www.nytimes.com /2024/09/10/books/review/nexus-yuval-noah-harari.
html
Book Review: ‘Nexus,’ by Yuval Noah
Harari
Dennis Duncan ⋮ 7-8 minutes ⋮ 9/10/2024
Advertisement
SKIP ADVERTISEMENT
nonfiction
Pulling Back the Silicon Curtain
Yuval Noah Harari’s study of human communication may be
anything but brief, but if you can make it to the second half, you’ll
be both entertained and scared.
Yuval Noah Harari sounds the alarm on our A.I. future.
“When the tech giants set their hearts on designing better
algorithms, they can usually do it,” he writes. But will they?
Credit...Philip Cheung for The New York Times
Dennis Duncan
Dennis Duncan is the author of “Index, A History of the.”
Sept. 10, 2024
When you purchase an independently reviewed book through our
site, we earn an affiliate commission.
NEXUS: A Brief History of Information Networks From the
Stone Age to AI, by Yuval Noah Harari
In the summer of 2022, a software engineer named Blake Lemoine
was fired by Google after an interview with The Washington Post in
which he claimed that LaMDA, the chatbot he had been working
on, had achieved sentience.
A few months later, in March 2023, an open letter from the Future
of Life Institute, signed by hundreds of technology leaders including
Steve Wozniak and Elon Musk, called on A.I. labs to pause their
research. Artificial intelligence, it claimed, posed “profound risks to
society and humanity.”
The following month, Geoffrey Hinton, the “godfather of A.I.,” quit
his post at Google, telling this newspaper that he regretted his life’s
work. “It is hard to see how you can prevent the bad actors from
using it for bad things,” he warned.
Over the last few years we have become accustomed to hare-eyed
messengers returning from A.I.’s frontiers with apocalyptic
warnings. And yet, real action in the form of hard regulation has
been little in evidence. Last year’s executive order on A.I. was, as
one commentator put it, “directional and aspirational” — a shrewdly
damning piece of faint praise.
Image
Meanwhile, stock prices for the tech sector continue to soar while
the industry mutters familiar platitudes: The benefits outweigh the
risks; the genie is already out of the bottle; if we don’t do it, our
enemies will.
Yuval Noah Harari has no time for these excuses. In 2011, he
published “Sapiens,” an elegant and sometimes profound history of
our species. It was a phenomenon, selling over 25 million copies
worldwide. Harari followed it up by turning his gaze forward with
“Homo Deus,” in which he considered our future. At this point,
Harari, an academic historian, became saddled with a new
professional identity and a new circle of influence: A.I. expert,
invited into the rarefied echelons of “scientists, entrepreneurs and
world leaders.” “Nexus,” in essence, is Harari’s report from this
world.
First, it must be said that the subtitle — “A Brief History of
Information Networks From the Stone Age to A.I.” — is misleading.
Really, what we have is two separate books, neither brief. The first
200 pages are indeed historical in their way. Unfortunately, this is a
dizzying, all-in version of history that swerves unsatisfyingly from
Assyrian clay tablets to a 19th-century cholera outbreak to an
adaptation of the “Ramayana” on Indian TV to the Peasants’ Revolt
in medieval England to the Holocaust in Romania, and so on. It
doesn’t feel controlled, or even particularly expert — and the effect
is a little like a flight where the person sitting next to you is well-
read, hyper-caffeinated and determined to tell you his Theory of
Everything.
In a nutshell, Harari’s thesis is that the difference between
democracies and dictatorships lies in how they handle information.
Dictatorships are more concerned with controlling data than with
testing its truth value; democracies, by contrast, are transparent
information networks in which citizens are able to evaluate and, if
necessary, correct bad data.
All of this is sort of obvious-interesting, while also being too vague
— too open to objection and counterexample — to constitute a
useful theory of information. After a lot of time, we have arrived at a
loose proof of what we hopefully felt already: Systems that are self-
correcting — because they promote conversation and mutuality —
are preferable to those that offer only blind, disenfranchised
subservience.
In the end, however, this doesn’t really matter, because the second
half of the book is where the action is. The meat of “Nexus” is
essentially an extended policy brief on A.I.: What are its risks, and
what can be done? (We don’t hear much about the potential
benefits because, as Harari points out, “the entrepreneurs leading
the A.I. revolution already bombard the public with enough rosy
predictions about them.”) It has taken too long to get here, but once
we arrive Harari offers a useful, well-informed primer.
The threats A.I. poses are not the ones that filmmakers visualize:
Kubrick’s HAL trapping us in the airlock; a fascist RoboCop
marching down the sidewalk. They are more insidious, harder to
see coming, but potentially existential. They include the
catastrophic polarizing of discourse when social media algorithms
designed to monopolize our attention feed us extreme, hateful
material. Or the outsourcing of human judgment — legal, financial
or military decision-making — to an A.I. whose complexity becomes
impenetrable to our own understanding.
Echoing Churchill, Harari warns of a “Silicon Curtain” descending
between us and the algorithms we have created, shutting us out of
our own conversations — how we want to act, or interact, or govern
ourselves.
None of these scenarios, however, is a given. Harari points to the
problem of email spam, which used to clog up our inboxes and
waste millions of hours of productivity every day. And then,
suddenly, it didn’t. In 2015, Google was able to claim that its Gmail
algorithm had a 99.9 percent success rate in blocking genuine
spam. “When the tech giants set their hearts on designing better
algorithms,” writes Harari, “they can usually do it.”
Even in its second half, not all of “Nexus” feels original. If you pay
attention to the news, you will recognize some of the stories Harari
tells. But, at its best, his book summarizes the current state of
affairs with a memorable clarity.
Parts of “Nexus” are wise and bold. They remind us that democratic
societies still have the facilities to prevent A.I.’s most dangerous
excesses, and that it must not be left to tech companies and their
billionaire owners to regulate themselves.
That may just sound like common sense, but it is valuable when
said by a global intellectual with Harari’s reach. It is only frustrating
that he could not have done so more concisely.
NEXUS: A Brief History of Information Networks from the
Stone Age to AI | By Yuval Noah Harari | Random House | 518 pp.
| $35
Advertisement
SKIP ADVERTISEMENT
