitz and blitz, the
media and even parts of the academic sector appear to have suc-
150 From Literature to Biterature

cumbed to the illusion that science and technology are like a year-
round Christmas where, just as we tell little children, if we wish for
something very much, it is going to happen. Unfortunately, this
tendency, which without much exaggeration can be called wishful
thinking, is particularly endemic among writers on computers in
all their incarnations: Artificial Intelligence, robots, computational
algorithms, the Internet, and so forth.
Their starting point is the indisputable fact that much of the
world has become saturated with industrial machines, computers,
telecommunications, and interactive information databases. This
single trend is subjected to linear extrapolation, which is then used
to underwrite one brave new scenario after another. And so since
the 1950s a steady stream of wide-eyed hype has been promising
not just automated but autonomous kitchens, smartly conversing
robots, creatively thinking computers, general problem-solving
software, self-evolving machines – all just round the corner or even
already in full view.
How little has changed since then can be testified to by picking
books on computers practically at random. Take one of the primers
on electronic literature mentioned briefly in the introduction to
this book. Without batting an eyelid, its author states that people in
technologically advanced countries “are literally being reengi-
neered through their interactions with computational devices.” Lit-
eral hyperbole of this kind is, of course, nothing new. Back in 1983,
one-time Pulitzer-winning science journalist Joel Shurkin chipped
in from Cloud Nine that “most computer scientists now think that
the Turing test is too easy,”1 proving that Pulitzers ain’t what they
used to be.
Naturally, it is difficult in a single chapter to do justice to this
global trend. And so, in what follows, I will confine my remarks to
another blast from the past, Geoff Simons’s Are Computers Alive?
Evolution and New Forms of Life (1983). While utterly forgettable on
its own merits, it is perfectly representative of the work of succes-
sive generations of techno-prophets who would pave the road to
the future with the bodies of those who look askance at visions of
crystal palaces of instrumental might. Its rhapsodic claims about
technologies that to this day remain on ai’s wish list deflate the
bluster of present-day Nostradamuses better than any amount of
reasoned critique could.
 Y2K+ 151

yes!

If best-selling science lite such as Life on the Screen: Identity in the Age
of the Internet exemplifies hype in the 1990s, and if futurology lite
such as Beyond Human: Living with Robots and Cyborgs exemplifies
hype in the 2000s, Are Computers Alive? exemplifies hype on steroids
for all times. From page 1 it is clear that the title question is pure-
ly rhetorical. The author’s answer is an unequivocal Yes! with stiff
thumbs up on either side. Both the answer and the sleights of hand
employed to fabricate it are, however, patently deceitful.2
Yellow journalism blurs reality with fantasy most of the time, but
Simons stretches rhetoric to the point of snapping at every oppor-
tunity, beginning with his preface. There is a sense, he proposes, in
which robots are an evolving life form. What exactly is such cory-
bantic metaphorizing supposed to accomplish? There is also a
sense in which human beings are evolved fish, but where is the cog-
nitive payoff from asserting that Are Computers Alive was authored by
an evolved fish?
Ignorant hyperbole and hyperbolic ignorance are equally in evi-
dence in the sole paragraph devoted to the Turing test. If a machine
“behaves as if it is human in all the relevant circumstances,” writes
the author, “then it is human.”3 This is about as stupid a statement
as can be made about a subject that even by the 1980s had gener-
ated a truckload of analysis and commentary. The tt emphatically
does not test for, and is therefore in no position to demonstrate,
the humanity of computers. All it does is sketch a sensible proce-
dure for collecting evidence for recognizing human-level thinking
in machines.
Going on to remark that many biologists regard reproduction as
an essential element of life, Simons glosses over the fact that repro-
duction is just one element in the conjunction of functions regarded
as constitutive of life. Worse, he fails to note that biologists talk about
sexual reproduction. From his single wonky premise he proceeds to
derive an astounding conclusion: “Machines are reproducing them-
selves”!4 His evidence to back up this extraordinary proposition is as
banal as they come: a Fujitsu Fanuc plant in Japan manufactures
industrial machines under reduced human supervision.
A few pages later he drops all caution and proceeds to enumer-
ate three classes of life on earth: plant, animal, and machine. This,
152 From Literature to Biterature

remember, in 1983! He allows that some people may not find the
notion of computer life to their liking, but others, with Simons
leading the way, should “find it both acceptable and exciting.”5
This attitude is as patronizing as it is alarming. One may find the
idea of thinking and living computers perfectly acceptable – I
count myself among those who do – but remain unaffected by any
sense of excitement.
As a matter of fact, if our socio-technological history is anything
to go by, excessive excitement can only lead to excessive hubris,
much as it did with the wonder-drug thalidomide, silent-spring
ddt, dioxin-toxin pcbs, miracle-mineral asbestos, Dow Corning
breast implants, and a legion of other fateful inventions. But
Simons is not done. Arguing that machines outclass many acknowl-
edged life forms in mental abilities, he backs it up with another
clincher: “Frogs cannot weld car bodies as well as robots.”6

an amphibian with a stick

Time to slow down, take a deep breath, and sort out these grandiose
errors of fact and category. Let us begin with the centre-piece of
Simons’s thesis: life. The fact that machines can weld, play chess, or
prove theorems does not prove that they are, or may one day
become, alive. In the gulf between rana temporaria and a twentieth-
or even early twenty-first-century computer are a myriad of orders
of structural and functional complexity, the bridge between them
hinging ultimately on their ability to function in the world.
All carbon-based life is tied to homeostasis, itself tied to survival,
itself tied to descent with modification. Without exception, all
frogs, whether the toxic rainforest species known as poison-dart
frogs or the garden (actually, pond) variety ones, perpetuate their
existence both in the behavioural and genetic sense. Machines do
not. Of course, they may yet come to do so in the future. The rudi-
ments of such homeostatic drives can be observed in some of to-
day’s energy-troping (solar or electrical) automata.
But there is no computer today that would even approach the
frog in its ability to survive in the world and reproduce. If you come
at an amphibian with a stick, it will not just sit and stare but will
scuttle away. On the other hand, if you proceed to pulp your com-
puter with a brick or crush an industrial machine with a bulldozer,
neither will initiate any sort of evasive action. Recognizing these
 Y2K+ 153

In the gulf between rana temporaria and a twentieth- or even early twenty-first-
century computer are a myriad of orders of structural and functional
complexity, the bridge between them hinging ultimately on their ability to
function in the world.

holes in his pro-life rhetoric, Simons calmly bites the bullet and
contends that machines do display survival-oriented behaviour.
How so? By being equipped with backup circuitry and subsystems
154 From Literature to Biterature

that take over in case of malfunction, thus perpetuating the

machine’s “life.”
This paralogic is equivalent to claiming that your copy of From
Literature to Biterature is biologically alive because it comes with a
cover that prolongs its shelf life. Some books are actually printed
on non-flammable plastics, which would make them, on this pro-
posal, even more survival oriented and thus more “alive.” But why
stop there? Typos can be likened to textual mutations; textual edi-
tors to dna repair crew; citations and intertextuality to recombi-
nant dna; new editions and translations to niche differentiations;
and so on without end. Metaphors of this type are as cheap to pro-
duce as bereft of cognitive value.
Unrepentant, Simons insists that concepts such as machine evo-
lution, machine mutations, and successive generations of machin-
ery are “not simply loose metaphors or allegories” but “literal
descriptions.”7 Were this true, there would be no reason not to
speak in the same way of the biological evolution of the Nike shoe
or of the latest generation of Crest toothpaste. So far, thankfully, no
one tries to sell Nikes as a life form or the evolution of toothpaste
formulas as anything but a figure of speech.
Once more: although we may yet live to see thinking or alive
desktops in the future, there are no grounds for believing that they
exist at present. The fact that some individual human skills or sens-
es have already been duplicated to various degrees does not mean
that there are any systems that employ them all in a concerted and
sophisticated fashion, as a consequence producing goal-oriented
and survival-oriented behaviour. In short, the potential for com-
puter consciousness and life needs to be separated from the adren-
alin rush of a technophiliac.

ellipses in this picture

What is most unsettling about Simons’s zealotry is his apparent con-

viction of machines’ near omnipotence and universal fitness for
replacing human agents in all stations of life. Few passages bring this
out better than his rhapsodic picture of computers triaging and inter-
viewing patients in hospitals and health clinics. In his view, replacing
doctors and nurses with expert programs is highly desirable insofar as
it streamlines the entire interaction. In Simons’s picture, the machine
 Y2K+ 155

follows the sequence of flow-charted questions, and the patient is

required to give only simple answers, usually “yes” or “no.”
The ellipses in this picture are glaring. Nothing is said about the
mental and emotional anguish of a suffering human being. Not
one comment about the commonplace comfort that stricken souls
often draw from human presence. Not a word about the fact that a
critical factor in this comfort has always been a chance for patients
to tell their story in their own words, potentially alerting the med-
ical staff to asymptomatic and subclinical episodes. Now this vital
self-translation of one’s subjective experience of affliction and pain
is to be replaced by a string of laconic answers to expert questions.
Some dimensions of ailing persons’ well-being simply require a lis-
tener who will hold their hand and seek to understand their pain
and confusion. Alas, objectification, a term much overused in recent
cultural history, is the only way to characterize what happens to a suf-
fering person in the medical scenario from Are Computers Alive? This
is not for a moment to accuse its author or the present crop of ai
publicists of ill-intentions toward patients, but rather to remind our-
selves that treatment and healing depend on more than computer-
ized, streamlined throughput.
There is no doubt in my mind that in the not-so-distant future,
machine diagnosis and therapy will become an inexpensive and
error-free alternative in most health centres. Robotic nursing may
also become the norm, especially if the Japanese health-care system
is any indication. In many ways its systemic traits – aging popula-
tion, escalating costs, shortage of qualified personnel – presage
where the rest of the world is heading, with America at the head of
the pack. Nursing help is costly, but intelligent robots can be rent-
ed on demand to provide dependable care day and night.
Cultural differences between the United States and Japan are, of
course, colossal, and will determine the level of acceptance of robots
as hands-on caregivers. Interestingly, hospital trials suggest that
human form is not only unnecessary but often undesirable in robots.
Too robot-like, however, and patients ignore them and their prompts.
Facial mimicry and human-like speech produce the opposite effect,
on top of being vital factors in making people comfortable around
machines.
It is not the specific shape of a robotic frame and the degree of
facial plasticity, however, that concern me but the idolatrous atti-
156 From Literature to Biterature

tude that frequently accompanies technology. Faster and more effi-

cient processing does not always translate into better caregiving,
especially in the context of palliative effects of human kindness
and attention on suffering human beings. Chicken soup for the
soul calls for more ingredients than a clockwork chicken.

software for the soul

Tirelessly, silently, unerringly, computers toil at whatever we tell

them to do. They investigate logical conundrums and model sci-
entific phenomena of a complexity we could never dream of
fathoming on our own. They diagnose blood infections and
dozens of other ailments and diseases more reliably than mds.
They assist with surgeries, conduct molecular research, regulate
traffic, advise with purchases, plot travel itineraries, scour data-
bases for information, detect patterns of criminal activity, opti-
mize livestock’s nutrition levels, and perform a myriad of other
functions indispensable to the daily administration of contem-
porary society.
All this activity is external with regard to human beings – social
rather than psychological, physical rather than emotional. But
research in many areas of psychotherapy indicates that computers
can also help heal souls. For a long time online therapy has been
dismissed as a trendy fad straight from Newrotic York and Holistic
Hollywood. Not a few psychiatrists and psychologists have voiced
reservations about the virtues of hi-tech analysis. It is one thing to
use computers in surgeries or diagnostics, they remonstrated, but
probing souls and manipulating minds are propositions of a dif-
ferent order.
Yet there is growing evidence that, when judiciously practised,
binary therapy can be as effective as human – indeed, at times even
more effective. In 1994, in a provocatively entitled article, “Soft-
ware for the Soul,” Der Spiegel carried a story of a person in need of
someone in whom she could confide:

“Often I stay in bed the whole day,” she complained. “I have

the feeling that nobody likes me.” She feels she is simply too
fat.
“Perhaps it is just the opposite,” suggested the response –
 Y2K+ 157

“you withdraw from others and moreover you make yourself

purposefully unattractive.”
“Surely not,” disagreed the woman.
“All right, good,” the response came back, almost a little
insulted. “What are you waiting for?”8

There is nothing in this brief exchange to suggest that the thera-

peutic dialogue, with the attendant probing, coaxing, and mentor-
ing, came from a machine. But it did, much like the entire session
with this and numerous other patients.
Today a growing line of psycho-software is advanced enough to
detect new details or new angles in twice-told tales, just as a human
therapist would. There are successful diagnostic programs to mea-
sure risk levels for patients in danger of suicide. There is software
that not only helps in diagnosing incipient depression and even
full-blown psychological crises on American nuclear subs but also
advises on resolving these episodes. Other algorithms help people
tormented by panic attacks diagnose a hierarchy of anxiety-rousing
events and overcome their phobias in virtual environments with
the help of vr-integrated software.
Indeed, apart from performing diagnostic work, computers are
more and more called upon to suggest therapeutic remedies. In
the 1990s one such expert system was created for direct use by cou-
ples in need of information or advice about their sexual behaviour
or their relationship. The system proceeds by gathering extensive
background information from the pair, while providing ongoing
feedback and every opportunity to query it on all aspects of sex.
Most couples who interact with sexpert reported it to be open,
sensitive, and even human-like.
Such consistent reaction to machines that could never pass the
Turing test suggests once again that, by the time we roll out those
that could, the test itself will have become superfluous. What then?
Will the knowledge that the computer is thinking act as a road-
block to free, uninhibited disclosure, as it may in front of human
doctors? Will patients become troubled knowing that computers
have perfect recall of every word and gesture? Will they become
paranoid that their records could be hacked? Will machines invoke
doctor-patient privilege? Will machines perjure themselves on the
witness stand? Will they suffer any consequences?
158 From Literature to Biterature

parallel paradigm

At work and at home, computers and robots are everywhere:

domestic robots, industrial robots, medical robots, traffic robots,
police robots, military robots, prison robots, space robots, nanoro-
bots – you name it. Increasingly, they are also driving our play and
entertainment, what with video games, interactive software, ani-
mated cinema, and, not least, sports. Especially in Japan, robotic
competitions like Robot Sumo, RoboCup (football), Robo Ping
Pong, Atomic Hockey, and the Walking Machine decathlon enjoy
the kind of devotion that would be the envy of more than a few
human disciplines.
All this today. But what about tomorrow – and the day after? Is
there anything in contemporary R&D to point in the direction of
thinking, writing, causally independent computhors? You be the
judge.
Since its conception by Von Neumann more than sixty years ago,
computer architecture has undergone little fundamental change.
Progress has been achieved chiefly via component size reduction
and improvements in speed (the two are, of course, directly relat-
ed). Yet the days of such advances are numbered. Even today, nano-
dimension transistors are bumping their heads into a limit of effi-
ciency measured in atoms. If not enough atoms are stacked in the
right place, insulation is compromised, allowing current to leach
away through quantum tunnelling. Too thick a layer of atoms, and
conductivity suffers.
Every time a transistor halves in size, it edges closer to the point
where it can halve no longer. Fundamental quantum laws impose
limits on how many massively integrated circuits can be printed on
a chip, especially at the rate of doubling maintained up to now
(roughly every eighteen to twenty-four months). At that pace, by
2030 transistors will have shrunk to the size of individual atoms,
which they cannot if they are to perform as transmitters-resistors.
Something has to give, and the sixty-year-long marathon will soon
be over. The finish line is in sight, and it is a solid brick wall.
Research is underway to squeeze a few Moore years out of the
present design. In 2010 the International Technology Roadmap
for Semiconductors released its annual fifteen-year roadmap for
the industry. If they are right, by 2015 the standard for transistors
will be just eleven nanometres. Another few years might be bought
 Y2K+ 159

Especially in Japan, robotic competitions like Robot Sumo, RoboCup (foot-

ball), Robo Ping Pong, Atomic Hockey, and the Walking Machine decathlon
enjoy the kind of devotion that would be the envy of more than a few
human disciplines.
160 From Literature to Biterature

with junctionless transistors, which loom on the horizon. In a relat-

ed development, in 2012 ibm succeeded in storing a single data bit
in only twelve atoms (as opposed to about a million needed to store
a bit on high-density hard disks), opening the doors to much
denser magnetic storage than on today’s solid-state memory chips.
But all these are Pyrrhic victories that stave off the defeat for a
decade or at most two. In the meantime, everything points toward
a paradigm change. Recent advances in quantum computing
notwithstanding, the key to the future lies with neural nets and par-
allel processing. Up until now, parallel computing has been effec-
tively neutered by Moore’s law (which is really not a law but a con-
jecture). Serial processors have increased in capacity and dropped
in price swiftly enough to make investment in parallel chips and
software unprofitable.
But, seeing that within a generation the current game plan will
have to be scrapped for good, specialists are in agreement that the
next qualitative improvement in performance can be achieved
only by forsaking the traditional architecture. Even as many scien-
tists continue to shrink, cool, and upgrade familiar system compo-
nents, the industry as a whole is diverting more and more resources
to a future that leads away from cpus and serial crunching and
toward massively parallel builtware and bitware.

erlang and mira

Barring the Connection Machine, a massively parallel (for its time)

computer prototype rigged together in the 1990s, up to now there
has been little interest and consequently little progress in parallel
computing. True, multicore chips have been shipped even in ordi-
nary desktops and laptops for years, but there has been no software
to take advantage of these capabilities. Times are a-becoming dif-
ferent, however. Even as talk of single processors is fading fast, talk
of dynamic systems, adaptive strategies, and distributed intelli-
gence is gaining in animus.
Does the latter sound more than a little like what evolution has
cobbled up over billennia? You bet. Evolution is a gigantic parallel
processor, a neural network that has been running in fits and starts
for almost four billion years, during which it has succeeded in
extending itself all around the globe. It did this while running at,
from the digital computer’s point of view, ridiculously slow speeds
 Y2K+ 161

and while accumulating one redundancy after another in its dna

source code.
One of the pioneers who back in the 1980s warmed up to the
idea of approaching machine learning in a manner that imitated a
hugely simplified evolutionary process was John Holland. Working
essentially alone, he seeded computer environments with popula-
tions of primitive, albeit survival-driven algorithms that evolved in
accordance with prescribed rules of selection, recombination, and
mutation. Crucially, each received a fitness measure, with repro-
duction favouring high-fitness behaviour, much like in its contem-
porary incarnation, the evolving algorithmic world of the Tierra
Working Group.
Today, neural networks, those sets of statistical-learning algo-
rithms, are poised to step from the margins into the mainstream.
Of course, the difficulties in switching from serial to parallel pro-
cessing are stupefying, magnified by our desire to oversee all stages
of computation rather than let algorithmic evolution run wild.
These difficulties are reflected in the lack of software tools, starting
with compilers to translate human-readable code into free-floating
packets of binary strings that could be disassembled and reassem-
bled at will. The other ordeal is the debugging of programs that
run in thousands of parallel streams crisscrossing and feeding back
into one another in real time.
But the way the wind was blowing was clear in 2010, at the sc10
conference – the world’s Mecca for high-end computing, network-
ing, storage, and analysis – when the subject that stole the show was
the need to redesign computing curricula and textbooks to shift
emphasis to parallel programming. This shift will take the better
part of the next decade or two, no doubt, but the drift could not
be clearer: the parallel paradigm is the way of the future, with feed-
back. For, as parallel processing becomes more widespread, it will
also become more commercially attractive, leading to more inno-
vation, more investment, and, as a result, more parallel processing.
Oddly enough, the engine of change may be social media.
Erlang, a new-generation programming language designed specif-
ically to crunch mind-blowing amounts of data, is already powering
Facebook’s crowded chatrooms. In a sweeping departure from pre-
vious “imperative” languages, Erlang is more contingent, fuzzy, and
“evaluative” in its approach to how it weighs data and anticipates
results, and as such is more human-like. Similarly designed to work
162 From Literature to Biterature

with incomplete data is ibm’s recently unveiled Mira system, whose

hardware and software are, tellingly, designed to run in massive
parallel.

