How Technology Is Destroying Jobs

Economic theory and government policy will have to be rethought if technology is

indeed destroying jobs faster than it is creating new ones.

Given his calm and reasoned academic demeanor, it is easy to miss just how
provocative Erik Brynjolfsson’s contention really is. Brynjolfsson, a professor at
the MIT Sloan School of Management, and his collaborator and coauthor Andrew
McAfee have been arguing for the last year and a half that impressive advances in
computer technology—from improved industrial robotics to automated translation
services—are largely behind the sluggish employment growth of the last 10 to 15
years. Even more ominous for workers, the MIT academics foresee dismal
prospects for many types of jobs as these powerful new technologies are
increasingly adopted not only in manufacturing, clerical, and retail work but in
professions such as law, financial services, education, and medicine.

That robots, automation, and software can replace people might seem obvious to
anyone who’s worked in automotive manufacturing or as a travel agent. But
Brynjolfsson and McAfee’s claim is more troubling and controversial. They believe
that rapid technological change has been destroying jobs faster than it is creating
them, contributing to the stagnation of median income and the growth of
inequality in the United States. And, they suspect, something similar is happening
in other technologically advanced countries.

Perhaps the most damning piece of evidence, according to Brynjolfsson, is a chart

that only an economist could love. In economics, productivity—the amount of
economic value created for a given unit of input, such as an hour of labor—is a
crucial indicator of growth and wealth creation. It is a measure of progress. On the
chart Brynjolfsson likes to show, separate lines represent productivity and total
employment in the United States. For years after World War II, the two lines
closely tracked each other, with increases in jobs corresponding to increases in
productivity. The pattern is clear: as businesses generated more value from their
workers, the country as a whole became richer, which fueled more economic
activity and created even more jobs. Then, beginning in 2000, the lines diverge;
productivity continues to rise robustly, but employment suddenly wilts. By 2011, a
significant gap appears between the two lines, showing economic growth with no
parallel increase in job creation. Brynjolfsson and McAfee call it the “great
decoupling.” And Brynjolfsson says he is confident that technology is behind both
the healthy growth in productivity and the weak growth in jobs.
It’s a startling assertion because it threatens the faith that many economists place in
technological progress. Brynjolfsson and McAfee still believe that technology
boosts productivity and makes societies wealthier, but they think that it can also
have a dark side: technological progress is eliminating the need for many types of
jobs and leaving the typical worker worse off than before. Brynjolfsson can point
to a second chart indicating that median income is failing to rise even as the gross
domestic product soars. “It’s the great paradox of our era,” he says. “Productivity
is at record levels, innovation has never been faster, and yet at the same time, we
have a falling median income and we have fewer jobs. People are falling behind
because technology is advancing so fast and our skills and organizations aren’t
keeping up.”

Brynjolfsson and McAfee are not Luddites. Indeed, they are sometimes accused of
being too optimistic about the extent and speed of recent digital advances.
Brynjolfsson says they began writing Race Against the Machine, the 2011 book in
which they laid out much of their argument, because they wanted to explain the
economic benefits of these new technologies (Brynjolfsson spent much of the
1990s sniffing out evidence that information technology was boosting rates of
productivity). But it became clear to them that the same technologies making many
jobs safer, easier, and more productive were also reducing the demand for many
types of human workers.

Anecdotal evidence that digital technologies threaten jobs is, of course, everywhere.
Robots and advanced automation have been common in many types of
manufacturing for decades. In the United States and China, the world’s
manufacturing powerhouses, fewer people work in manufacturing today than in
1997, thanks at least in part to automation. Modern automotive plants, many of
which were transformed by industrial robotics in the 1980s, routinely use machines
that autonomously weld and paint body parts—tasks that were once handled by
humans. Most recently, industrial robots like Rethink Robotics’ Baxter (see “The
Blue-Collar Robot,” May/June 2013), more flexible and far cheaper than their
predecessors, have been introduced to perform simple jobs for small manufacturers
in a variety of sectors. The website of a Silicon Valley startup called Industrial
Perception features a video of the robot it has designed for use in warehouses
picking up and throwing boxes like a bored elephant. And such sensations as
Google’s driverless car suggest what automation might be able to accomplish
someday soon.
A less dramatic change, but one with a potentially far larger impact on
employment, is taking place in clerical work and professional services.
Technologies like the Web, artificial intelligence, big data, and improved analytics—
all made possible by the ever increasing availability of cheap computing power and
storage capacity—are automating many routine tasks. Countless traditional white-
collar jobs, such as many in the post office and in customer service, have
disappeared. W. Brian Arthur, a visiting researcher at the Xerox Palo Alto Research
Center’s intelligence systems lab and a former economics professor at Stanford
University, calls it the “autonomous economy.” It’s far more subtle than the idea of
robots and automation doing human jobs, he says: it involves “digital processes
talking to other digital processes and creating new processes,” enabling us to do
many things with fewer people and making yet other human jobs obsolete.