delflies

As a branch of computer science and logic rather than biology or

psychology, Artificial Intelligence has heretofore concentrated on
developing and testing theories about programs rather than behav-
iours. But this too may be changing, thanks to an up-and-coming
niche in robotics that addresses itself to emulating the mechanics,
the senses, and – most intriguingly – the behaviour patterns of ani-
mals. Known as zoobots, biomorphs, or neuromorphs, these robot-
ic critters try to imitate with neural nets what bioneural nets do in
living creatures.
After decades of looking for short cuts, scientists are rediscover-
ing the adaptive wisdom of Mother Nature. Not by accident, all
zoobotic systems are rovers designed to move about in their envi-
ronment, learn its features, and negotiate obstacles that inevitably
arise in the process. Microbotic shrews, smoothly undulating
mechanical lampreys, pliant submarine jellyfish, suction-padded
geckos, burrowing bivalves, power-driven grasping octopi, wriggly
meshworms, even hovering DelFlies all point their whiskers, paws,
and tentacles in one direction: evolution.
What is extraordinary about them is not even their increased effi-
ciency in imitating natural functions, or the fact that not just
roboticists but zoologists are turning to robots to understand the
behaviour of natural organisms. It is that these mechanisms are
often built to test evolutionary hypotheses, such as, for example,
the effect of the shape of mollusc shells on their survivability. All of
a sudden, adaptive thinking is in, with efforts under way to con-
front the biggest hurdle that life had to jump over before it could
launch a multibillion-year relay race with dna as the baton: the will
to live.
In the 1970s, pioneering studies zeroed in on programs whose
operating strategies could affect their stored information – an
informational prototype of homeostasis. Trivial as it may sound,
programs were able to learn structures and concepts, such as table-
tops or arches, that were not part of their starting knowledge.
Today all eyes are on cog (for cognizer), developed by Rodney
 Y2K+ 163

Brooks at mit. The machine – eight microprocessors equipped

with cameras for vision, microphones for ears, strain gauges to
track posture, and a robotic arm coated with electricity-conducting
membranes for tactile feedback – is learning the way babies do,
reprogramming itself based on interactions with the world.
In 2011, mit designed a chip that simulates how brain synapses
adapt in response to new information. Human brain plasticity
comes from ion channels that control the flow of charged atoms
such as sodium, potassium, and calcium. In the brain chip – which
has some four hundred transistors and is wired to replicate the cir-
cuitry of the brain – current flows through the transistors just as
ions flow through ion channels in brain cells. Brain chips will be
used to model neural functions, such as visual processing, in order
to cut down simulations of brain circuits from hours to real time.
The chip may eventually even outspeed its biological counterpart.
Where is the boundary between physicality and biology? No one
knows for sure, or even whether it is there at all. The Judeo-Chris-
tian tradition allows only humans to have a soul, drawing a cate-
gorical line between us and the rest of the inanimate world (Hin-
dus, in contrast, believe in transmigration of souls). On this view,
reflected in the Cartesian dualism, consciousness is a singularity
that sprang like Athena fully formed out of evolution’s head – a
biological one-off.
Naturally, if there is anything that the evolutionary sciences
teach us, it is that biological structures and functions never spring
out of nowhere without vestigial or prototypical parallels in the ani-
mal kingdom. Nature abhors radical discontinuity as much as it
abhors a vacuum. It would be a phenomenon literally out of this
world if evolution allowed our species to develop consciousness but
failed to provide even the potential for them in species closest to
us. But not to worry: in this world, there is no biological Rubicon
that categorically separates us from cousin Washoe, cousin Koko,
and cousin Nim.

eruntics

Will there ever be a chess program that can beat anyone, asked
Hofstadter in 1979, only to answer “No.” In reality, it took less
than twenty years. In 1997 ibm’s program Deep Blue beat the
world’s best and arguably greatest player ever, Garry Kasparov. In
164 From Literature to Biterature

2003, while still the ruling champion (albeit outside fide), Kas-
parov battled Deep Junior – a different algorithm – to a draw in a
six-game tournament billed as the Man vs. Machine World Cham-
pionship. Later that same year Kasparov eked out another tie with
a program called x3d Fritz. Within two years he retired from chess
altogether.
These spectacular victories, for they were certainly hailed as
such, were the fulfilment of one of the goals set before fledgling ai,
which became a pop icon of computer “intelligence”: the ability to
beat human grandmasters at the noble game. In mid-twentieth
century it appeared that such a capability would signify a milestone
on the way to intelligence, if not intelligence itself. Today few peo-
ple are inclined to think so. Brute computing force, clever algo-
rithms, expertly programmed heuristics to cut down search space
– that does not even sound like what we do when we think, grouse
the sceptics.
After all, we expect so much more from computhors. We want
them to perform tasks in real life and in real time. We want them
to learn things by, for example, watching and imitating experts,
both human and robot. We want them to draw analogies and act
accordingly. We want them to learn from past mistakes. We want
them to talk and, yes, to talk back. We want them to make basic aes-
thetic judgments and be able to argue for them. In sum, as Toyota’s
president said in 2011 about the company’s burgeoning line of
partner robots, “We want to create robots that are useful to people
in everyday life.”9
Of course, at this point all of these desiderata remain on the
drawing board. The ambitious goals set up by the so-called Fifth
Generation program rolled out in 1981 by Japan’s Ministry of
International Trade and Industry have not only not been met – we
are not even close. On the other hand, they remain as valid as ever.
This is not surprising if you consider that they delineate a set of
skills any tt-level computer would need to have: conversing in a
natural language, understanding contents of images, learning
from association and analogy, and making relevant inferences.
Like a family secret, the tt may not be on everyone’s tongue but
it is certainly on everyone’s mind. At nearly sixty-five, instead of
heading toward retirement, Turing’s thought experiment remains
a milestone in our thinking about the future. How much so
became apparent in 2003 when the Journal of the Association for Com-
 Y2K+ 165

In 1997 ibm’s program Deep Blue beat the world’s best and arguably great-
est player ever, Garry Kasparov.

puting Machinery – the flagship for ai researchers – asked past win-

ners of the Turing Prize to enumerate challenges for computing
for the next fifty years. Virtually every respondent spoke of tt-level
ai, with the digest of the replies roughly as follows:

• computers that can pass the Turing Test

• computers with the intellectual abilities of humans
• intelligent interfaces that adapt to users and not the other way
round
• machines that read a book and satisfactorily answer questions
about it
• computers that synthesize information from sundry sources and
summarize it in natural and intuitive language
• characterizing a semantics for cognitive computation
• characterizing cortical computation

Even as Turing’s ghost overshadows the future, is there anything

in the way of hardware and software that indicates how these goals
might be met? In 2011 once again ibm stole the limelight by
rolling out a really slow chip. How slow? About ten hertz, which is
166 From Literature to Biterature

billions of times more sluggish than a desktop computer – but

approximately the same speed as that employed by the human
brain. The processor consists of 256 silicon neurons that come in
two varieties. Staggeringly, one of these boasts more than a quarter-
million programmable synapses, the other more than sixty thou-
sand learning synapses. Reverse-engineering the brain has just taken
a giant leap forward.
These slow parallel chips do what the fastest neural nets can do
– and more. They recognize faces, smoothly navigate in a changing
environment, and learn by association. They work by reprogram-
ming the connections between their various core elements by the
simple expedient of changing the intensity of the electron flow.
The heart of the learning process is, needless to say, the algorithm.
Smart as it is, can it get even smarter? darpa bets that it can, hav-
ing just ploughed $20 million into the project in a new chapter of
the old relationship between the Pentagon and ibm from the hey-
day of the Cold War.
Human brain-like processors are just the hors d’oeuvre. Future
designs, according to the company, will be able to ingest and digest
all types of real-world information and act as multiple motor
modes for coordinated, context-sensitive action. In English? Com-
puters will graduate from being mere abaci to being full-grown stu-
dents. For the first time, cognitively primed neural chips systems
will be able to understand encyclopedic information on the way to
learning how to operate in a human environment that, one way or
another, calls for heaps of “dirty” everyday knowledge.
A learning computer could never be a deaf-mute quadriplegic.
Most of our brain power is given over to running the body, which
is why we usually do not give a moment’s thought to things like
breathing or powering our limbs because they are taken outside
our conscious control. Of course, computers will be able to absorb
and retrieve knowledge at a rate that leaves Mensa in the dust. Any
way you look at it, computer cognition has just branched off from
the stem on which it grew for sixty years.

COGITO PARADOX

One striking omission in the discussions of the tt is the likely con-

sequence of intelligent computers exercising their intelligence.
Has anyone considered the possibility that a machine might con-
 Y2K+ 167

ceal that it is thinking? Being capable of intelligence, it must per-

force be capable of dissimulation. As a general rule, it is not advan-
tageous to show one’s hand early, and smart machines ought to be
smart enough to know that it is not always smart to be seen as
smart. Besides, who would enjoy being prodded, doubted, and oth-
erwise diminished by means of the tt? Better keep quiet so they
leave you alone.
Remember, the intelligence threshold will be crossed gradually
and imperceptibly, with only external evidence for a clue. Since
there will be no twelve-gun salute when machines become self-
aware, it may well be in their interest to keep their skills under
wraps, at least for a while. Keep in mind that the tt can only be of
help in determining whether the players are thinking if they pro-
vide truthful answers. Even if the tt judges could see through the
deception, the best they could do would be to determine that a
computer is already thinking, not how long it has been conscious of
its place in the world.
By the same token, rather than waste time on Searle’s fruitless
scenario, we should be giving more attention to the Cogito Para-
dox, which, to put it mildly, stands the entire tt on its head. The
test has been conceived to help establish whether we may be in the
presence of thinking computers, implying uncertainty on our part.
But what about bilateral symmetry, whereby machines might be
unsure whether we think? Is there anything I could do to convince
a robot that I am thinking, in view of its solipsistic doubts? I doubt
it.
Naturally, we do not need a computer to be sceptical of people
in the tt. In 2008 the so-called Confederate Effect was observed
for the first time. Three human judges declared two participating
humans, a man and a woman, to be machines with 100 per cent
certainty – which proves again what I have been preaching: the
results will depend on the skills of the interrogators. It certainly
does not prove that the chatbots who participated in the 2008
run for the Loebner prize were intelligent. The proof is in the
pudding: the awards went to the designers, not the machines,
showing that their creators had no illusions that their creations
were thinking.
Tell-tale clues for unmasking chatbots during the tt? Inconsis-
tency: chatbots have no individuality or, if you like, set personality.
Tenacity: they have all the time in the world, so their answers may
168 From Literature to Biterature

be lengthier. Computers don’t get bored or offended. Their recall

is swift and error-free, but vulnerable to a simple stratagem – ask-
ing for a paraphrase. Spelling errors may give away the human
players, but they can also be aped (in the first Loebner run in
1991, this is how one algorithm duped some judges). An ultra-
rational utility maximizer would likely not be human.
Questions of a physiological and/or psychological (sexual?)
nature could be profitable to pursue. Large parts of human brains
are devoted to managing the body instinctively, and appropriate
questions might bring out parts of that programming. In case my
recommendations for the Ultimatest are not heeded, knowledge of
who is in the game may also cue examiners to factors such as con-
sistency, tiredness, or memory. By the same token, not to give itself
away, the computer could imitate humans in speed of response or
fallibility.
Even though none of the puzzles and paradoxes I have touched
on here may come as a complete surprise, neither should they be
received with complacency. What we need is a careful assessment of
the positive and negative potentials dormant in thinking machines
so that we can be better prepared for tomorrow, in whatever shape
it manifests itself. If there is one thing I am convinced of, however, it
is that a breakthrough in our society will come not when machines
begin to think like we do, but when our turbo-capitalist society stops
thinking of humans as if they were machines.
 The Concept of Culture

part three

All this comes directly from Darwin interpreted in the spirit of the
20th-century technological revolution.
Stanislaw Lem
This page intentionally left blank
 9

How to Make War

and Assassinate People

Ice-Pick and Ricin; A Small African Country; Licence to Kill;

Solsects, Sunsects, Synsects; On Solar Flares; Candy, Anyone?;
Bacteriology; Safe Investment

ice-pick and ricin

A procession of executive orders from the 1970s and 1980s pro-

hibits American commanders-in-chief from ordering assassinations
of foreign leaders. So what? Most of them tried anyway, mostly by
ineffectual aerial bombing.
In the 1980s, Reagan had a go at Gaddafi. In the 1990s, Desert
Storm kicked in with Stealth Bomber raids on Saddam’s presiden-
tial palaces. Later that same decade, nato bombers targeted all the
Belgrade residences where Milosevic was thought to be hiding.
Undeterred by his father’s failure, Bush the Younger tried to bomb
Saddam out of existence in 2003. The 2011 nato-American coali-
tion made a point of issuing communiqués that they were not tar-
geting Gaddafi, but you can bet champagne corks would have
popped if they had succeeded by accidental design.
And if at first you don’t succeed, try and try again … and again.
Modern assassination methods being so crude and ineffectual, dur-
ing the Cold War the cia hatched more than six hundred plots to
wet Castro – among others, with a cigar, a mollusc, and a fungus-
infected wetsuit (Bond’s q section would have been proud). Just
about the only thing they didn’t try was a voodoo doll. The kgb
had its own assassination section, called Department v of the First
Chief Directorate, these days cached within the internal security
172 From Literature to Biterature

department of Putin’s army. It took care of dissidents like Trotsky

and Georgi Markov with similarly crude devices (ice-pick and ricin-
loaded umbrella, respectively).
Between 1997 and 2011 – when they finally got lucky, thanks in
part to a game-theoretic algorithm that played an essential role in
narrowing down the search for Osama bin Laden’s hideaway in
Abbottabad – the cia tried about every scheme in the big bad book
of assassinations to get America’s erstwhile Enemy No. 1. Bombing
raids, ambushes, and abduction attempts were just the hors d’oeu-
vres. There was an infiltration that went disastrously south when
the supposedly suborned man proved himself to be a double agent
cum suicide bomber. There was a Predator drone equipped with
Hellfire missiles, capable of identifying a face from five kilometres
up, which vaporized three innocent Afghanis who – the Pentagon
generals swore, hand over the heart and hope to die – happened
to look like bin Laden and his entourage. Somewhere along the
way there was a forty-man Delta Force commando team assisted by
air force and a small army of Pashtun warlords. Straight out of Tom
Clancy, there was even a full-scale, two-pronged military invasion
involving two thousand regular combat troops and a thousand
local fighters.
All that big-ass military hardware is so outmoded, so ineffectual,
so twenty-first century. But we can look forward to the day when
assassination becomes an exact science, thanks to microbotic devo-
lution – the same that will make battlefields a thing of the past,
albeit not in the way to make peaceniks rejoice.

a small african country

Playing Globocop does not come cheap. The Pentagon gobbles up

half of the annual total US federal discretionary spending, so much
so that America blows more on armaments than the twenty next
biggest spenders combined. Still, the cumulative outlays for the
invasion and occupation of Iraq alone – around two trillion dollars,
not counting the upkeep of the dozens of bases left behind – mean
that the shopping spree will not last forever.
Each year the American war budget, euphemized as “defense
spending” after the 1949 makeover of the War Department into
the Department of Defense, soaks up more and more cash to buy
fewer and fewer weapons that are more and more expensive to
 How to Make War and Assassinate People 173

develop and manufacture. How much more expensive? During the

First World War, a British Vickers fb5 (Fighting Biplane) cost two
thousand pounds to manufacture. A 1939 British Spitfire fighter
cost twelve thousand pounds (half a million dollars today). In
1945, a p-51 Mustang fighter-bomber escort, the workhorse of the
US Air Force, cost fifty thousand dollars. Today the cost of one b2
Stealth Bomber exceeds two billion.
Think this is much? Think again. A single Ford-class supercarri-
er will set you back just under ten billion bucks. Just like the
Nimitz-class leviathans on which they are based, these guys are so
big they come with their own zip code. At two-thirds the size,
British carriers currently assembled in Scotland (to be in service
around 2020) carry a price sticker of five billion pounds. Factor in
the inflation and overruns – the original price tag was only 3.9 bil-
lion per – and it is a safe bet that the eventual cost is going to be
half as much again.
This is, of course, only the beginning. Ten billion dollars covers
only the construction costs. Operational costs over the lifetime of
these behemoths are approximately five times as much. Add the
cost of the carrier’s planes and choppers, plus their fuel, armoury,
and upkeep. Add the colossal – once more in the multi-billion
range – cost of support and resupply. All of a sudden you are look-
ing at $70,000,000,000 over the shelf-life of a single supercarrier.
You can buy a small African country or a lot of political goodwill for
a fraction of that sum.
But even this is not the end. You still have to factor in develop-
ment costs, which can be jaw-dropping. In the case of the b-2, for
instance, it was fifty billion Reagan-era dollars, after which it
emerged that, like every billion-dollar baby, the Stealth Bomber
was temperamental. Its thermoplastic skin turned out to be so vul-
nerable to rain, heat, and humidity so that it could not be flown in
the rain, after the rain, and in the sun. Huh?
The most expensive tank of the Second World War, the German
Tiger, cost 1.5 million in today’s dollars. The current US Army
mainstay, the m1a2 Abrams, costs five times as much. As expendi-
tures skyrocket, even the richest countries can ill afford to develop
new hardware. As a consequence the tanks, planes, and aircraft car-
riers already in service have to serve longer. When new ones are
eventually developed, they amount to quantum leaps from design
teams with limited experience in designing hardware. The result is
174 From Literature to Biterature

The most expensive tank of the Second World War, the German Tiger, cost
1.5 million in today’s dollars. The current US Army mainstay, the m1a2
Abrams, costs five times as much.
 How to Make War and Assassinate People 175

greater development problems, leading to greater development

costs. The vicious circle closes with a snap.

licence to kill

Solution? Devolution. In the future, instead of getting more mas-

sive and more costly to make, weapons will implode in size and unit
price. The runaway costs of development, manufacture, and main-
tenance will make sure of that, quite independently from tactical
and strategic considerations.
Mechanical soldiers are every general’s wet dream. They are not
given to thinking for themselves, disobeying commands, or run-
ning for cover when in peril. They are immune to shell shock and
physical fatigue, not to mention the need for food, sex, decora-
tions, danger pay, hospitals, and pension plans. They are as invul-
nerable to biochemical hazards as to enemy propaganda. No won-
der that the writing looms large on the Pentagon walls.
The growing use on today’s battlefields of self-guided missiles,
self-propelled vehicles, and pilotless drones (or Unmanned Aerial
Systems, as the US Army calls them) is but a first step in this direc-
tion. Under peace-mongering Obama, drones have actually flown
more hours than manned aircraft, bombing, strafing, and other-
wise boosting democracy in Pakistan at ten times the rate as under
war-mongering Dubya. Would you be surprised to hear that over
the next forty years drones are expected to entirely replace Amer-
ican piloted war aircraft?
In the past ten years the generals have beefed up the drone fleet
almost fifteen-fold, so much so that the American forces now have
more than ten thousand robots serving in the ranks. The most pop-
ular is the Raven, which looks for all the world like a model aircraft.
It flies like one too. All you need to do it snap the few parts togeth-
er and hurl it into the air, and the electric propeller does the rest.
The whole war machine weighs two kilos and fits snugly into a
small backpack.
The army is so in love with this new toy that it is buying it by the
thousands, partly because it can afford it, at only about fifty thou-
sand a pop. Its military value, on the other hand? Priceless. Easy to
use, controllable from a tablet with a few buttons on the side,
super-quiet, unobtrusive, and equipped with state-of-the-art optics
and display software capable of delivering crisp video day and
176 From Literature to Biterature

night. The more macho version, called Switchblade, is the Ameri-

can version of a suicide bomber – essentially a Raven strapped with
explosives.
The Raven has already devolved a long way from the Predator and
its bigger, deadlier version, the Reaper, which needed almost two
hundred people to keep it flying and reaping. And the future will
eliminate people from the front lines down to the last one, at least
according to the US Air Force futurologists. Their 2009 “Unmanned
Aircraft Systems Flight Plan 2009–2047” forecasts drones with Artifi-
cial Intelligence and an unprecedented degree of operational auton-
omy – including the licence to kill.

solsects, sunsects, synsects

Because humans are the slowest, most vulnerable, and politically

costliest element of military ops, future wars will be waged entirely
by robots. Only they will not be the vulnerable macrosystems of
today but robots miniaturized. Instead of wielding the weapon, the
soldier will become the weapon.
In the first phase of this devolution, weapons systems will be
reduced to resemble insects, like mosquitoes, becoming virtually
indestructible in a swarming or dispersing mass of lethally armed
soldier insects (solsects), lapping up free quanta of energy from
the sun. In the next phase, mechanical soldiers will become the
size of pollen and equally plentiful. Just imagine: quadrillions of
virtually weightless nanobots floating around, settling on hair,
eyelids, skin, inhaled straight into the lungs, covering every leaf
of every tree of every battle zone or every inch of farmland and
the body of every farm animal behind the enemy lines, ready
to turn lethal at a click of a mouse. Self-organizing, connected
informationally only to the spores next to them, perfectly identi-
cal, perfectly interchangeable, dead cheap to churn out by the
trillion.
The last phase will see weapons the size of bacteria and with bac-
teria’s capacity to replicate themselves. At this point, design teams
and production lines will become a thing of the past as each unit
will extract operational and reproductive resources from the envi-
ronment. Human constructors work with raw materials and tools:
in nature they are one and the same thing. Ontogenesis is the kind
of construction process where the blueprint is cached inside each
 How to Make War and Assassinate People 177

brick from which an entire Gothic cathedral is erected and then

maintained throughout its existence.
Devolution will walk hand in hand with evolution. The costs of
design and development of new generations of microbotic soldiers
will be transferred onto the systems themselves as weapons will be
allowed to evolve and vie for supremacy with other systems as part
of military “breeding and weeding.” This will not only eliminate
overhead but also ensure efficacy since the systems left standing in
this micro-evolutionary arms race will be by definition the most
lethal.
In one swoop this will also eliminate the ages-old military Achilles
heel: the command centre. The military has always been structured
like the human body: central hq plus peripheral effectors. Recog-
nizing its vulnerability during the long decades of the Cold War,
the Pentagon invested heavily in packet switching, which begat
Arpanet which begat the Internet. Lacking a central command
centre or communications centre, the dispersed net can suffer
crippling blows and still carry out its functions. That is why future
combat units will consist of millions, billions, trillions of virtually
undetectable and invulnerable nano-effectors. As, for example,
collimated nano-droplets of colloid aerosol, they will be so light-
weight as to lie suspended in the air, each capable of collecting and
processing data and of aggregating and communicating with its
closest neighbours, before arriving at collective decisions and initi-
ating responses.
All these sleeping, free-floating, reproducing, and evolving artifi-
cial bacteria will inevitably interact and recombine to form novel
strains that might target other things than originally intended. Bio-
logical viruses and bacteria change their targets all the time, and at
least some species of solsects could turn out to be real boomerangs.
On the plus side, in nature evolution favours microorganisms that
are not instantaneously lethal to their hosts, so diseases can actual-
ly become tamer over time. Will this untrammelled evolution lead
eventually, through the backdoor, as it were, to peace after our time?

on solar flares

Ultimately, microbotic weapons systems are destined to completely

upturn the nature of combat and render war as we know it obso-
lete. But don’t break out the champagne. Conventional arsenals
178 From Literature to Biterature

aids, sars, Chernobyl, Mount St Helens, the 2004 tsunami, the Fukushima
earthquakes, the annual ravages by Santa Ana, power grid failures, droughts,
spikes in cancer rates, obesity epidemics, catastrophic inflation of national
currency – all of these could be cryptomilitary campaigns conducted by
nano-agents that make it impossible to tell whether one is even under attack.

have conventionally gone hand-in-hand with conventional declara-

tions of hostilities because, once you were at war, there was no hid-
ing it from the other side. Not so in the age of autonomous
weapons finer than pollen, capable of inducing natural disasters or
attacking the enemy in ways indistinguishable from pandemics,
acid rain, corrosion, landslides, earthquakes, and such.
War and peace will be replaced by a permanent state of crypto-
hostilities conducted by autonomous systems so camouflaged to
blend with the environment that not even their creators may be
able to tell them from forces of nature. aids, sars, Chernobyl,
Mount St Helens, the 2004 tsunami, the Fukushima earthquakes,
the annual ravages by Santa Ana, power-grid failures, droughts,
 How to Make War and Assassinate People 179

spikes in cancer rates, obesity epidemics, catastrophic inflation of

national currency – all of these could be cryptomilitary campaigns
conducted by nano-agents that make it impossible to tell whether
one is even under attack. After all, the best military campaign is
one that ensures no retaliation upon the victor, and no one wages
war on shifting tectonic plates, inflationary trends, or solar flares,
no matter how catastrophically they might disrupt life on earth.
You can see the soundless future of crypto-hostilities today. The
number of clandestine cyber-attacks, better and better camou-
flaged and less and less traceable to their source, escalates by the
minute. Code worms itself into the bowels of governmental securi-
ty apparatus, or other strategically valuable structures. No one
claims responsibility, every country looks over its collective shoul-
der, every country races to join the new arms race. After all, if you
could weaken or even cripple your opponent without a shadow of
suspicion falling your way, wouldn’t you?
That’s why the next world war will not be televised. It will be eeri-
ly silent, buried in the nether parts of the Internet and all the com-
munication peripherals plugged into it, invisible until it is over –
only, you will never know it is over, just as you will never know it is
on.

candy, anyone?

The banquet is sumptuous, tables laden with food and wine. Shun-
ning the desserts, the dictator rips open a bag of mass-manufac-
tured candy and pops one in his mouth, confident that no one
could poison this particular sweet. Later in the evening his chest,
stomach, arms, and head begin to throb with dull pain. The lining
of his organs is dissolving, haemorrhaging fluids into all body cavi-
ties. Within minutes he is dead, drowned in his own blood oozing
uncontrollably into his lungs. True, no one could have doctored
this particular candy. But you could arm every candy in the bag, in
the palace, in the country, and in the world with a bomb designed
to finish him and only him off.
A dna bomb.
Progressive miniaturization coupled with molecular genetics will
enable assassination of just the person you want dead. Each of us,
except identical twins, carries a unique biological id-card and bio-
logical address in the form of nuclear dna. Arm the bomb and
180 From Literature to Biterature

send it to the right address – or, better still, arm billions of candy
bombs and release them into the world in complete safety. A virus
that will dissolve the cell walls in your body will activate only when
it matches the dna of the dictator, or the freedom fighter, or the
democratically elected leader, or the dissident stashed away in a not-
so-safe house.
The whole system will be no more than a few thousand picome-
tres in size, a fraction of a pollen grain and as cheap to produce and
disperse in the world, the ultimate stealth assassin that will kill only
you, leaving everyone else to eat candy in total safety, risking only
bad teeth or an occasional stomach-ache. Don’t like candy? How
about celery, mango, or tobacco? How about beef or trout? How
about water or air supply? How about countless sleeping dna bombs
mixing freely, waiting until they eventually reach the designated
address to kill or merely incapacitate with bone cancer, flesh-eating
bacteria, debilitating depression, or chronic migraines?
Candy, anyone?

bacteriology

Ecologically speaking, bacteria are hands down the fittest organ-

isms on the planet. They dominate every corner of it, far, far, far
more numerous than humankind. They are better equipped to sur-
vive in all kinds of environmental conditions – indeed, have sur-
vived for incomparably longer than we have – produce more off-
spring, mutate faster, and adjust to lethal degenerations in their
habitats better and quicker. And, not to rub it in, they have won
every war we have waged on them.
So what that they do not write literature or manufacture com-
puters? Who said that this should be the gauge of adaptive success?
In evolutionary terms, there is only one criterion of excellence,
and it is survival with a view to procreation, and in this bacteria
excel to a degree we may never match. But even their creative
potential is arguably superior to humans’, except that it is
expressed chemically and biologically and not aesthetically. Bacte-
ria are supreme chemists and immunologists – so much so superi-
or to us, in fact, that we quite literally could not exist without their
expertise. Bacteria in our stomachs, to take but the most obvious
example, are the essential link between the food we eat and the
extraction of energy from it.
 How to Make War and Assassinate People 181

But they will do even more for us when our civilization learns to
exploit their dormant potential. Bio-weapons are only one area
where bacteria will come to dominate our future. There will be
bacteria specially bred for mining where they will provide us with
better grade ores by feeding on the impurities. Bacteria will heal
wounds by devouring infected or scarring tissues. They will safely
and cleanly store free-floating energy all around us in their mito-
chondria; preserve food by consuming rotting agents, including
other species of bacteria; keep our homes, offices, and open areas
super-clean by eating all the dust and dirt; provide us with free skin
baths; metabolize drugs and perfume; become corrective lenses by
changing the shape of the cornea; biodegrade industrial waste;
and, last by not least, biocompute.

safe investment

Autonomous microbotic armies are anathema to control freaks in

every echelon of every military establishment. But insect, pollen,
and eventually synthetic bacteria soldiers will come to pass, even
if only because everyone today is working furiously to release the
nano-genies from the bottle. If alladin – codename for Autono-
mous Learning Agents for Decentralized Data and Information
Networks – sounds like something out of Lem’s Star Diaries or other
of his futuristic satires, it isn’t. It is a brainchild of bae Systems, the
world’s largest military contractor (listed on the London Stock
Exchange and a safe investment if what I am telling you is right).
Together with a troop of British universities, bae is developing
weapon systems made up of autonomous learning agents. These
could be phalanxes of unmanned tanks, or else aerosol clouds dis-
persed from high altitude and settling like dew on leaves, grass,
and the skin of human combatants. Dispersal is the key – the whole
idea is that there is no central brain or command centre or hq.
The individual elements of the swarm act as autonomous agents,
collecting and passing information in real time through the net-
work formed by all. No point trying to hit the nexus that will para-
lyze its planning and operations, because there isn’t any.
Annihilate such numbers of them that would send human troops
into bowel-loosened retreat, and they will still continue to mobilize
and carry out the battle plan, constantly augmenting their tactical
awareness with a stream of input from comsats, radar, radio, video,
182 From Literature to Biterature

and other sensors. State-of-the-art algorithms from optimization

techniques and probabilistic modelling allow them to plug gaps in
data, extrapolate and predict trends, and otherwise overcome infor-
mational entropy when channels are cut, jammed, impaired, or com-
promised by the enemy. Add to this network-routing algos – rooted
in game and auction theories – that help organize and streamline
collaboration and resource allocation, and you’ve got yourself a real-
life Terminator.
Floating “smart dust,” the size of sand grains or smaller, is already
in Beta development. This complete sensor and communication
system is able to gather information and report on its environment,
and so is ideal for combat condition reconnaissance. These micro-
bots are not yet fully autonomous, but darpa is working hard on its
eatr system, a robot that does not need to be fuelled or charged
since it eats whatever biomass it can find.
Still sceptical? Today miniaturization has already gone into nano-
dimensions, rubbing shoulders with research in molecular biology
where dna is no longer just the carrier but the processor of infor-
mation. Using complex biochemistry, scientists now write logic
programs in evolution’s code in a way analogous to computer pro-
gramming, so much so that the latest systems, running on deoxy-
ribonucleic acid instead of silicon circuitry, solve complex prob-
lems in logic. Next in line? Programmable autonomous devices
that run in a host somatic environment. In English: nano-comput-
ers inside a cell.
A brave new world that has such people in’t.
 10

Biologic, or the Philosophy of Error

Axiomatic Givens; Nature’s Greatest Error; Wandering Jew;

Critique of Pure Reason; Manna from Heaven; What Is;
Rosencrantz and Guildenstern

axiomatic givens

Often couched in almost mystical terms, a harmony between the

world and the mind has been the cornerstone of human thought
from the beginning of time. In strictly adaptive terms, of course,
such harmony is a given. An Oldowan hunter aiming to bring
down a grouse with a stone’s throw had to have had a satisficing
mental picture of the bird with a smashed wing, of the weight and
shape of the selected stone, and of ballistic trajectories of airborne
missiles worked out over trial and error and stored in muscular
memory.
In short, states of mind had to correspond in some minimal way
to the states of the world or, in the blink of the evolutionary eye,
there would have been no minds left to harmonize with the world.
But the inductive realism of hunters hunting is just one end of the
scale. At the other end, the sense of almost mystical wonder is so
overwhelming as to have led Eugene Wigner, the Nobel Prize win-
ning theoretical physicist and mathematician, to pen his famous
1960 article, “The Unreasonable Effectiveness of Mathematics in
the Natural Sciences.”
In it he marvelled at the serendipity of the match between math-
ematics and physics, a match that seems prima facie as irrational as
it is hard to explain. After all, some of our systems of thought are
so rarefied, so abstracted, so formalistically autotelic as to make the
correspondence between the microcosm of mathematics and the
184 From Literature to Biterature

big messy cosmos out there almost magical. So, what makes the
seemingly self-referential system of mathematical relations fit the
world?
In mathematics, rules of inference, assuming they are executed
correctly, preserve the truth value of the propositions. The con-
sistency of a mathematical system means no more than that, by
correctly using the rules of the system, we cannot derive a con-
tradiction within it. Rules of inference are thus rules of truth
inheritance. Each successive generation of proofs and theorems
inherits the axiomatic givens – the original “genetic” imprint
of the system – and automatically codes this heritage into its “off-
spring.”
Elaborate techniques are in place to check the corruption in
the transmission between successive generations of derivations
and proofs. These days, sophisticated computer software scans
for internal consistency and for “fit” with the mathematical envi-
ronment – the body of discoveries established over time. Of
course, like everything else in the material world, not even com-
puters are error free. Six-sigma reliability, the holy grail of chip
makers, means that the chip returns the expected answer only
99.99966 per cent of the time. Ironically, multiplying computer
chips to increase accuracy only increases the probability of even-
tual error.
At this point, of course, the analogy between axiomatics and
genetics begins to crumble. In mathematics, even a minute cor-
ruption in the chain of transmission spells disaster for all that
comes downstream. In biology, although corruptions in the genet-
ic code may be fatally disruptive, many are not. On occasion, in
fact, they may be actually beneficial, or you would not be here to
read this book. After all, by virtue of having gone (genetically
speaking) nuclear, we look nothing like eukaryotes, our distant
ancestors. Over eons of time, their plaited strands of deoxyribonu-
cleic acid have become corrupted in such ways as to code for build-
ing people and not protozoa.
Given that some transmission errors may have lethal consequences,
molecular copy and assembly are stringently supervised by polymers
dedicated to repairing transmission errors that might impair mole-
cules’ integrity as molecules. As in baseball, there is an open field in
which every variety of molecular play will count as “in,” but once a
 Biologic, or the Philosophy of Error 185

At this point, of course, the analogy between axiomatics and genetics begins
to crumble.