It is this onslaught of digital processes, says Arthur, that primarily explains how
productivity has grown without a significant increase in human labor. And, he says,
“digital versions of human intelligence” are increasingly replacing even those jobs
once thought to require people. “It will change every profession in ways we have
barely seen yet,” he warns.

McAfee, associate director of the MIT Center for Digital Business at the Sloan
School of Management, speaks rapidly and with a certain awe as he describes
advances such as Google’s driverless car. Still, despite his obvious enthusiasm for
the technologies, he doesn’t see the recently vanished jobs coming back. The
pressure on employment and the resulting inequality will only get worse, he
suggests, as digital technologies—fueled with “enough computing power, data, and
geeks”—continue their exponential advances over the next several decades. “I
would like to be wrong,” he says, “but when all these science-fiction technologies
are deployed, what will we need all the people for?”

New Economy?

But are these new technologies really responsible for a decade of lackluster job
growth? Many labor economists say the data are, at best, far from conclusive.
Several other plausible explanations, including events related to global trade and the
financial crises of the early and late 2000s, could account for the relative slowness
of job creation since the turn of the century. “No one really knows,” says Richard
Freeman, a labor economist at Harvard University. That’s because it’s very difficult
to “extricate” the effects of technology from other macroeconomic effects, he says.
But he’s skeptical that technology would change a wide range of business sectors
fast enough to explain recent job numbers.
Employment trends have polarized the workforce and hollowed out the middle
class.

David Autor, an economist at MIT who has extensively studied the connections
between jobs and technology, also doubts that technology could account for such
an abrupt change in total employment. “There was a great sag in employment
beginning in 2000. Something did change,” he says. “But no one knows the cause.”
Moreover, he doubts that productivity has, in fact, risen robustly in the United
States in the past decade (economists can disagree about that statistic because there
are different ways of measuring and weighing economic inputs and outputs). If he’s
right, it raises the possibility that poor job growth could be simply a result of a
sluggish economy. The sudden slowdown in job creation “is a big puzzle,” he says,
“but there’s not a lot of evidence it’s linked to computers.”

To be sure, Autor says, computer technologies are changing the types of jobs
available, and those changes “are not always for the good.” At least since the 1980s,
he says, computers have increasingly taken over such tasks as bookkeeping, clerical
work, and repetitive production jobs in manufacturing—all of which typically
provided middle-class pay. At the same time, higher-paying jobs requiring creativity
and problem-solving skills, often aided by computers, have proliferated. So have
low-skill jobs: demand has increased for restaurant workers, janitors, home health
aides, and others doing service work that is nearly impossible to automate. The
result, says Autor, has been a “polarization” of the workforce and a “hollowing
out” of the middle class—something that has been happening in numerous
industrialized countries for the last several decades. But “that is very different from
saying technology is affecting the total number of jobs,” he adds. “Jobs can change
a lot without there being huge changes in employment rates.”

What’s more, even if today’s digital technologies are holding down job creation,
history suggests that it is most likely a temporary, albeit painful, shock; as workers
adjust their skills and entrepreneurs create opportunities based on the new
technologies, the number of jobs will rebound. That, at least, has always been the
pattern. The question, then, is whether today’s computing technologies will be
different, creating long-term involuntary unemployment.

At least since the Industrial Revolution began in the 1700s, improvements in

technology have changed the nature of work and destroyed some types of jobs in
the process. In 1900, 41 percent of Americans worked in agriculture; by 2000, it
was only 2 percent. Likewise, the proportion of Americans employed in
manufacturing has dropped from 30 percent in the post–World War II years to
around 10 percent today—partly because of increasing automation, especially
during the 1980s.

hile such changes can be painful for workers whose skills no longer match the
needs of employers, Lawrence Katz, a Harvard economist, says that no historical
pattern shows these shifts leading to a net decrease in jobs over an extended
period. Katz has done extensive research on how technological advances have
affected jobs over the last few centuries—describing, for example, how highly
skilled artisans in the mid-19th century were displaced by lower-skilled workers in
factories. While it can take decades for workers to acquire the expertise needed for
new types of employment, he says, “we never have run out of jobs. There is no
long-term trend of eliminating work for people. Over the long term, employment
rates are fairly stable. People have always been able to create new jobs. People
come up with new things to do.”