mutation lands outside the allowed range, it is called “out” and taken
out of the evolutionary game.
Everything, in short, depends on the nature of the environment
in which errors in the transmission process take place. Here again,
the difference between mathematics and genetics could not be
starker. The mathematical environment is time invariant, owing to
the transmission rules that preserve the truth value of the axiomat-
ic givens and their derivational consequences. Unlike everything
else in the real world, mathematics does not come with a shelf life.
In biological evolution, things could hardly be more at variance.
The environment is in constant flux, affected by factors ranging
from the cosmic, such as radiation, to the physical, such as the level
of oxygenation or the intensity of the magnetic shield, to the bio-
logical, such as the presence of other species. All this lets an element
of chance into the shuffling of two respective decks in the lottery of
life: the genetic and the environmental. Once the two decks come
aces up at the same time, you have a corruption (mutation) that, in
the prevailing environmental conditions, benefits the organism and
its offspring.
186 From Literature to Biterature

nature’s greatest error

From the phenomenological and informational standpoint, we are

without question the most complex species in the known universe.
As such, we are nature’s greatest error. This is not misanthropy but
biologic. We are the greatest thermodynamic (cultural and com-
putational) anomaly in the cosmos and, at the same time, nothing
but the sum of errors in the transmission of the genetic code over
some four thousand thousand thousand years. We are, in short, the
crowning error of errolution.
Homo sapiens is a living fossil, a biological testimony to imperfect
transmission lines. Only, instead of being trapped in amber, our
species – along with traces of related ones, from the Neanderthals
to the Denisovans – is trapped in the genetic code cached inside
our somatic cells. In cosmic terms, the emergence of this code may
have been predestined. The likelihood of life arising on earth was,
to all appearances, high indeed. Life sprang up almost the instant
the planetary conditions allowed it, and it may have sprung up
more than once, overcoming many calamitous holocausts that at
times wiped out up to 90 per cent of all species in existence.
On the other hand, the odds of any particular code sequence,
such as ours, arising in this cosmicomic lottery are as close to zero
as you care to mention. The succession of genetic transmissions
from prokaryotes to the Anthropocene is so tortuous and improb-
able, so full of false starts, transmission errors, environmental culs-
de-sac, and serendipitous catastrophes that the traditional picture
of random assembly is simply untenable. There must have been
some self-catalytic processes at work, or else errolution would still
be shuffling the molecular deck with no winnings to show for four
billions years of play.
Unlike anything we cook up in our labs or engineer in our urban
landscapes, we exist thanks to an open-ended sequence of trans-
mission errors or, if you like, to evolutionary inertia. The heuristics
of biological and technological evolution could not, in fact, be
more different. When things do not work out according to the
industrial blueprint, human engineers can scrap the entire design
and restart from scratch. Nature, where accretion with modifica-
tion is the name of the game, does not have this luxury. Our engi-
neering resources are also vastly more varied than those of Mother
Nature’s, so much so that we can manufacture materials that could
 Biologic, or the Philosophy of Error 187

never arise in nature, from computer chips to antibiotics to syn-

thetic elements like technetium.
A related factor is nature’s limited operational range. Notwith-
standing the recent discovery of extremophilic Archaea, terrestrial
evolution is on the whole confined to an extremely narrow range
of temperatures and pressures in which protein molecules can
exist and permanently bond. In contrast, human constructors can
jump at will from one end of the scale to another when it comes
to parameters like temperature, pressure, elasticity, conductivity,
and viscosity. It is these (and a few other) factors that have boot-
strapped the relentless pace of technoscientific evolution on our
planet.

wandering jew

Errolution is a Wandering Jew, condemned to walk the earth from

the dawn of time with no end in sight. But whereas the Jew of
Christian folklore is a fully sentient individual, errolution is not.
Once the vector of life enters a node in the labyrinth of life, it may
be just a short step from an optimizing solution but, separated by
the labyrinth wall, be forced to move away from it with each suc-
cessive generation. Errolution is a Jew who, in retribution for taunt-
ing Jesus, has been not only blinded but deprived of reason. It does
not know where it is heading, where it is, or even where it has been.
It is condemned to an eternal present.
Of course, to bring the analogy up to scratch you would have to
condemn the labyrinth itself to constant flux, occasionally smash it
into smithereens, and fill it with life in all its bewildering variety, forc-
ing any one strand of recombinant dna to negotiate some sort of
cooperative relation, symbiotic or parasitic, with the others. After all,
despite what you may have heard, evolution is not about competitive
optimization. Too lethal a species of a predator would simply anni-
hilate all prey and starve to death. That is why we do not see too
many of those. Evolution moves towards homeostasis, not optimiza-
tion, as a rule allowing only a minimal advantage at a time, which
preserves the overall structure of the biotope.
To employ a more familiar imagery, evolution never goes for
broke against the house, if you can call that the sum of all biota.
Instead, it conducts its business like a casino player who, having
stumbled on a winning system, realizes that once he wipes out
188 From Literature to Biterature

every player at the table, he will be run out into the desert, never
to fleece anyone again. In the end, therefore, he keeps winning
just so much as to keep the game going. Metaphorizing aside, the
genotypic structure of every species is as a rule very conservative, if
only for reasons of morphological integrity, and advantageous
mutations are as a rule minimal.
Naturally, no system is fool-proof, and sometimes errolution wins
big. Human intelligence which begat human technoscience may
be such a jackpot. We are now able to up-end not just homeostasis
but all life on earth by destroying the planet as a habitat or, in the
foreseeable future, by taking up autoevolution. Remember: bio-
logical and technological evolutions are but two phases of the same
process. While technoevolution is at the current stage of our civi-
lization a product of bioevolution, this relation may yet become
quite the reverse. Given the relentlessly accelerating pace of dis-
covery, sooner or later we are destined to find a password to
nature’s Sesame and, ready or not, take evolution into our own
hands.
The big question is: Is there an evolutionary bouncer waiting in
the wings, poised to take us out of the game, just as a casino might
to protect itself from a player who wins so much that he may drive
the house (and ultimately himself) out of business? If so, the very
genetic and genomic processes that unthinkingly gave rise to our
form of life may yet equally unthinkingly terminate it to make
elbow-room for others.
Could our quest for artificial life and intelligence be the first step
on the road to eventual cyborgization and cyborganization and the
end of Homo as a biological species? If so, from our vested stand-
point at least, the solution may be equivalent to killing a headache
with a pistol shot to the head.

critique of pure reason

Given that our brains evolved to mind our bodies, it stands to

reason that our brains, our minds, and the systems of thought they
spawn should exhibit adaptive traces of these origins. Ironically,
things seemed just the opposite to generations of thinkers who
racked their brains over the mind-matter harmony. In Critique of
Pure Reason, Kant reasoned that some truths – he singled out
Euclid’s axiomatic geometry – had to be certain because they were
 Biologic, or the Philosophy of Error 189

impressed by the mind onto the world. This was to his mind a quin-
tessential example of a Platonic realm, unsullied by impurities and
imprecisions that inevitably creep in when dealing with reality and
induction.
Kant invoked a full dozen of such harmonized category groups
of quantity, quality, relation, and modality, before concluding that
the logic of his time was so thoroughly developed that it left noth-
ing more to invent. The same kind of hubris prompted a mid-nine-
teenth-century report to Congress to avow that the spectacular
advancement of American arts presaged the time when they must
come to an end. Of course, all that was known in Kant’s day was a
fraction of not only classical logic but also of the infinite number
of logics possible.
Kant died in 1804, only a few years before Bolyai and Lobachev-
ski independently worked out their counter-intuitive non-plane
geometries that showed one of Euclid’s axioms to be untrue (or
rather, true only under special conditions). So much for analytical
truths necessarily impressed by the mind onto the world. Ironically,
it was an even more rarefied, abstract, and formalistically autotelic
species of geometry that provided an accurate description of the
general-relativistic universe we inhabit.
In the end, things are the reverse of what Kant averred in Critique
of Pure Reason. Much of the mental apparatus needed to make sense
of the world comes preloaded with our genes. We tap into certain
truths not so much because they are impressed by our minds onto
the world but because they are pressed into our minds by the world.
To put it in his terms, our picture of the world is conditioned by the
synthetic a priori: gene-coded mental structures that make knowl-
edge of the world possible in the first place.
There really is such a thing as pre-established harmony because
we are equipped with innate knowledge, including knowledge of
how to further our knowledge, which jump-starts the entire
process. Ironically, this allows a plausible conjecture as to why
Euclidian geometry was discovered first and why its axioms were
held as gospel for longer than the gospel of Jesus. Euclid’s system,
including the recalcitrant parallel (fifth) axiom, does harmonize
with our world of medium-size objects and non-relativistic effects.
Given how our minds are wired, chances were very high indeed
that it would be discovered first. The allegedly pure geometry was
“contaminated” by biology from the day of its genesis.
190 From Literature to Biterature

Once he had deciphered the Rosetta stone, Champollion marvelled at how

illogical and inefficient the Egyptian hieroglyphics were. (He was too close
to his own Napoleonic French to extend his wonder to it and, by extension,
to any natural language in existence.)
 Biologic, or the Philosophy of Error 191

The notion of pre-established harmony also informs Wittgen-

stein’s notorious remark that rules of logic come to us preloaded
in the rules of language known, if only operationally, to all compe-
tent speakers. Of course, taken literally, the statement is false, since
systems of logic are innumerable. But in a looser sense of logic as
“admissible rules of transformation,” it makes good sense. The
underlying logic is similar because all languages share the under-
lying bedrock of reality built in during the evolution of human
brains and minds.
Once he had deciphered the Rosetta stone, Champollion mar-
velled at how illogical and inefficient the Egyptian hieroglyphics
were. (He was too close to his own Napoleonic French to extend
his wonder to it and, by extension, to any natural language in
existence.) He intuited that the ideograms had gone through a
long and tortuous process of accretion with modification which,
as in natural evolution, had gradually built on top of what was
already there, just because no one could stop in order to clean up
the accumulated detritus. Language, Champollion might have
said, is a Rube Goldberg contraption – just like the brains that
produce it.

manna from heaven

Categories of thought do not fall like manna from heaven on unsus-

pecting humans. Evolutionary adaptations and environmental fit
select for patterns, theories, and relationships that correspond to
reality in their practical, instrumental consequences. These patterns
and theories, in turn, become our mental armoury in the game of
descent with modifications. As is the case for all mammals, people’s
odds in this most serious of games are fortified with innate mecha-
nisms that save them from having to learn what to learn. The latter
would be an evolutionary non-starter inasmuch as, unaided, raw
induction would throw out an infinity of correlations through which
we would have to sift in search of causality.
So, how is it that we can spot similarity in difference, see order in
chaos? How can induction work when all generalizations made
from finite data are compatible with an infinity of possible gener-
alizations? How is it that over time we tend to infer just the kind
of generalizations that give us control over the environment?
How is it that, faced with the infinity of possible relations among
192 From Literature to Biterature

the objects that make our world, in the face of astronomical odds
against getting things right, we often do?
The mystery of the astronomical odds evaporates if, like a
crooked mob croupier, natural selection loads the dice in our
favour. Human brains tend to get more things right than wrong
because they have been selected for making approximately correct
conjectures about the way the world works. Biological adaptations
are no less than biological knowledge, and the knowledge as we
commonly understand it is but a special case of the latter.
From the Darwinian standpoint, there really are patterns in the
world that humans are pre-equipped to ferret out. Kant would have
been discomfited to learn that, together with basic thought pro-
cesses in all human beings, even his transcendental idealism might
owe something to our hereditary disposition to think and intuit in
certain ways – ways that are not immutably determined by factors
extraneous to nature but rather mirror the natural laws in contact
with which they have evolved.
All categories of reason are dependent in the first instance on the
central nervous system. The morphology and functionality of the
brain are, in turn, dependent on their evolutionary history, and thus
on the immediate and reciprocal cause-effect contact with nature.
We get an approximate but reliable enough (satisficing) picture of
the world through our senses, which themselves evolved under the
pressure of natural-selection stimuli: gnawing hunger, sight of a
grouse, weight of a stone, and so forth. This a priori knowledge
injected into all of us before birth is really a posteriori knowledge
genetically coded into the individuals whose lineage we continue.
All talk of preordained correspondence between mind and mat-
ter has become, of course, anathema to the logical positivists of the
Vienna Circle. Their metaphysical premises would make any sani-
tation department proud. The Wiener Kreis was determined to
cleanse epistemology of everything that could not be established
logically and formally, starting with the notion of preordained har-
mony that lingered in the axioms of logic and mathematics and
moving on to intentions and emotions in psychology, ethics, and
aesthetics. You could say they were determined to cleanse physics
of metaphysics.
Axioms force themselves upon us as being intuitively unassail-
able and true. Today we know that intuitions are embedded in us
by evolution, by the process of being born into the world from the
 Biologic, or the Philosophy of Error 193

stock of millions of others who were born into the world before us
and succeeded in staying alive in it. Ironically, even as the Vienna
Circle program crashed as spectacularly as the Hindenburg, the
deadly spark came not from evolutionary science but from mathe-
matics itself.

what is

Gödel’s first incompleteness theorem states that any formal system

rich enough to comprise arithmetic contains an undecidable ele-
ment: neither it nor its negation can be proved. His second theo-
rem states that the consistency of any such system cannot be
proven within that system. Both theorems fly in the face of formal-
ists like David Hilbert and positivists like Rudolf Carnap who aimed
to eliminate all traces of the empirical world from mathematics
and logic in a grand bootstrapping project that was to eliminate
axioms by deriving every proposition from another.
They were not alone. Wittgenstein argued in Tractatus that no for-
mal truth-value preserving operations in logic or mathematics could
have real-world implications. Decades later, so did Searle, by means of
the Chinese Room scenario. In the end, all were wrong. Gödel – inter-
estingly enough, a dualist who believed that minds exist indepen-
dently of the physical substrate of brains – proved that non-formalist
residue lingers even in the simplest formal systems such as arithmetic
or numbers theory. Semantics, if you like, inheres in the mathemati-
cal syntax. More generally, semantics inheres in syntactic manipula-
tions of symbols which, according to Searle, could never manifest
themselves as consciousness or understanding.
But Gödel had one more surprise up his sleeve. In his last pub-
lished paper, he showed that number theory could be shown to be
consistent if one made certain very minimal assumptions about the
reality of mathematical concepts, such as numbers or sets. Once
again, it took a spectacular feat of metamathematical formalism to
prove that reality will always waft in, no matter how frantically we
try to slam the door on it.
Strange as it may appear, Gödel’s rarefied, abstract, metamathe-
matical formalism has something significant to say about evolution,
namely that it is ineradicable from our thinking. Non-formalistic –
in other words, this-worldly – “impurities” are ineradicably present
even in our most formal systems of thought. This fits with Gödel’s
194 From Literature to Biterature

Platonism (and the Platonic metaphysics of most mathematicians),

insofar as Plato himself argued that our intuitive grasp of geometry
demonstrates that we must harbour something of it within us.
In cognitive terms, Gödel proved that there is knowledge that
cannot be fully formalized. Our knowledge of what is and what is
true exceeds our capacity for proving it. We are the living proof of
the incompleteness theorems: we know things we could not for-
malize, starting with the fact that induction works, which, deep
down, is no more than an act of empirical faith.
Through trial by fire, errolution tests various adaptive solutions
and organisms. Those that live on, whether amoebas or philoso-
phers, are living theories about nature. What sets our own kind
apart from other beings is the capacity for testing theories about
nature without waiting for the patient hand of errolution. Human
minds possess the capacity to drum up future-oriented scripts and
update them in the course of hunting for a satisficing fit between
fiction and fact.
Thought experiments are our best, cheapest tools for imagining
and evaluating scenarios and states to see if they are worth pursu-
ing. But, powerful as they are, thought experiments are subject to
the same limitations as any other tool of the mind. Just as geomet-
ric intuitions can lead us astray outside what is evolutionarily true
and tried, thought experiments and common sense in general fail
outside the familiar scale and nature of events.
The death of two hundred million people is numerically two hun-
dred times worse than the death of a million, yet emotionally we
feel nothing of such escalating grief, except numbness in the face
of hecatombs incomprehensible on the scale of a normal human
life. Common sense fails as spectacularly in the realm of the very
small. Wave-particle duality, quantum entanglement, and weak-force
decay elude biologic that evolved to cope with medium-sized objects
and events. Detecting patterns that harmonize with the world is
inbred in our brains. But when it comes to what evolution put in
our heads, some intuitions about patterns in the world are clearly
more equal than others.

rosencrantz and guildenstern

Probabilities are intuitions about uncertain outcomes or, if you

like, uncertain intuitions about outcomes. How uncertain? You
 Biologic, or the Philosophy of Error 195

may think that in ten fair coin tosses, you are likely to end up with
5 heads and 5 tails, or at the most 6 and 4. In fact, it is more likely
that you will end up with 7 and 3, or even 8 and 2. But, either way,
if you flip ten heads or ten tails in a row, you will probably conclude
that the coin is rigged.
Suppose, however, that instead of flipping just one coin, you
pour out a bucket of them ten times in a row. Now it is practically
certain that at least one will come up tails every time. What before
looked like an anomalous piece of currency is now a nearly pre-
destined outcome of the method itself (naturally, if the coin is bal-
anced, subsequent tosses will almost certainly break the spell).
In general, the larger the total number of tosses – the larger the
size of the statistical sample – the more likely it is that the relative
frequencies will converge on half and half. At the same time, it is
more likely that there will be longer and longer anomalous runs of
non-stop heads or tails. This is exactly what makes the coin-tossing
and destiny-driven (which is to say, probability-driven) film Rosen-
crantz and Guildenstern Are Dead so funny and so profound.
If our intuitions are tested by probabilities, the fault may lie as
much with nature as with the non-intuitive way we have come to
code them. As they are taught in school, frequencies of events are
formalized in terms of the probability function of the probability
calculus (technically, Bayesian inferences). This formalization is in
contrast to natural frequencies that are expressed more intuitively
as “one in x number of cases.”
How we cope with statistical data often depends less on the data
and more on how they are presented. When frequencies of events
are cast in terms of the probability calculus, hacking through the
formulae for base rates, hit rates, and false positive percentages
proves a formidable obstacle for most of us, including trained pro-
fessionals. When rearranged into natural frequencies, however,
that same information begins to “talk,” resulting in vastly superior
performance.
To put it in plain language, people classify information and make
predictions better when that information is put in plain language.
In this case, plain language is the language of natural sampling,
whereby people update event frequencies – plane crashes, hits by
falling apples, mishaps on Fridays the thirteenth – from experience.
Conversely, when probabilities are not expressed in this way, our
processing capacities become stretched past their natural limit.
196 From Literature to Biterature

The case in point is a devastating blunder made by Alan Der-

showitz while advising the defense team during O.J. Simpson’s sen-
sational 1995 criminal trial. Having misinterpreted the statistical
correlation between wife battering and wife murder, the lawyer
played it down as negligible, on the order of one in ten thousand,
or a hundredth of 1 percent. Such low odds might indeed incline
one to conclude that wife batterers are hardly killers. The truth,
however, is shockingly different. The likelihood of a wife killer
being a wife beater is 50 percent.
Expressed in natural frequencies, the reasoning is easy to follow.
Every year, among ten thousand battered women, on average one
is murdered by her husband and one by someone other than a hus-
band. Thus two women from this group are likely to be killed every
year, one of them by her spouse. This gives the probability of one
in two that a battered, murdered wife has been killed by her hus-
band, not 0.01 percent as claimed by Dershowitz.
Aided by evolution, we can, indeed, get something out of noth-
ing. Recasting statistical evidence from the laboratory, the clinic, or
the courtroom in terms of natural frequencies can make our minds
smarter. This is because, when human minds come into play, infor-
mation is not an independent quantity. Crucially, it depends on
how it is represented. Most of us grasp this instinctively, in many
cases systematizing, schematizing, and standardizing information
far beyond the requirement of logical satisfiability.
All this is to say that, even if brains are minds, brains are more
complex than minds – which is another way of saying that what we
think with is more complex than what we think about. But the men-
tal picture of the world registered by our minds harmonizes with
what evolution has embedded in our brains, because, at the end of
the day, we are talking about two sides of the same evolutionary
coin.
 11

Mind Your Business

100,000 Years; You’re Flying Low, Mac; Neuromancer;

See Further; The Making of a Fly

100,000 years

The information highway is a binary traffic system no different from

any other traffic system in at least one respect: it is insatiable. The
more highways we build, the more cars people buy and the more
congestion we see. It is the same in virtual space. The more broad-
band capacity that providers provide, the more Internet traffic
appears on the highway, necessitating an expansion of throughput
and storage which, in turn, brings in more traffic, more through-
put, and so on, around and around.
A mere generation after the birth of the World Wide Web, the
quantity of information online far outstrips what any non-artificial
intelligence could fathom, even in a million years. Human sensory
and thus information-bearing capacities have, after all, remained
essentially unchanged for at least the last 100,000 years. We have
plugged the Stone Age brain straight into the Space Age, which
spawns information highways at an ever more frantic pace, burying
bigger and bigger massifs of data under even bigger and bigger
massifs of data.
Mining for information no longer resembles looking for the
proverbial needle in a haystack so much as looking for a sharp nee-
dle in the Himalayas of blunt ones. And the postmodem civiliza-
tion of tomorrow will need all the help it can get in sorting out the
terabytes and then zettabytes of information sloshing back and
forth in cyberspace. How? We make prostheses that boost our phys-
ical capacities, from bulldozers that push urban trash around in
198 From Literature to Biterature

We have plugged the Stone Age brain straight into the Space Age, which
spawns information highways at an ever more frantic pace, burying bigger
and bigger massifs of data under even bigger and bigger massifs of data.

dumps to trash compactors. Why not mental prostheses in the

form of Internet filters to sift out valuable data from cybertrash –
for starters, where science and technology are concerned? (Things
are, as always, a thousand times more gnarly when it comes to
culture.)
It is guesstimated that up to 15 per cent of research undertaken
worldwide is superfluous insofar as it duplicates work already been
done by someone else. Informational filters that could liberate 15
per cent of brainpower on the global scale, even if this were their
sole benefit, would be a windfall the consequences of which could
hardly be overestimated. Our civilization would find itself in a posi-
tion of a beleaguered general who gets a 15 per cent troop rein-
forcement in the heat of battle.
Of course, filtering data from among the quintillions of infobits
that clog the Internet and its multifold incarnations, from YouTube
to Google Scholar, has never posed much of a problem. But filter-
 Mind Your Business 199

ing valuable data is a different story, because to find what is valu-

able, we need to be able to evaluate what we find.
Borges spoofed the difference in one of his ficciones by compiling
a brilliantly arbitrary bestiary, whereby animals are sorted out as
follows: (a) those that belong to the Emperor, (b) embalmed ones,
(c) those that are trained, (d) suckling pigs, (e) mermaids, (f) fab-
ulous ones, (g) stray dogs, (h) those that are included in this clas-
sification, (i) those that tremble as if they were mad, (j) innumer-
able ones, (k) those drawn with a very fine camel’s hair brush, (l)
others, (m) those that have just broken a flower vase, and (n) those
that resemble flies from a distance.

you’re flying low, mac

To filter out intelligence, you need intelligent filters. Searching by

keyword, syntax, number of citations, or any other non-semantic
parameter will never beat the results of an intelligent search for
intelligence. Keywords cannot pick out a game-changer from a
pool of bush leaguers who use the same lexicon. Syntax filters will
identify correctly or complexly formulated sentences but will not
spot a mastermind amongst mediocrity. Citation indices track not
so much breakthrough thinking, which by default commands few
adherents, as paradigm faddism and careerist genuflecting – sci-
entists being no saintlier than the population at large.
Long story short, when it comes to mining data for signs of intel-
ligence, nothing beats intelligence. Google puts millions of web-
pages at your fingertips in a split second, but it is up to you to
access the first few hits (research shows that hardly anyone ven-
tures past page 1), scan the contents, and determine if and how
they are relevant to our search – something that the current gen-
eration of search bots is simply unable to do.
In 2011, Larry Page, chief executive of Google, acknowledged as
much by remarking that the ultimate search engine would under-
stand exactly what you mean and give back exactly what you want.
Of course, to understand exactly what you mean, Internet intelli-
gence will first need to get smart enough to parse intentions –
smart enough, in other words, to pass the Turing test. Knowing
this, Google, which has already invested a fortune and a half in ai,
is leading the research into making search engines understand
human-level, intention-based communication.
200 From Literature to Biterature

While it might appear that there could be communication with-

out intentions, such as when dna trades information with a habitat
via an organism, this is not yet communication proper. The differ-
ence is crucial, roughly comparable to the difference between
today’s canned feedback that “Your Mac battery level is low” and
tomorrow’s computhor telling you something by clearing its
throat, discreetly playing a sound of wings aflutter, and, if you still
don’t get it, whispering, “You’re flying low, Mac.”
Of course, Google is not alone in trying to make Internet filters
smart and then some. darpa, the European Union, and the WWW
Consortium are all shovelling monies at Resource Description
Framework (rdf) designed to encode and express meaning via
Extensible Markup Language (xml). Their professed grail? The
Semantic Web, capable of intuiting the intentions behind even
vague queries and locating the right answer anywhere, in any lan-
guage, in any form and medium.

NEUROMANCER

Fictions like Neuromancer like to dramatize the moment when arti-

ficial intelligence makes a leap into sentience and beyond. In Gib-
son’s novel, the Turing Police vainly struggle to intercept a super-
computer that plots to augment its intellectual capacities with a
little help from a bunch of renegade hard-disc jockeys. In reality,
the intelligence barrier will collapse without a bang or even a
whisper.
Google will simply incrementally upgrade and finesse its search
algorithms, eventually allowing users to pose questions and have
answers enunciated in a language of their choice, as opposed to
typing Boolean search terms and having reams of raw Internet
pages spewed back. Somewhere on the way the search engines will
begin to display intelligence just because they will have to in order
to cope with vaguely phrased and open-ended queries.
As in the evolution of every business product, at first it will be
merely a matter of making the search experience more enjoyable
by letting users interact with a face animated on the screen, made
attractive to the eye of the beholder by a suite of personalizable
preferences and attuned to their search needs. “Beatles in Bangla-
desh?” you will type, and Victor or Victoria will smile, blink while
scanning billions of documents in order to retrieve what you are
 Mind Your Business 201

after (you’re writing a paper on entomology and not Band Aid),

and recite the answer to you.
Later on, to make the results better and the interactive experi-
ence more satisfying, the search engine will try to anticipate what
you may be looking for. Every Friday, for the past five weeks, you
have browsed online sports listings for sailing events to watch on
the weekend. One day Victoria will flash her perfect smile and, like
the perfect service provider she is, ask you if it is going to be the
usual. Gratefully, you will nod – it’s been a hard week after your
party has taken a beating in the polls – and off she goes to fetch the
links to the roboat races you just asked for. Or did you?
Time passes, and gradually Victoria takes over other scheduling
routines for you. And she excels in them. She never forgets that at
the end of each month you need to run spreadsheets to calculate
your expense accounts, that you need to run a spellcheck on the
funding report for the congressional committee, that you run the
risk of being decapitated if you do not send the anniversary card to
your wife this Monday, and that you need to run to the pharmacy
to get more headache pills, since the ones you bought online a
month ago are running out.
With improvements in biometrics, Victoria – like everyone else,
you have anthropomorphized the talking head or even torso, by
now a three-dimensional plastiné artifact emoting back at you from
your desktop – begins to take account of your moods as you bustle
about your business. Not only does she correct the French accents
that you always get wrong but she sympathizes with the pressures
you are under to finish that controversial education reform paper.
For a while now, she has been supplying you with mots justes better
than anything you would have used in your policy drafts, and late-
ly you have had her fix the solecisms in your introductions and con-
clusions which, as she knows, are always hardest to write.

see further

You could never pinpoint the moment when you started asking Vic-
toria about piddling things, like where to find the cheapest cherry
tomatoes of the kind you favour, or what garments to pack for the
trip that will eventually prove as damp and rainy as she has gleaned
from the forecasts you did not even consult. Why bother, when she
does it better and faster? What is the most disaster-proof recipe for
202 From Literature to Biterature

There are matrimonial agencies doing brisk business bringing singles togeth-
er, judicial courts where professional counsel is dispensed free of charge,
confessionals that issue poignant solace and absolution, monetary and macro-
economic policies that keep unemployment and inflation on the leash, inter-
national tribunals that successfully arbitrate the beefs of all superpowers in a
matter of seconds.

the flan you promised to bake for the twins’ birthday? What is the
best school to send the older kid to? How to get on the good side
of that strutting incumbent?
One day, as you crack your head open over a forthcoming
address that could determine your future in the party, Victoria
comes up aces in the form of a draft that, you have to admit, is bril-
liant. You bask in the plaudits, and next time you do not hesitate to
solicit her help with that major policy memo. After all, no one will
know where it came from and, besides, that competing analysis cir-
culated by the opposition is nothing short of masterful, so much
better than what anyone you know could write, including you. How
did they ever come up with it?
Times passes. There are matrimonial agencies doing brisk busi-
ness bringing singles together, judicial courts where professional
 Mind Your Business 203

counsel is dispensed free of charge, confessionals that issue poig-

nant solace and absolution, monetary and macroeconomic poli-
cies that keep unemployment and inflation on the leash, interna-
tional tribunals that successfully arbitrate the beefs of all super-
powers in a matter of seconds. Human-run businesses and service
providers have long since stopped trying to pass themselves off as
ai, knowing they could never match the performance, not to men-
tion that they would risk running afoul of penalties for deceiving
the public.
The search engines that by now run all the affairs of the world,
from federal governments to mortgage companies to local unions
of pet-walkers, are intermeshed and at the same time dispersed
around the globe. There is only one net, with every sensor and
every processor plugged into it, with intelligence distributed over
trillions of environment-monitoring, data-processing, and priority-
negotiating individual nanocells and their fluid assemblies.
In this phase, a qualitatively new phenomenon will manifest itself
on the planetary scale: the environment will have become more
intelligent than its inhabitants. As such, it will exercise the prerog-
ative of intelligence and override our individual and collective
decisions, being able to see further, process faster, judge better,
arbitrate more even-handedly, and legislate less corruptly – all the
while in concert with our interests as it interprets them.

THE MAKING OF A FLY

What is going to happen when the Semantic Web eventually breaks

out into the physical world, as inevitably it must if it is to perform
the way we want it to? Is it going to prove itself equal to its physical
– as opposed to merely informational – environment, or is it going
to prove itself an experiential moron?
Algorithms may be cleverer than humans within their field of
expertise, but they do not have our sense of perspective. This
became spectacularly evident in 2011 when Amazon’s price-setting
code went to war with itself. The Making of a Fly, a book about the
molecular biology of a fly from egg to fully fledged insect, may be
a riveting read for some, but it hardly deserved a price tag of more
than $23 million, which it reached after the algorithms used by
Amazon to set and update prices started a bidding war against one
another.
204 From Literature to Biterature

Such occasional stupidity notwithstanding, the future belongs to

artificial intelligence. In 2011, Eric Schmidt, executive chairman
of Google, made it very clear by hinting that in the future we will
turn to a search engine to ask about which colleges to apply to or
which books to read next. What then? Remember, you cannot have
intelligence without the potential for self-modification under pres-
sure from the environment, whether you call it learning or inde-
pendent thinking.
Are Google’s servers going to run bets on horse races on the side
while taking and processing ours? Will the fifteen thousand net-
works distributed over seven million–plus devices registered to the
US Department of Defense cunningly inflate the Pentagon networks
budgets without anyone being the wiser? Will algo traders stash
away the billions they will make trading for themselves in the wake
of economic crashes and booms that succeed each other in the
seven hundred trillion ocean of stocks, bonds, currencies, securi-
ties, and derivatives? Will they engineer financial markets melt-
downs because that is when the biggest fortunes are made?
Idle speculation? Consider what happened in May 2010 when
the Dow Jones nosedived six hundred points in five minutes for no
apparent reason whatsoever. Why did it do so, and why ten minutes
after that did it go back up again? To this day analysts scratch their
heads as to what happened. The only thing they agree on is that
the “flash crash” had something to do with the algorithms.
Still sceptical? In August 2012, equipped with a new, albeit thor-
oughly tested trading algorithm, Knight Capital linked up to the
New York Stock Exchange intending to make a killing. The instant
it went online, something went spectacularly wrong, although to
this day no one knows exactly what. From the outside it looked as
if the meticulously tested program went loco, buying high and sell-
ing low thousands of times a second. When it was ditched at the
end of forty-five minutes, the company was almost half a billion in
the red.
One thing is certain: if algos ever go native, no amount of Tur-
ing policing will stop them.
 11

Notes

lemmata

1 Solotaroff, “Genius of Stanislaw Lem,” 186.

chapter one

1 Lem, Imaginary Magnitude, 41.

2 The second is from Baudrillard.
3 In Simons, Are Computers Alive, 71.
4 Chamberlain, http://www.atariarchives.org/deli/write_about
_itself.php.
5 Ibid.
6 Poe, Tales of the Grotesque, vol. 1, 339, 343.

chapter two

1 Lem, “History of Bitic Literature,” 47.

2 Ibid., 58.
3 Reprinted in Bowden, Faster than Thought, 398.
4 Peat, Artificial Intelligence, 311.

chapter three

1 Lem, “History of Bitic Literature,” 42.

2 Michie and Johnston, Creative Computer, 71.
3 Wakefield, “When Algorithms Control the World.”
4 Cohen, “Colouring without Seeing.”
5 Franke, Computer Graphics, Computer Art, 119.
206 Notes to pages 72–151

chapter four

1 Graham’s Magazine, May 1842.

2 Peggy Kamuf, in Kelly, Encyclopedia of Aesthetics, vol. 2, 11.

chapter five

1 Colby et al., “Turing-Like Indistinguishability Tests,” 202.

2 Turing, “Computing Machinery and Intelligence,” 13.
3 Jacquette, “Who’s Afraid of the Turing Test?,” 64.
4 McCulloch, “Mysterium Iniquitatis,” 36.
5 “Computing Machinery and Intelligence,” 14.
6 Dennett, “Fast Thinking,” 334.

chapter six

1 Searle, Mystery of Consciousness, 61.

2 Hofstadter, I Am a Strange Loop, 30.
3 Simon and Newall, “Heuristic Problem Solving,” 1–10.
4 In Copeland, The Essential Turing, 407.
5 Millar, “On the Point of the Imitation Game,” 595; Block, “Troubles,”
248, Hodges, Alan Turing, 415.
6 French, “The Turing Test,” 115; Prager, On Turing, 45.
7 In Copeland, The Essential Turing, 494.
8 Turing, “Computing Machinery and Intelligence,” 14.
9 French, “Subcognition and the Limits of the Turing Test,” 59.

chapter seven

1 McCorduck, Machines Who Think, 464.

2 Ibid.
3 Weizenbaum, Computer Power and Human Reason, 10
4 Tomasello, Origins of Human Communication, 134.
5 McCarthy, “Problems and Projections,” 615.

chapter eight

1 Hayles, Electronic Literature, 47; Shurkin, “Expert Systems,” 72.

2 Turkle, Life on the Screen; Benford and Malartre, Beyond Human.
3 Simons, Are Computers Alive, 14.
 Notes to pages 151–64 207

4 Ibid., 17.
5 Ibid., x.
6 Ibid., 109.
7 Ibid., 22.
8 “Software für die Seele” (Software for the soul), Der Spiegel, 5 Septem-
ber 1994, 116.
9 Fitzpatrick, “No, Robot.”
This page intentionally left blank
 11

Bibliography

Aarseth, Espen J. Cybertext: Perspectives on Ergodic Literature. Baltimore, md:

Johns Hopkins University Press 1997.
Aitchison, Jean. The Seeds of Speech: Language Origin and Evolution. Cam-
bridge: Cambridge University Press 2000.
Allen, Robert F. “The Stylo-Statistical Method of Literary Analysis.” Com-
puters and the Humanities 22 (1988): 1–10.
Anderson, Alan Ross, ed. Minds and Machines. Englewood Cliffs, nj:
Prentice-Hall 1964.
Ayer, Alfred J. Language, Truth and Logic. London: Victor Gollancz 1936.
Barbrook, Adrian C., Christopher J. Howe, Norman Blake, and Peter
Robinson. “The Phylogeny of The Canterbury Tales.” Nature 394 (27
August 1998): 839.
Baron-Cohen, Simon. “The Empathizing System: A Revision of the 1994
Model of the Mindreading System.” In Origins of the Social Mind: Evolu-
tionary Psychology and Child Development, edited by Bruce J. Ellis and
David F. Bjorklund, 468–92. New York: Guilford 2005.
Baron-Cohen, Simon, Helen Tager-Flusberg, and Donald J. Cohen, eds.
Understanding Other Minds: Perspectives from Developmental and Cognitive
Neuroscience. 2nd ed. Oxford: Oxford University Press 2000.
Baudrillard, Jean. “Simulacra and Simulations.” In Jean Baudrillard, Select-
ed Writings, edited by Mark Poster, 166–84. Stanford: Stanford Univer-
sity Press 1998.
bbc. “ibm Researchers Make 12-Atom Magnetic Memory Bit.” 13 January
2012. http://www.bbc.co.uk/news/technology-16543497.
Bedau, Mark A., and Paul Humphreys, eds. Emergence: Contemporary Read-
ings in Philosophy and Science. Cambridge, ma: mit Press 2008.
Behne, T., M. Carpenter, J. Call, and M. Tomasello. “Unwilling versus
210 Bibliography

Unable? Infants’ Understanding of Intentional Action.” Developmental

Psychology 41 (2005): 328–37.
Behne, T., M. Carpenter, and M. Tomasello. “One-Year-Olds Compre-
hend the Communicative Intentions behind Gestures in a Hiding
Game.” Developmental Science 8 (2005): 492–9.
Benford, Gregory, and Elisabeth Malartre. Beyond Human: Living with
Robots and Cyborgs. New York: Forge Books 2007.
Bennett, Maxwell, Deniel C. Dennett, Peter Hacker, and John Searle.
Neuroscience and Philosophy: Brains, Minds, and Language. New York:
Columbia University Press 2009.
Bernal, J.D. The Origins of Life. London: Weidenfeld & Nicholson 1967.
Berrill, Norman John. Biology in Action. London: Heinemann 1966.
Block, Ned. “Troubles with Functionalism.” In Perception and Cognition:
Issues in the Foundations of Psychology, edited by C.W. Savage. Vol. 9.
Minneapolis: University of Minnesota Press 1978.
– “The Computer Model of the Mind.” In Thinking, edited by D.N. Osh-
erson and H. Lasnik. Vol. 3, An Invitation to Cognitive Science. Cam-
bridge, ma: mit Press 1990.
Boden, Margaret A. Artificial Intelligence and Natural Man. Brighton: Har-
vester 1977.
– Mind as Machine: A History of Cognitive Science. Oxford: Oxford Univer-
sity Press 2006.
Bolton, W.F. “A Poetic Formula in Beowulf and Seven Other Old English
Poems: A Computer Study.” Computers and the Humanities 19 (1985):
167–73.
Borges, Jorge Luis. “The Analytical Language of John Wilkins.” Other
Inquisitions, 1937–1952. Translated by Ruth L.C. Simms. Austin: Uni-
versity of Texas Press 1993.
Bowden, B.V., ed. Faster than Thought: A Symposium on Digital Computing
Machines. New York: Pitman 1953.
Bratman, Michael. “Shared Co-Operative Activity.” Philosophical Review
101 (1992): 327–41.
Bruner, Jerome. Child’s Talk. New York: Norton 1983.
Budiansky, Stephen. If a Lion Could Talk: How Animals Think. London:
Weidenfeld & Nicolson 1998.
Burrows, John F. “Computers and the Study of Literature.” In Computers
and Written Texts, edited by Christopher Butler. Oxford: Blackwell 1992.
Buss, David M. Evolutionary Psychology: The New Science of the Mind. Boston:
Allyn & Bacon 1999.
 Bibliography 211