Still, Katz doesn’t dismiss the notion that there is something different about today’s
digital technologies—something that could affect an even broader range of work.
The question, he says, is whether economic history will serve as a useful guide. Will
the job disruptions caused by technology be temporary as the workforce adapts, or
will we see a science-fiction scenario in which automated processes and robots with
superhuman skills take over a broad swath of human tasks? Though Katz expects
the historical pattern to hold, it is “genuinely a question,” he says. “If technology
disrupts enough, who knows what will happen?”

Dr. Watson

To get some insight into Katz’s question, it is worth looking at how today’s most
advanced technologies are being deployed in industry. Though these technologies
have undoubtedly taken over some human jobs, finding evidence of workers being
displaced by machines on a large scale is not all that easy. One reason it is difficult
to pinpoint the net impact on jobs is that automation is often used to make human
workers more efficient, not necessarily to replace them. Rising productivity means
businesses can do the same work with fewer employees, but it can also enable the
businesses to expand production with their existing workers, and even to enter new
markets.

Take the bright-orange Kiva robot, a boon to fledgling e-commerce companies.

Created and sold by Kiva Systems, a startup that was founded in 2002 and bought
by Amazon for $775 million in 2012, the robots are designed to scurry across large
warehouses, fetching racks of ordered goods and delivering the products to
humans who package the orders. In Kiva’s large demonstration warehouse and
assembly facility at its headquarters outside Boston, fleets of robots move about
with seemingly endless energy: some newly assembled machines perform tests to
prove they’re ready to be shipped to customers around the world, while others wait
to demonstrate to a visitor how they can almost instantly respond to an electronic
order and bring the desired product to a worker’s station.

A warehouse equipped with Kiva robots can handle up to four times as many
orders as a similar unautomated warehouse, where workers might spend as much as
70 percent of their time walking about to retrieve goods. (Coincidentally or not,
Amazon bought Kiva soon after a press report revealed that workers at one of the
retailer’s giant warehouses often walked more than 10 miles a day.)

Despite the labor-saving potential of the robots, Mick Mountz, Kiva’s founder and
CEO, says he doubts the machines have put many people out of work or will do so
in the future. For one thing, he says, most of Kiva’s customers are e-commerce
retailers, some of them growing so rapidly they can’t hire people fast enough. By
making distribution operations cheaper and more efficient, the robotic technology
has helped many of these retailers survive and even expand. Before founding Kiva,
Mountz worked at Webvan, an online grocery delivery company that was one of
the 1990s dot-com era’s most infamous flameouts. He likes to show the numbers
demonstrating that Webvan was doomed from the start; a $100 order cost the
company $120 to ship. Mountz’s point is clear: something as mundane as the cost
of materials handling can consign a new business to an early death. Automation can
solve that problem.
Meanwhile, Kiva itself is hiring. Orange
balloons—the same color as the robots—
hover over multiple cubicles in its sprawling
office, signaling that the occupants arrived
within the last month. Most of these new
employees are software engineers: while the
robots are the company’s poster boys, its
lesser-known innovations lie in the complex
algorithms that guide the robots’ movements
and determine where in the warehouse
products are stored. These algorithms help
make the system adaptable. It can learn, for
example, that a certain product is seldom
ordered, so it should be stored in a remote
area.

Though advances like these suggest how

some aspects of work could be subject to
automation, they also illustrate that humans
still excel at certain tasks—for example,
packaging various items together. Many of
the traditional problems in robotics—such
as how to teach a machine to recognize an
object as, say, a chair—remain largely
intractable and are especially difficult to
solve when the robots are free to move
about a relatively unstructured environment
like a factory or office.

Techniques using vast amounts of

computational power have gone a long way
toward helping robots understand their
surroundings, but John Leonard, a professor
of engineering at MIT and a member of its
Computer Science and Artificial Intelligence
Laboratory (CSAIL), says many familiar
difficulties remain. “Part of me sees
accelerating progress; the other part of me
sees the same old problems,” he says. “I see how hard it is to do anything with
robots. The big challenge is uncertainty.” In other words, people are still far better
at dealing with changes in their environment and reacting to unexpected events.

For that reason, Leonard says, it is easier to see how robots could work with
humans than on their own in many applications. “People and robots working
together can happen much more quickly than robots simply replacing humans,” he
says. “That’s not going to happen in my lifetime at a massive scale. The
semiautonomous taxi will still have a driver.”

One of the friendlier, more flexible robots meant to work with humans is Rethink’s
Baxter. The creation of Rodney Brooks, the company’s founder, Baxter needs
minimal training to perform simple tasks like picking up objects and moving them
to a box. It’s meant for use in relatively small manufacturing facilities where
conventional industrial robots would cost too much and pose too much danger to
workers. The idea, says Brooks, is to have the robots take care of dull, repetitive
jobs that no one wants to do.