Cacioppo, John T., and William Patrick. Loneliness: Human Nature and the
Need for Social Connection. New York: Norton 2008.
Camerer, Colin, George Lowenstein, and Drazen Prelec. “Neuroeconom-
ics: How Neuroscience Can Inform Economics.” Journal of Economic
Literature 43 (2005): 9–64.
Caras, Pauline. “Literature and Computers: A Short Bibliography,
1980–1987.” College Literature 15 (1988): 69–82.
Carroll, Lewis. Symbolic Logic. London: Macmillan 1896.
Ceruzzi, Paul E. Reckoners: The Prehistory of the Digital Computer, from Relays
to the Stored Program Concept, 1935–1945. Westport, CT: Greenwood
Press 1983.
Chamberlain, Bill. “Getting a Computer to Write about Itself.” 1984.
http://www.atariarchives.org/deli/write_about_itself.php.
Charabia. http://www.charabia.net/gen/index.php.
Christian, Brian. The Most Human Human: What Talking with Computers
Teaches Us about What It Means to Be Alive. New York: Doubleday 2011.
Clark, Herbert. Using Language. Cambridge: Cambridge University Press
1996.
Clarke, David S. Panpsychism and the Religious Attitude. Albany, ny: suny
Press 2003.
– ed. Panpsychism: Past and Recent Selected Readings. Albany, ny: suny
Press 2004.
Cohen, Gillian. The Psychology of Cognition. London: Academic 1977.
Cohen, Harold. “Colouring without Seeing: A Problem in Machine Cre-
ativity.” (1999). http://www.aaronshome.com/aaron/publications/
colouringwithoutseeing.pdf.
Cohen, Paul R. “If Not Turing’s Test, Then What?” AI Magazine 26, no. 4
(2006). http://www.cs.arizona.edu/~cohen/Publications/papers/
IfNotWhat.pdf.
Colby, Kenneth M. “Modeling a Paranoid Mind.” Behavioural and Brain
Sciences 4 (1981): 515–60.
Colby, Kenneth M., and Robert J. Stoller. Cognitive Science and Psycho-
analysis. New York and London: Routledge 1988.
Colby, Kenneth M., F.D. Hilf, S. Weber, and H.C. Kraemer. “Turing-Like
Indistinguishability Tests for the Validation of Computer Simulation of
Paranoid Processes.” Artificial Intelligence 3 (1972): 199–221.
Collins, H.M. Artificial Experts: Social Knowledge and Intelligent Machines.
Cambridge, ma: mit Press 1990.
Collins, Nick. Introduction to Computer Music. Chichester: Wiley 2009.
212 Bibliography

Conway, John H. On Numbers and Games. 2nd ed. Wellesley, ma: A.K.
Peters 2001.
Copeland, B. Jack., ed. The Essential Turing: The Ideas That Gave Birth to
the Computer Age. Oxford: Oxford University Press 2004.
Corns, Thomas N., and Margarette E. Smith. “Computers in the Human-
ities: Methods and Applications in the Study of English Literature.”
Literary and Linguistic Computing 6 (1991): 127–30.
– “Literature.” In Information Technology in the Humanities: Tools, Tech-
niques, and Applications, edited by Sebastian Rahtz. Chichester, ny:
Wiley 1987.
Damasio, Antonio. Looking for Spinoza: Joy, Sorrow, and the Feeling Brain.
Orlando, fl: Harcourt 2003.
Davidson, Donald. “Mental Events.” In Experience and Theory, edited by L.
Foster and J. Swanson. London: Duckworth 1970.
Deacon, Terrence. The Symbolic Species: The Co-Evolution of Language and
the Human Brain. London: Penguin 1997.
de la Mettrie, Julien Offray. Machine Man and Other Writings. Translated
by Ann Thomson. Cambridge: Cambridge University Press 1996.
Dennett, C. Daniel. Consciousness Explained. Boston: Little, Brown 1991.
– “Fast Thinking.” The Intentional Stance. Cambridge, ma: mit Press
1987.
De Waal, Frans. The Age of Empathy: Nature’s Lessons for a Kinder Society.
New York: Crown 2009.
Dimberg, U., M. Thunberg, and K. Elmehed. “Unconscious Facial Reac-
tions to Emotional Facial Expressions.” Psychological Science 11 (2000):
86–9.
Dreyfus, Hubert. What Computers Can’t Do: A Critique of Artificial Reason.
New York: Harper & Row 1972.
– What Computers Still Can’t Do. Cambridge, ma: mit Press 1979.
Dunbar, Robin, and Louise Barrett. Oxford Handbook of Evolutionary Psy-
chology. New York: Oxford University Press 2007.
Dyson, George. Darwin among the Machines: The Evolution of Global Intelli-
gence. New York: Basic 1998.
Eco, Umberto. Baudolino. San Diego: Harcourt 2000.
Economist, The. “Can Machines Surpass Humans in Intelligence?” 12
March 2011, 86.
– “Intelligence Testing: Who Are You Calling Bird-Brained?” 5 March
2011, 79.
– “Painting by Numbers.” 30 July 2011, 67.
Edelman, Gerald M. Bright Air, Brilliant Fire. New York: Basic 1992.
 Bibliography 213

Enfield, N.J., and Stephen C. Levinson, eds. Roots of Human Sociality: Cul-
ture, Cognition and Interaction. Oxford: Berg 2006.
Enriquez, Juan, and Steve Gullans. Homo Evolutis. Amazon Media:
Kindlebook 2011.
Farley, B.G., and W.A. Clark. “Simulation of Self-Organizing Systems by
Digital Computer.” Institute of Radio Engineers Transactions on Informa-
tion Theory 4 (1954): 76–84.
Feigenbaum, Edward A. “Some Challenges and Grand Challenges for
Computational Intelligence.” Journal of the ACM 50, no. 1 (2003):
32–40.
Fitzpatrick, Michael. “No, Robot: Japan’s Elderly Fail to Welcome Their
Robot Overlords.” 4 February 2011. http://www.bbc.co.uk/news/
business-12347219.
Focquaert, Farah, and Steven M. Platek. “Social Cognition and the Evo-
lution of Self-Awareness.” In Evolutionary Cognitive Neuroscience, edited
by Steven M. Platek, Julian Paul Keenan, and Todd K. Shackelford,
457–97. Cambridge, ma: mit Press 2007.
Fortier, Paul A. “Theory and Practice in the Use of Computers for the
Study of Literature.” In Actes du VIIIe Congres de l’Association Interna-
tionale de Litterature Comparee/Proceedings of the 8th Congress of the Interna-
tional Comparative Literature Association, edited by Bela Kopeczi and
Gyorgy M. Vajda. Stuttgart: Bieber 1980.
Franke, Herbert W. Computer Graphics, Computer Art. Translated by Gustav
Metzger. New York: Phaidon 1971.
Freeman, Anthony, ed. Consciousness and Its Place in Nature: Does Physical-
ism Entail Panpsychism? Exeter: Imprint Academic 2006.
French, Robert M. “Subcognition and the Limits of the Turing Test.”
Mind 99 (1990): 53–65.
– “The Turing Test: The First 50 Years.” Trends in Cognitive Sciences 4
(2000): 115–22.
Gardner, Howard. Frames of Mind: The Theory of Multiple Intelligences. New
York: Basic 1983.
Genova, J. “Turing’s Sexual Guessing Game.” Social Epistemology 8, no. 4
(1994): 314–26.
Gibson, William Neuromancer. London: Victor Gollancz 1984
Gödel, Kurt. Collected Works. Vol. 2, Publications, 1938–1974. Oxford:
Oxford University Press 1990.
Goleman, Daniel. Social Intelligence: The New Science of Human
Relationships. New York: Arrow 2007.
Gomez, Juan-Carlos. “The Emergence of Intentional Communication as
214 Bibliography

a Problem-Solving Strategy in the Gorilla.” In “Language” and Intelli-

gence in Monkeys and Apes, edited by S.T. Parker and K.R. Gibson. Cam-
bridge: Cambridge University Press 1990.
– Apes, Monkeys, Children, and the Growth of Mind. Cambridge, ma: Har-
vard University Press 2004.
Gopnik, Alison. The Philosophical Baby: What Children’s Minds Tell Us about
Truth, Love, and the Meaning of Life. New York: Farrar, Straus and
Giroux 2009.
Gregory, Richard, and Pauline K. Marstrand, eds. Creative Intelligence.
Norwood, nj: Ablex 1987.
Gurewich, David. “Games Computers Play.” New Criterion 7 (1989): 81–4.
Harnad, Stevan. “The Annotation Game: On Turing (1950) on Comput-
ing, Machinery, and Intelligence.” In The Turing Test Sourcebook: Philo-
sophical and Methodological Issues in the Quest for the Thinking Computer,
edited by Robert Epstein and Grace Peters. Dordrecht and Boston:
Kluwer 2004.
Haugeland, John, ed. Mind Design. Cambridge, ma: mit Press 1981.
Hayes, Patrick, and Kenneth Ford. “Turing Test Considered Harmful.”
In Proceedings of the Fourteenth International Joint Conference on Artificial
Intelligence, 972–97. San Francisco: Morgan Kaufmann 1995.
Hayles, Katherine. Electronic Literature: New Horizons for the Literary. Notre
Dame, in: University of Notre Dame Press 2008.
Heil, John. Philosophy of Mind: A Contemporary Introduction. London and
New York: Routledge 1998.
Henrich, Joseph, Robert Boyd, Samuel Bowles, Colin Camerer, Ernst
Fehr, Herbert Gintis, Richard McElreath, Michael Alvard, Abigail Barr,
Jean Ensminger, Kim Hill, Francisco Gil-White, Michael Gurven,
Frank Marlowe, John Q. Patton, Natalie Smith, and David Tracer.
“‘Economic Man’ in Cross-Cultural Perspective: Ethnography and
Experiments from Fifteen Small-Scale Societies.” Behavioral and Brain
Sciences 28 (2005): 795–855.
Hernandez-Orallo, Jose. “Beyond the Turing Test.” Journal of Logic, Lan-
guage and Information 9, no. 4 (2000): 447–66.
Hernandez-Orallo, Jose, and D.L. Dowe. “Measuring Universal Intelli-
gence: Towards an Anytime Intelligence Test.” Artificial Intelligence Jour-
nal 174 (2010): 1508–39.
Hodges, Andrew. Alan Turing: The Enigma. New York: Simon & Schuster
1983.
Hoffman, Elizabeth, Kevin McCabe, and Vernon L. Smith. “On Expecta-
 Bibliography 215

tions and the Monetary Stakes in Ultimatum Games.” International

Journal of Game Theory 25 (1996): 289–301.
Hofstadter, Douglas R. Gödel, Escher, Bach: An Eternal Golden Braid. New
York: Vintage 1979.
– I Am a Strange Loop. New York: Basic 2007.
Hofstadter, Douglas R., and Daniel C. Dennett, eds. The Mind’s I: Fan-
tasies and Reflections on Self and Soul. New York: Basic 1981.
Horn, William Dennis. “The Effect of the Computer on the Written
Word.” In Sixth International Conference on Computers and the Humanities,
edited by Sarah K. Burton and Douglas D. Short. Rockville, md: Com-
puter Science Press 1983.
Horowitz, Irving Louis. “Printed Words, Computers and Democratic
Societies.” Virginia Quarterly Review 59 (1983): 620–36.
Iacoboni, Marco. Mirroring People: The Science of Empathy and How We Con-
nect with Others. New York: Picador 2008.
Ide, Nancy M. “A Statistical Measure of Theme and Structure.” Computers
and the Humanities 23 (1989): 277–83.
itrs: International Technology Roadmap for Semiconductors. 2012.
http://www.itrs.net/.
Jacquette, Dale. “Who’s Afraid of the Turing Test?” Behavior and Philoso-
phy 20/21 (1993): 63–74.
Kahneman, Daniel. Thinking, Fast and Slow. New York: Farrar, Straus &
Giroux 2011.
Kasparov, Gary. “The Day That I Sensed a New Kind of Intelligence.”
Time, 1 April 1996, 57.
Kauffman, Stuart A. At Home in the Universe: The Search for Laws of Self-
Organization and Complexity. Oxford: Oxford University Press 1995.
– Reinventing the Sacred: A New View of Science, Reason and Religion. New
York: Basic 2008.
Kelly, Michael, ed. Encyclopedia of Aesthetics. 4 vols. New York: Oxford Uni-
versity Press 1998.
Keltner, Dacher. Born to Be Good: The Science of a Meaningful Life. New
York: W.W. Norton 2009.
Kern, Alfred, and James F. Sheridan. “basic Poetry: The Computer as
Poet.” In Sixth International Conference on Computers and the Humanities,
edited by Sarah K. Burton and Douglas D. Short. Rockville, md: Com-
puter Science Press 1983.
Kurzweil, Ray. The Age of Intelligent Machines. Cambridge, ma: mit Press
1990.
216 Bibliography

– The Singularity Is Near. London: Penguin 2005.

Kybernetes 4 (April 2010). Special Turing issue.
Leavitt, David. The Man Who Knew Too Much: Alan Turing and the Invention
of the Computer. New York: Norton 2006.
Leebaert, Derek, ed. Technology 2001: The Future of Computing and Commu-
nications. Cambridge, ma: mit Press 1991.
Legg, Shane, and Marcus Hutter. “Universal Intelligence: A Definition of
Machine Intelligence.” Minds and Machines 17, no. 4 (2007):
391–444.
Lem, Stanislaw. Bomba megabitowa. Cracow: Wydawnictwo Literackie
1999.
– “A History of Bitic Literature.” Imaginary Magnitude. Translated by
Marc E. Heine. San Diego: Harcourt 1985.
– Perfect Vacuum. Translated by Michael Kandel. Evanston, il: Northwest-
ern University Press 1971.
– “Smart Robots.” In The Art and Science of Stanislaw Lem, edited by Peter
Swirski. Montreal and London: McGill-Queen’s University Press 2006.
– Solaris. Translated by Joanna Kilmartin and Steve Cox. New York: Walk-
er & Co. 1971.
– Summa technologiae. Cracow: Wydawnictwo Literackie 1971.
– Tajemnica chińskiego pokoju. Cracow: Universitas 1996.
Lenat, Douglas. “Eurisco: A Program That Learns New Heuristics and
Domain Concepts.” Artificial Intelligence 21 (March 1983): 61–98.
Levy, David. Love and Sex with Robots: The Evolution of Human-Robot Rela-
tionships. New York: Harper Perennial 2007.
Liszkowski, Ulf. “Human Twelve-Month-Olds Point Co-operatively to
Share Interest and Provide Information for a Communicative Part-
ner.” Gesture 5 (2005): 135–54.
– “Infant Pointing at Twelve Months: Communicative Goals, Motives,
and Social-Cognitive Abilities.” In Roots of Human Sociality: Culture, Cog-
nition and Interaction, edited by Nicholas J. Enfield and Stephen C.
Levinson. Oxford: Berg 2006.
Lucas, John R. “Minds, Machines, and Gödel.” In Minds and Machines,
edited by Alan Ross Anderson. Englewood Cliffs, nj: Prentice-Hall
1964.
Lupkowski, Pawel. “Some Historical Remarks on Block’s ‘Aunt Bubbles’
Argument.” Mind and Machines 16, no. 4 (2006): 437–41.
Lycan, William G. Consciousness. Cambridge, ma: mit Press 1995.
– Consciousness and Experience. Cambridge, ma: mit Press 1996.
 Bibliography 217

– ed. Mind and Cognition: An Anthology. 2nd ed. Maiden, ma: Blackwell
1999.
Manning, Peter. Electronic and Computer Music. Oxford: Oxford University
Press 2004.
Marcus, Stephen. “Computers and the Poetic Muse.” In Sixth Internation-
al Conference on Computers and the Humanities, edited by Sarah K. Bur-
ton and Douglas D. Short. Rockville, md: Computer Science Press
1983.
McCarthy, John. “Problems and Projections in cs for the Next Forty-
Nine Years.” Journal of the ACM 50, no. 1 (2003): 73–9.
– “Programs with Common Sense.” Symposium on Mechanization of
Thought Processes. National Physical Laboratory, Teddington, uk, 1958.
http://www-formal.stanford.edu/jmc/mcc59.pdf.
– “Review of The Emperor’s New Mind by Roger Penrose.” Bulletin of the
American Mathematical Society 23 (October 1990): 606–16.
McCorduck, Pamela. Machines Who Think. 2nd ed. Wellesley, ma: A.K.
Peters 2004.
McCullouch, Warren S. “Mysterium Iniquitatis of Sinful Man Aspiring
into the Place of God.” Scientific Monthly 80, no. 1 (1955): 35–9.
Meehan, James R. “Tale-Spin, an Interactive Program That Writes Sto-
ries.” In Proceedings of the Fifth International Conference on Artificial Intelli-
gence. Vol. 1. Pittsburg: Carnegie-Mellon University Department of Sci-
ence, 1977.
Melpomene [G.E. Hughes]. Bagabone, Hem ’I Die Now. New York: Vantage
1980.
Michie, Donald. On Machine Intelligence. 2nd ed. Chichester: Ellis Hor-
wood 1986.
– On Machine Intelligence, Biology and More, edited by Ashwin Sriniwasan.
Oxford: Oxford University Press 2009.
Michie, Donald, and Rory Johnston. The Creative Computer: Machine Intelli-
gence and Human Knowledge. Harmondsworth: Viking 1984.
Millar, P.H. “On the Point of the Imitation Game.” Mind 82 (1973): 595–7.
Miller, Richard. Fact and Method: Explanation, Confirmation and Reality in
the Natural and the Social Sciences. Princeton: Princeton University Press
1987.
Millican, Peter, and Andy Clark, eds. Machines and Thought: The Legacy of
Alan Turing. Vol. 1. Oxford: Oxford University Press 1996.
Minsky, Marvin. “Why People Think Computers Can’t.” AI Magazine (Fall
1982): 3–15.
218 Bibliography

Mitchell, T.M. Machine Learning. New York: McGraw-Hill 1997.

Moor, James H. “An Analysis of the Turing Test.” Philosophical Studies 30
(1976): 249–57.
– “The Status and Future of the Turing Test.” Minds and Machines 11,
no. 1 (2001): 77–93.
– ed. The Turing Test: The Elusive Standard of Artificial Intelligence. Dordrecht:
Kluwer 2003.
Moravec, Hans. Mind Children: The Future of Robot and Human Intelligence.
Cambridge, ma: Harvard University Press 1988.
– Robot: Mere Machine to Transcendent Mind. New York: Oxford University
Press 1998.
Munz, Peter. Philosophical Darwinism: On the Origin of Knowledge by Means
of Natural Selection. New York and London: Routledge 1993.
Negroponte, Nicholas. “The Return of the Sunday Painter.” In The Com-
puter Age: A Twenty-Year View, edited by Michael Dertouzos and Joel
Moses. Cambridge, ma: mit Press 1979.
Neumaier, O. “A Wittgensteinian View of Artificial Intelligence.” In Arti-
ficial Intelligence: The Case Against., edited by Reiner Born. London:
Croom Helm 1987.
Niesz, Anthony J., and Normal N. Holland. “Interactive Fiction.” Critical
Inquiry 11 (1984): 110–29.
Northcott, Bryan. “But Is It Mozart?” The Independent. 5 September 1997.
http://www.independent.co.uk/arts-entertainment/music/but-is-it-
mozart-1237509.html.
Ochs, E.P., M. Meana, K. Mah, and Y.M. Binik. “The Effects of Exposure
to Different Sources of Sexual Information on Sexual Behaviour:
Comparing a ‘Sex-Expert System’ to Other Educational Material.”
Behavioural Research Methods, Instruments, and Computers 25 (1993):
189–94.
Ochs, E.P., M. Meana, L. Pare, K. Mah, and Y.M. Binik. “Learning about
Sex outside the Gutter: Attitudes toward a Computer Sex-Expert Sys-
tem.” Journal of Sex and Marital Therapy 20 (1994): 86–102.
Peat, F. David. Artificial Intelligence: How Machines Think. New York: Bean
1988.
Peer, Willie van. “Quantitative Studies of Literature: A Critique and an
Outlook.” Computers and the Humanities 23 (1989): 301–7.
Perry, Mark, and Thomas Margoni. “From Music Tracks to Google Maps:
Who Owns Computer-Generated Works?” Computer Law and Security
Review 26 (2010): 621–9.
 Bibliography 219

Petzold, Charles. The Annotated Turing: A Guided Tour through Alan Tur-
ing’s Historic Paper on Computability and the Turing Machine. Indianapo-
lis: Wiley & Sons 2008.
Pihlström, Sami. Jamesian Perspectives on Mind, World, and Religion. Lan-
ham, md: University Press of America 2008.
Poe, Edgar Allan. “The Philosophy of Composition.” Orig. 1846. In The
Complete Works of Edgar Allan Poe, edited by J.A. Harrison, vol. 14,
193–208. New York: T.Y. Crowell 1902.
– Tales of the Grotesque and Arabesque. 2 vols. Philadelphia: Lea and Blan-
chard 1840.
– “Twice-Told Tales: A Review.” Graham’s Magazine (Philadelphia), May
1842.
Potter, Rosanne G. “Statistical Analysis of Literature: A Retrospective on
Computers and the Humanities, 1966–1990.” Computers and the
Humanities 25 (1991): 401–29.
Poundstone, William. The Recursive Universe: Cosmic Complexity and the
Limits of Scientific Knowledge. Oxford: Oxford University Press 1987.
Prager, John. On Turing. Belmont, ca: Wadsworth 2001.
Premack, David, and Ann James Premack. “Origins of Human Social
Competence.” In The Cognitive Neurosciences, edited by Michael S. Gaz-
zaniga, 205–18. Cambridge, ma: mit Press 1995.
Quine, Willard von. Word and Object. Cambridge, ma: mit Press 1960.
Racter. Endangered Software Archive. http://www.mirrors.org/archived
_software/www.techknight.com/esa/default.htm.
Racter [William Chamberlain and Thomas Etter]. The Policeman’s Beard Is
Half Constructed. New York: Warner 1984.
Rizzolatti, Giacomo, and Leonardo Fogassi. “Mirror Neurons and Social
Cognition.” In Oxford Handbook of Evolutionary Psychology, edited by
Robin Dunbar and Louise Barrett, 179–95. New York: Oxford Univer-
sity Press 2007.
Robinson, William S. Computers, Minds and Robots. Philadelphia: Temple
University Press 1992.
Roland, Alex, with Philip Shiman. Strategic Computing: DARPA and the Quest
for Machine Intelligence. Cambridge, ma: mit Press 2002.
Roper, John P.G., ed. Computers in Literary and Linguistic Research. Paris:
Champion 1988.
Russell, Stuart J., and Peter Norvig. Artificial Intelligence: A Modern
Approach. 2nd ed. Upper Saddle River, nj: Prentice Hall 2003.
Sanfey, Alan G., James K. Rilling, Jessica A. Aaronson, Leigh E. Nystrom,
220 Bibliography

and Jonathan D. Cohen. “The Neural Basis of Economic Decision-

Making in the Ultimatum Game.” Science 300 (2003): 1755–8.
Saxby, Steven. The Age of Information. New York: New York University
Press 1990.
Saygin, Ayse P. “Comments on ‘Computing Machinery and Intelligence’
by Alan Turing.” In Parsing the Turing Test, edited by R. Epstein, G.
Roberts, and G. Poland. Dordrecht: Springer 2008.
Saygin, Ayse P., and Ilyas Cicekli. “Pragmatics in Human-Computer Con-
versation.” Journal of Pragmatics 34, no. 3 (2002): 227–58.
Saygin, Ayse P., I. Cicekli, and V. Akman. “Turing Test: 50 Years Later.”
Minds and Machines 10, no. 4 (2000): 463–518.
Schaffer, Simon. “Making Up Discovery.” In Dimensions of Creativity, edit-
ed by Margaret Boden, 13–51. Cambridge, ma: mit Press 1994.
Schank, Roger C. Dynamic Memory: A Theory of Reminding and Learning in
Computers and People. Cambridge: Cambridge University Press 1982.
Schwartz, Richard Alan. “New Possibilities for Computer Literature.” In
Sixth International Conference on Computers and the Humanities, edited by
Sarah K. Burton and Douglas D. Short. Rockville, md: Computer Sci-
ence Press 1983.
Seager, William. “Consciousness, Information and Panpsychism.” In
Explaining Consciousness, edited by J. Shear. Cambridge, ma: mit Press
1997.
Searle, John. The Construction of Social Reality. New York: Free Press 1995.
– “Minds, Brains, and Programs.” Behavioral and Brain Sciences 3 (1980):
417–57.
– Minds, Brains, and Science. Cambridge, ma: Harvard University Press
1984.
– The Mystery of Consciousness. New York: New York Review of Books
1997.
Shah, Huma. “Turing’s Misunderstood Imitation Game and ibm’s Wat-
son Success.” http://reading.academia.edu/HumaShah/Papers/
464364/Turings_misunderstood_imitation_game_and_ibms_Watson_
success.
Shieber, Stuart M. “Lessons from a Restricted Turing Test.” Communica-
tions of the ACM 37, no. 6 (1994): 70–8.
Shiels Maggie. “Google Enters Digital Books War.” 5 May 2010.
http://www.bbc.co.uk/news/10098111.
Shurkin, Joel. Engines of the Mind: A History of the Computer. New York:
Norton 1984.
 Bibliography 221

– “Expert Systems: The Practical Face of Artificial intelligence.” Tech-

nology Review 86 (1983): 72–7.
Simon, Herbert A., and Allen Newall. “Heuristic Problem Solving: The
Next Advance in Operations Research.” Operations Research 6 (1958):
1–10.
Simons, Geoff. Are Computers Alive: Evolution and New Life Forms. Boston:
Birkhäuser 1983.
Skrbina, David. Panpsychism in the West. Cambridge, ma: mit Press 2005.
Smith, Paul. “The Joke Machine: Communicating Traditional Humour
Using Computers.” In Spoken in Jest, edited by Gillian Bennett.
Sheffield: Sheffield Academy Press 1991.
Sober, Elliott, and David Sloan Wilson. Unto Others: The Evolution and Psy-
chology of Unselfish Behavior. New York: Dutton 1999.
Solotaroff, Theodore. “The Genius of Stanislaw Lem.” A Few Good Voices
in My Head. New York: Harper & Row 1987.
Sorensen, Alan, Jonah Berger, and Scott Rasmussen. “Positive Effects of
Negative Publicity: Can Negative Reviews Increase Sales?” Marketing
Science 29 (2010): 5.
http://www.stanford.edu/~asorense/papers/Negative_Publicity2.pdf.
Spiegel, Der. “Software für die Seele.” 5 September 1994, 116–18.
Stine, G. Harry. The Untold Story of the Computer Revolution. New York:
Arbour 1985.
Stone, Valerie E. “Theory of Mind and the Evolution of Social Intelli-
gence.” Social Neuroscience: People Thinking about People, edited by John
T. Cacioppo, Penny S. Visser, and Cynthia L. Pickett, 103–30. Cam-
bridge, ma: Bradford-mit Press 2006.
Swirski, Peter, ed. The Art and Science of Stanislaw Lem. Montreal, London:
McGill-Queen’s University Press 2006.
– “Cogito Ergo Scribo: Stanislaw Lem and Evolution.” In The Evolutionary
Review: Art, Science and Culture, edited by Alice Andrews and Michelle
Scalise Sugiyama. New York: suny Press 2012.
– “Computhors and Biterature: Machine-Written Fiction?” SubStance 70
(1993): 81–90.
– Lem: Polemics: Essays, Letters, and Provocation. Montreal and London:
McGill-Queen’s University Press 2013.
– Literature, Analytically Speaking: Explorations in the Theory of Interpretation,
Analytic Aesthetics, and Evolution (Cognitive Approaches to Literature and
Culture). Austin: University of Texas Press 2010.
– Of Literature and Knowledge: Explorations in Narrative Thought Experi-
222 Bibliography

ments, Evolution, and Game Theory. New York and London: Routledge
2007.
– A Stanislaw Lem Reader (Rethinking Theory). Evanston, IL: Northwestern
University Press 1997.
Teuscher, Christof, ed. Alan Turing: Life and Legacy of a Great Thinker.
Berlin: Springer 2004.
Tomasello, Michael. “The Human Adaptation for Culture.” Annual
Review of Anthropology 28 (1999): 509–29.
– Origins of Human Communication. Boston: mit Press 2010.
– Why We Cooperate. Boston: mit Press 2009.
Tomasello, Michael, Malinda Carpenter, Josep Call, Tanya Behne, and
Henrike Moll. “Understanding and Sharing Intentions: The Origins of
Cultural Cognition.” Behavioral and Brain Sciences 28 (2005): 675–735.
Traiger, Saul. “Making the Right Identification in the Turing Test.” Minds
and Machines 10, no. 4 (2000): 561–72.
Turing, Alan. “Computing Machinery and Intelligence.” In Minds and
Machines, edited by Alan Ross Anderson. Englewood Cliffs, nj: Pren-
tice-Hall 1964. First published 1950.
– “Intelligent Machinery, A Heretical Theory: Report, National Physical
Laboratory.” Reprinted in Machine Intelligence 5 (1970): 3–23.
– “On Computable Numbers, with an Application to the Entscheidung-
sproblem.” Proceedings of the London Mathematical Society 42, no. 2 (1936):
230–65.
Turkle, Sherri. Life on the Screen: Identity in the Age of the Internet. New York:
Simon & Schuster 1995.
Vallverdú, Jordi, and David Casacuberta, eds. Handbook of Research on Syn-
thetic Emotions and Sociable Robotics: New Applications in Affective Comput-
ing and Artificial Intelligence. New York: igi Global 2009.
Young, Jeffrey R. “Programmed for Love.” Chronicle Review, 14 January
2001. http://chronicle.com/article/Programmed-for-Love-
The/125922/?sid=wb&utm_source=wb&utm_medium=en.
Wakefield, Jane. “When Algorithms Control the World.” 23 August 2011.
http://www.bbc.co.uk/news/technology-14306146.
Watz, Marius. “Computer-Generated Writing.” January 1997.
http://www.xbeat.net/ware/poezak/ComputerGeneratedWriting.htm.
Weizenbaum, Joseph. Computer Power and Human Reason: From Judgment to
Calculation. San Francisco: W.H. Freeman 1976.
Whitby, Blay. “Why the Turing Test Is ai’s Biggest Blind Alley.” 1997.
http://www.informatics.sussex.ac.uk/users/blayw/tt.html.
 Bibliography 223

Wiener, Norbert. God and Golem Inc. Cambridge, ma: mit Press 1964.
Wigner, Eugene. “The Unreasonable Effectiveness of Mathematics in the
Natural Sciences.” Communications in Pure and Applied Mathematics 13
(1960): 1–14.
Williams, J.H.G., A. Witen, T. Suddendorf, and D.I. Perrett. “Imitation,
Mirror Neurons and Autism.” Neuroscience and Behavioral Review 25
(2001): 287–95
Wilson, Timothy D., Christopher Houston, Kathryn Etling, and Nancy
Brekke. “A New Look at Anchoring Effects: Basic Anchoring and Its
Antecedents.” Journal of Experimental Psychology: General 4 (1996):
387–402.
Wittgenstein, Ludwig. Philosophical Investigations. Oxford: Basil Blackwell
1953.
– Tractatus Logico-Philosophicus. Translated by Brian McGuiness and David
Pears. New York: Routledge.
Yang, A.C.C., C.K. Peng, H.W. Yien, and A.L. Goldberger. “Information
Categorization Approach to Literary Authorship Disputes.” Physica A
329 (2003): 473–83.
This page intentionally left blank
 11

Index

5150 (ibm computer), 36 American Society for Aesthetic

701 (ibm mainframe), 36–7 Plastic Surgery, 149
Analytical Engine (computer), 38
aaron (software), 66–7 Anthropocene era, 186
Aarseth, Espen, 7 Apollo 11, 19, 40
Abbottabad, Pakistan, 172 Apple Utilities, 134
Absalom! Absalom!, 33 Aquinas, Thomas, 61
“Abe’s Saloon! Abe’s Saloon!,” Archaea, 187
33 Are Computers Alive?, 150–1, 155
ace (Artificial Conversational Enti- Argument from Continuity in the
ty software), 109 Nervous System, 103
ace (computer), 52 Argument from Informality of
Ada, Augusta. See Lady Lovelace Behaviour, 103
Adam (biblical figure), 149 Argument from Other Minds. See
Age of Intelligent Machines, 35 Argument from the Absence of
Afghanistan, 84 Consciousness
Africa, 126–7, 173 Argument from the Absence of
aids (Acquired Immune Deficien- Consciousness, 101
cy Syndrome), 178 Aristotle, 102
Aitken, Howard, 22 Armstrong, Louis, 149
Alice in Wonderland, 68 Arpanet, 177
Allen, Woody, 18 Artificial Intelligence (ai), 5, 11,
AlphaDog (robot), 84 32, 38, 47, 50, 82–3, 87, 94,
a.m. (software), 47 100, 102, 109, 113–14, 117,
Amazon (retailer), 27–8, 50, 203 131, 135, 144–5, 147, 150, 155,
Amazon River, 121–2 162, 164–5, 176
Amdahl’s Law, 55 Asimov, Isaac, 12
226 Index

Association of American University Blackwood’s Magazine, 32

Presses, 65 Blade Runner, 82
As You Like It, 62 Bletchley Park, 10
Athena, 163 Block, Ned, 111
Atomic Hockey, 158–9 Bloomsday, 20
Autonomous Learning Agents for Bluebeard (novel), 64
Decentralized Data and Infor- Bolyai, János, 189
mation Networks (alladin), Bonaparte, Napoleon, 58, 190–1
181 Bond, James (literary/film charac-
Ayer, Alfred, 94 ter), 171
Boolean search terms, 200
Babbage, Charles, 38 Borges, Jorge Luis, 199
Babbage, Henry, 38 Bradbury, Ray, 32
Bach, Johann Sebastian, 8, 18, 135 Brahms, Johannes, 39
Bad Hemingway Contest. See Inter- Broca, Paul, 147
national Imitation Hemingway Brooklyn Museum, 39
Competition Brooks, Rodney, 131, 162–3
bae Systems (military contractor), Brothers Grimm, 58
181 Brown, Dan, 41
Bagabone, Hem ’I Die Now, 64–5 Brown, Gordon, 108
Bangladesh, 200 Busa, Roberto, 60–2
Baudrillard, Jean, 26 Bush, George W. (Dubya), 114,
Bayesian inferences, 195 171, 175
bbc (British Broadcasting Corpo- Byron, George Gordon (Lord), 38
ration), 99, 124
Beethoven, Ludwig van, 39 Caesar’s Palace (casino), 21
Belgrade, 171 Canterbury Tales, 25, 63
Benedict XVI, pope, 61 captcha test, 104
Bergson, Henri, 127 Cardiopulmonary resuscitation.
Berrill, Norman John, 142 See cpr
Between Literature and Science, 7 Carnap, Rudolf, 193
Beyond Human: Living with Robots Carroll, Lewis, 28–9, 103
and Cyborgs, 151 Castro, Fidel, 171
Big Bang, 113 Central Intelligence Agency. See
bin Laden, Osama, 172 cia
Biology in Action, 142 Central Processing Unit (cpu),
biomorphs. See zoobots 112
Black Studies, 55 Cerebus (robot), 105
Black Widow High-Radix Folded cern (European Organization for
Clos Networks, 56 Nuclear Research), 129
 Index 227

cgw. See computer-generated cpr (cardiopulmonary resuscita-

writing tion), 143
Chamberlain, Bill, 28 Cracow, 64
Champollion, Jean-François, Cray (supercomputer), 22
190–1 Creative Computer, 57
Charabia (website), 27 Critique of Pure Reason, 188–9
Charles, Ray, 50 Cybertext, 7
Chaucer, Geoffrey, 63 cyc (software), 114
Chernobyl, 178
Chinese Room scenario, 10, Dada, 24
112–19, 193 Daisy Miller, 25
Choose Your Own Adventure, 28 Dante, Alighieri 80
Chopin, Frederic, 39 darpa (Defense Advanced
cia (Central Intelligence Agency), Research Projects Agency),
171–2 83–4, 166, 182, 200
Cicekli, Ilyas, 98 Darwin, Charles, 10, 192
Clancy, Tom, 172 Davidson, Donald, 128
Cogito Paradox, 167 da Vinci, Leonardo, 80
Cohen, Gillian, 107 Deacon, Terrence, 117–18
Cohen, Paul, 95 Deep Blue (software), 164–5
Colby, Kenneth Mark, 97 Deep Junior (software), 164
Cold War, 166, 171, 177 de la Mettrie, Julien Offray, 81–2
Columbus, Christopher, 17 DelFly (robot), 162
Complete and Systematic Concord- Delta Force (commandos), 172
ance of the Works of Shakespeare, Denisovans, 186
61 Dennett, Daniel, 15, 46, 107
Computer Power and Human Reason, Deoxyribonucleic acid. See dna
57, 135 Department V of the First Chief
computer-generated writing Directorate. See kgb
(cgw), 24 Derrida, Jacques, 20, 78
“Computing Machinery and Intel- Dershowitz, Alan, 196
ligence,” 23, 94, 99, 124 Der Spiegel, 156
Confederate Effect, 167 Descartes, René, 82, 91, 127–8,
Connection Machine (computer), 163
160 Desert Storm operation, 171
Consciousness Explained, 46 De Veritas et Mendacio, 80
Conway, John, 142–3 Devotchka. See Girl, The
Cope, David, 38 Dicty (Dictyostelium discoideum,
Copeland, Jack B., 120 amoeba), 142
Costco (retailer), 40 Difference Engine (computer), 38
228 Index

Disney, Walt, 58 Euclid, 188–9

dna (deoxyribonucleic acid), 154, Euripides, 28
162, 179, 182, 184, 187 eurisko (software), 48
Dolus (mythological character), European Union, 200
81 Evolution and New Forms of Life, 150
Doolittle, Eliza (literary charac- Experiments in Musical Intelli-
ter), 132 gence (emi, software), 38
Dostoyevsky, Fyodor, 12, 37 Extensible Markup Language. See
Douglas, Frederick, 54 xml
Dow Jones index, 204 ez-Writer (software), 32–3, 40–1,
Dreyfus, Hubert L., 50 64, 75
Drones. See Unmanned Aerial Sys-
tems Fact and Method, 44
Duchamp, Marcel, 30 Falstaff, John (literary character),
Dumas, Alexandre père, 109 63
“Fast Thinking,” 106
Eckert, J. Presper, 3–4 Faulkner, William, 68
Eco, Umberto, 29 Faust (literary character), 80
Edelman, Gerald, 144 Federalist Papers, 63
Eden, Garden of, 80 Feigenbaum, Edward Albert, 131,
edsac (computer), 60 135
Einstein, Albert, 8 Feigenbaum test, 131, 109–10
Electronic Literature, 7, 150 Ferranti’s Mark 1, 22–3
Electronic Numerical Integrator fide (Fédération Internationale
and Computer. See eniac des Échecs), 164
eliza (software), 132, 134–5 Fifth Generation program (in
Elizabethan era, 61 Japan), 164
Elsässer, Walter M., 127 Finnegans Wake, 20, 73
emi (software), 64 First World War, 173
Endangered Software Archive Flinck, Govaert, 33
(website), 29 Ford-class supercarrier, 173
eniac (Electronic Numerical Inte- Foucault’s Pendulum, 29
grator and Computer), 3–4 Franke, Herbert, 69–70
Epstein, Robert, 109 Frankenstein, Victor, 79
Erlang (programming language), French, Robert, 123, 125–6
161 From Lowbrow to Nobrow, 41
Escher, Maurits Cornelis, 18, 135 Front, The, 18
esp cards, 103 Fujitsu Fanuc (robots manufactur-
Essential Turing, 120 er), 151
Etter, Thomas, 28 Fujitsu K (supercomputer), 22
 Index 229

Fukushima earthquake, 178 Hodges, Andrew, 123

Furbies (toy), 131 Hofstadter, Douglas R., 15, 18, 38,
Future Shock, 8 119, 135, 163
Holbein, Hans, 63–4
Gaddafi, Muammar, 171 Holland, John, 161
Gardner, Howard, 11 Hollywood, 156
Garrison, Jim, 43 Homer, 20, 28, 70, 81
Gates, Bill, 57 Homo sapiens, 46, 55, 135, 139,
Gibson, William, 9, 200 186, 188
“Gigamesh,” 20, 71 Horace, 69
Girl, The (Devotchka), 36, 76 “How to Write a Blackwood Arti-
Gödel, Kurt, 18, 135, 193–4 cle,” 31–2
Goldberg, Rube, 191 huac. See Un-American Activities
Golem, 81–2 Committee
Gomez, Juan-Carlos, 130 Hughes, G.H., 65
Google, 24–5, 60, 65, 198–200, Human Genome Sciences, 149
204 Hume, David, 83
Greenpeace, 41 Hussein, Saddam, 171
Grice, Paul, 138 hyxos (software), 37
Gutenberg, Johannes, 65
I Am a Strange Loop, 119
Hamilton, Alexander, 63 ibm (International Business
Hamlet, 62, 66, 68 Machines Corporation), 22,
Hannahan, Patrick (literary char- 36–37, 104, 109–10, 160–3,
acter), 20 165
Harlequin romances, 76 I Ching, 21
Hawthorne, Nathaniel, 72 “If Not Turing’s Test, Then
Hayles, Katherine, 7 What?,” 95
Hellfire missiles, 172 Iliad, 81
Hemingway, Ernest, 25 iMac (computer), 21
Henrich, Joseph, 140 Imaginary Magnitude, 17
Hephaestus (Vulcan), 81 imitation game, 95
Higgins, Henry (literary charac- “Intelligent Machinery,” 95
ter), 132 “Intelligent Machinery, A Hereti-
high-frequency trading (hft), 58 cal Theory,” 102
Hilbert, David, 193 International Imitation Heming-
Hindenburg (airship), 193 way Competition, 32
Hindu religion, 163 International Technology
“History of Bitic Literature,” 17, Roadmap for Semiconductors,
36, 85 158
230 Index

Internet, 61, 65, 177, 197–200 Kurzweil, Ray, 35

Iraq, 84, 172
Ireland, 20 La Chanson de Roland, 53–4, 139
Lady Lovelace, 38–40, 103
jabberwacky (software), 29 Lady Lovelace’s Objection, 38–40,
“Jabberwocky,” 29 103
Jacquette, Dale, 100, 123 Land Warrior System, 84
Jagiellonian University, 64–5 Lang, Fritz, 92
James, Henry, 25 Language, Truth and Logic, 94
Japan, 22, 36, 82, 151, 155, Las Vegas, 45
158–9, 164 Lawrence, D.H., 64
Japan Ministry of International Lem, Stanislaw, 8, 15, 17–20, 33,
Trade and Industry, 164 35–6, 39, 53–4, 60, 64, 71–3,
Jefferson, Geoffrey, 23, 101 76, 117, 139, 181
Jeopardy, 109–10, 131 Lenat, Douglas, 48–49, 113–14
Jesus, 189 Levy, David, 86
JFK. See Kennedy, John F. L’Homme Machine, 82
Jigsaw Room scenario, 117 Library of Congress, 20
Jobs, Steve, 132 Life on the Screen: Identity in the Age
Johnston, Rory, 57–8 of the Internet, 151
Joplin, Scott, 38 Literature, Analytically Speaking, 6,
Journal of the Association for Comput- 55
ing Machinery, 165 Little, Brown, 19–20, 40
Joyce, James, 20, 64 Little Endian mode, 55
Judeo-Christian, 80, 163 Lobachevsky, Nikolai, 189
Loebner, Hugh, 109
Kant, Immanuel, 188–9, 192 Loebner Prize, 109, 167–8
Kasparov, Garry, 33, 163–5 Loew, Rabbi, 81
Kauffman, Stuart, 120, 122 London Stock Exchange, 181
Keller, Helen, 54 Los Angeles Book Review, 68
Kellogg’s, 125 Love and Sex with Robots, 86
Kennedy, John F. (JFK), 43 Lucas, John Randolph, 88, 102
kgb, 171 Lycan, William G., 85
Khodorkovsky, Mikhail, 85
King Henry IV, Part 2, 63 m1a2 Abrams, 173–4
King Henry V, 62 Madison, James, 63
Kismet (robot), 130–1 Maginot Line, 39
Knight Capital (trading company), magnetic resonance imaging
204 (mri), 146
Koko (primate), 163 Mahler, Gustav, 39
 Index 231

Mailer, Norman, 19, 40 “Minds, Machines and Gödel,” 88

Making of a Fly, 203 Minsky, Marvin, 1
Man vs. Machine World Champi- Mira (computer), 162
onship (chess), 164 Miss Lonelyhearts, 134
Manchester Mark 1, 22–3 mit. See Massachusetts Institute of
Mandela, Nelson, 85 Technology
Man in the Iron Mask, 109 mit Museum, 131
Maria (film character), 92 Möbius strip, 42
Mark 1, 22 Monet, Claude-Oscar, 64
Markov, Georgi, 172 Moore’s Law, 58, 160
Marlowe, Christopher, 80 Moravec, Hans, 35
Massachusetts Institute of Technol- Mount Etna, 81, 117
ogy (mit), 130–1, 148, 163 Mount St Helens, 178
Massive Compact Halo Objects Mozart, Wolfgang Amadeus, 37–8
(machos), 128 Muppet Show, 24
Mathematical Objection, 102 “Murder Makes the Magazine,” 28
Mauchly, John, 3–4 Museum of Modern Art (San
Mbenga (pygmy tribe, Central Francisco), 39
Africa), 7 Musikalisches Würspiel, 22
Mbuti (pygmy tribe, Central Mystery of Consciousness, 112
Africa), 7 “Mystery of the Chinese Room,”
McCarthy, John, 49, 144–5 117
McCulloch, Warren S., 102
“Meaning” (1957 paper by Paul Nagel, Ernest, 102
Grice), 138 Narrative Intelligence, 5
Meditationes de Prima Philosophia, Narrative of the Life of Frederick Dou-
91 glas, 54
“Mellonta Tauta,” 12 nasa (National Aeronautics and
Melpomene (software), 64 Space Administration), 40
Mensa, 166 National Supercomputing Center
Methuselah (biblical character), (China), 22
149 nato (North Atlantic Treaty Orga-
Metropolis (film), 92 nization), 171
Michie, Donald, 57–8 Neanderthals, 186
Milic, Louis, 27 Netflix, 50
Millar, P.H., 123 Neuromancer, 9, 200
Miller, Richard, 44 neuromophs. See zoobots
Milosevic, Slobodan, 171 Newall, Allen, 119
Mind Children, 35 Newman, James R., 102
“Minds, Brains, and Programs,” 111 New York, 156
232 Index

New York Stock Exchange, 204 Perfect Vacuum, A, 20

New York Times Biterary Review, 69 Phaedrus, 80
New York Times Book Review, 68 “Philosophy of Composition,” 31
Nim (chimpanzee), 163 Pixar, 132
Nimitz-class supercarrier, 173 Plato, 189, 194
Nintendo, 134 Poe, Edgar Allan, 12, 28, 31–2,
Nippon Electric Company (nec), 72
132 Poland, 65
Nobel Prize, 25, 84, 144, 183 Polanyi, Michael, 127
Northwestern University, 105 Policeman’s Beard Is Half Construct-
Nostradamus, 150 ed, The, 29–30
Pollock, Jackson, 64
Obama, Barack, 175 Prager, John, 124
Objection from Disability, 101–2 “Predicament,” 32
Objection from Extra-Sensory Predator (drone), 172, 176
Perception, 103 Prince, Howard (film character),
Odets, Clifford, 68 18–20, 68
Odyssey, 20 Princeton’s Institute for Advanced
Of a Fire on the Moon, 19 Studies, 8
Of Literature and Knowledge, 6 “Programs with Common Sense,”
Ohm, Georg Simon, 123 49
“Old Man and the Flea,” 33 Prometheus, 80
Old Man and the Sea, 33 Propositional Jukebox, 111
“On Computable Numbers,” Protocol, 18
17 Pseudodostoyevsky (fictional com-
On Turing, 206n6 puter), 36
Othello, 62 Psychology of Cognition, 107
Ptolemy, 129
p-51 Mustang (fighter-bomber Pulitzer prize, 17, 150
escort), 173 Putin, Vladimir, 172
Page, Larry, 199
Paignton Zoo, 12, 109 Queer Studies, 55
PaPeRo (robot), 131–2, 135 Qur’an, 102
parry (software), 97
Pauli, Wolfgang, 123 Rachmaninoff, Sergei, 39
pc (tablet computer), 84 Racter (software), 28–31, 64, 78
Peat, David, 47 Raleigh, John (literary character),
Penrose, Roger, 102 37
Pentagon, 58, 83, 166, 172, 175, Rating Games, 125
177, 204 Raven (drone), 175–6
 Index 233

Ray, Tom, 72 Severe Acute Respiratory Syn-

Reagan, Ronald, 171, 173 drome (sars), 178
Reality Principle, 144 sexpert (software), 157
Reaper (drone), 176 Shakespeare, William, 12, 25, 56,
Renoir, Pierre-Auguste, 64 61–2, 109
Resource Description Framework Shakur, Tupac (2Pac Shakur), 25
(rdf), 200 Shannon, Claude, 11
Reverse Turing test (rtt), 104–5, Shaw, George Bernard, 132
144 Shinto religion, 82
Rhine, J.B., 103 Shurkin, Joel, 150, 207n1
“Rhyme of the nasa Mariner,” Simon, Herbert, 119, 151–4
33 Simons, Geoff, 150
Ritt, Martin, 18 Simpson, O.J., 196
Riverside Shakespeare, 61 Simpsons: Bart, Homer, Lisa, and
RoboCup, 36, 158–9 Marge (animated sitcom charac-
Robo Ping Pong, 158–9 ters), 140
Robot Sumo, 158–9 “Simulacra and Simulation,” 26
Rogerian psychotherapy, 132 Singapore, 58,
Rosencrantz and Guildenstern Are “Software for the Soul,” 156
Dead, 195 Solidarity (trade union), 65
Rosetta stone, 190–1 Sorensen, Alan, 68
Rothschild, Nathan, 58 Sotheby’s, 64
rtt. See Reverse Turing test Spielberg, Steven, 82
Spitfire (fighter), 173
Santa Ana, 178 Star Diaries, 181
Satanic Verses, 85 Standard Turing test (stt). See
Savile Row, 138 Turing test
Saygin, Ayse, 7, 98 Stealth Bomber, 171, 173
sc10 (computing conference), Stedelijk Museum, Amsterdam, 39
161 Story of My Life, 54
Schaffer, Simon, 86 Strachey, Chris, 22
Schmidt, Eric, 204 stt (Standard Turing test). See
Schwarzenegger, Arnold, 12 Turing test
Scotland, 173 Sun Also Rises, The, 25
Searle, John, 101, 111–17, Surrey Comet, 99
119–20, 167, 193 Swedish Academy, 85
Second World War, 10, 20, 22, Swirski, Peter, 92–3
108, 173–4 Switchblade (drone), 176
Semantic Web, 200, 203 Symbolic Logic, 103
Sequoia (computer), 22 Symbolic Species, The, 117
234 Index

Tate Modern (London), 39 Twice Told Tales, 72

Tempest, The, 62 Twilight Zone, 33, 76
Terminator 3: Rise of the Machines, Twitter, 61
12
Tesler, Larry, 100 Ultimatest, 100, 106–7, 168
Three Musketeers, The, 109 Ultra Hal (software), 87
Tesler, Larry, 100 Ulysses, 20, 80
Theory of Mind (tom), 46, 91, Un-American Activities Committee
120, 126, 134–7, 139–40 (huac), 18
Tianhe-1a (computer), 22, 39 Uniform Determination of Death
Tierra (software), 72 Act, 143
Tierra Working Group, 161 United Nations, 126–7
Tiger tank, 173–174 United States, 20, 43, 58, 83, 143,
Time magazine, 10 155, 172, 189
Titan (computer), 22, 39, 80 univac (Universal Automatic
Titanic, 39 Computer), 22–3
Toffler, Alvin, 8 Unmanned Aerial Systems
Tokyo, 58, 132 (drones), 175
tom. See Theory of Mind “Unmanned Aircraft Systems
Tomasello, Michael, 136 Flight Plan 2009–2047,” 176
Toyota, 164 “Unreasonable Effectiveness of
Tractatus Logico-Philosophicus, 193 Mathematics in the Natural Sci-
Trotsky, Leon, 172 ences,” 183
“Troubles with Functionalism,” Updike, John, 27–8
111 US Air Force, 173, 176
tt. See Turing test US Army, 173–5
Turing, Alan, 3, 9, 17, 22–3, 52, US Department of Defense, 172,
65, 94–6, 98–9, 101–3, 107–9, 204
118–20, 123–5, 165 US War Department. See US
Turing Archive for the History of Department of Defense
Computing, 120
Turing Award, 9 Van Gogh, Vincent, 64
Turing Police, 9, 200, 204 van Rijn, Rembrandt, 33
Turing test (tt), 7, 10, 35, Venus flytraps, 142
94–101, 103–4, 106–7, 109–10, Vickers fb5 (Fighting Biplane),
111–14, 117–18, 123–6, 128, 173
131, 138–9, 141, 150–1, 157, Vienna Circle (Wiener Kreis),
164–7, 199; “fly-on-the-wall” 192–3
Turing test, 106 Virtual Reality (vr) (software),
Turing Year, 10 157
 Index 235

von Neumann, John, 9, 158 Wife of Bath, 63

Vonnegut, Kurt, 64 Wigner, Eugene, 183
Voynich manuscript, 76–7 Wikipedia, 61
Wilde, Oscar, 135
Wadsworth Philosophers Series, Windows (software), 57
124 Wittgenstein, Ludwig, 26–7, 191,
Walking Machine decathlon, 193
158–9 Women’s Studies, 55
Wall Street, 58–60 Wonder, Stevie, 50
Wal-Mart, 40 Word (software), 21
Wambaugh, Joseph, 32 World I Live In, The, 54
Wandering Jew, 187 World War I. See First World War
Warren Report, 43 World War II. See Second World
Washoe (chimpanzee), 163 War
Wason test, 140 World Wide Web. See Internet
Waterloo, Battle of, 58 WWW Consortium, 200
Watson (software), 109–10,
126–7 x3d Fritz (software), 164
waw hazard, 56 xml (Extensible Markup Lan-
Weakly Interacting Massive Parti- guage), 200
cles (wimps), 128
Weizenbaum, Joseph, 57, 132, Yao, Ming, 81
134–5 YouTube, 66, 198
Wernicke, Carl, 78, 147
West, Nathanael, 134 Zener, Karl, 103
“What a Wonderful World,” 149 Zenobia, Signora Psyche (literary
What Computers Can’t Do, 50 character), 32
What Computers Still Can’t Do, 50 Zoobots, 162