It’s hard not to instantly like Baxter, in part because it seems so eager to please.
The “eyebrows” on its display rise quizzically when it’s puzzled; its arms
submissively and gently retreat when bumped. Asked about the claim that such
advanced industrial robots could eliminate jobs, Brooks answers simply that he
doesn’t see it that way. Robots, he says, can be to factory workers as electric drills
are to construction workers: “It makes them more productive and efficient, but it
doesn’t take jobs.”

The machines created at Kiva and Rethink have been cleverly designed and built to
work with people, taking over the tasks that the humans often don’t want to do or
aren’t especially good at. They are specifically designed to enhance these workers’
productivity. And it’s hard to see how even these increasingly sophisticated robots
will replace humans in most manufacturing and industrial jobs anytime soon. But
clerical and some professional jobs could be more vulnerable. That’s because the
marriage of artificial intelligence and big data is beginning to give machines a more
humanlike ability to reason and to solve many new types of problems.

Even if the economy is only going through a transition, it is an extremely painful

one for many.

In the tony northern suburbs of New York City, IBM Research is pushing super-
smart computing into the realms of such professions as medicine, finance, and
customer service. IBM’s efforts have resulted in Watson, a computer system best
known for beating human champions on the game show Jeopardy! in 2011. That
version of Watson now sits in a corner of a large data center at the research facility
in Yorktown Heights, marked with a glowing plaque commemorating its glory days.
Meanwhile, researchers there are already testing new generations of Watson in
medicine, where the technology could help physicians diagnose diseases like cancer,
evaluate patients, and prescribe treatments.

IBM likes to call it cognitive computing. Essentially, Watson uses artificial--

intelligence techniques, advanced natural-language processing and analytics, and
massive amounts of data drawn from sources specific to a given application (in the
case of health care, that means medical journals, textbooks, and information
collected from the physicians or hospitals using the system). Thanks to these
innovative techniques and huge amounts of computing power, it can quickly come
up with “advice”—for example, the most recent and relevant information to guide
a doctor’s diagnosis and treatment decisions.

Despite the system’s remarkable ability to make sense of all that data, it’s still early
days for Dr. Watson. While it has rudimentary abilities to “learn” from specific
patterns and evaluate different possibilities, it is far from having the type of
judgment and intuition a physician often needs. But IBM has also announced it will
begin selling Watson’s services to customer-support call centers, which rarely
require human judgment that’s quite so sophisticated. IBM says companies will rent
an updated version of Watson for use as a “customer service agent” that responds
to questions from consumers; it has already signed on several banks. Automation is
nothing new in call centers, of course, but Watson’s improved capacity for natural-
language processing and its ability to tap into a large amount of data suggest that
this system could speak plainly with callers, offering them specific advice on even
technical and complex questions. It’s easy to see it replacing many human holdouts
in its new field.

Digital Losers

The contention that automation and digital technologies are partly responsible for
today’s lack of jobs has obviously touched a raw nerve for many worried about
their own employment. But this is only one consequence of what Brynjolfsson and
McAfee see as a broader trend. The rapid acceleration of technological progress,
they say, has greatly widened the gap between economic winners and losers—the
income inequalities that many economists have worried about for decades. Digital
technologies tend to favor “superstars,” they point out. For example, someone who
creates a computer program to automate tax preparation might earn millions or
billions of dollars while eliminating the need for countless accountants.
New technologies are “encroaching into human skills in a way that is completely
unprecedented,” McAfee says, and many middle-class jobs are right in the bull’s-
eye; even relatively high-skill work in education, medicine, and law is affected. “The
middle seems to be going away,” he adds. “The top and bottom are clearly getting
farther apart.” While technology might be only one factor, says McAfee, it has been
an “underappreciated” one, and it is likely to become increasingly significant.

Not everyone agrees with Brynjolfsson and McAfee’s conclusions—particularly the

contention that the impact of recent technological change could be different from
anything seen before. But it’s hard to ignore their warning that technology is
widening the income gap between the tech-savvy and everyone else. And even if
the economy is only going through a transition similar to those it’s endured before,
it is an extremely painful one for many workers, and that will have to be addressed
somehow. Harvard’s Katz has shown that the United States prospered in the early
1900s in part because secondary education became accessible to many people at a
time when employment in agriculture was drying up. The result, at least through
the 1980s, was an increase in educated workers who found jobs in the industrial
sectors, boosting incomes and reducing inequality. Katz’s lesson: painful long-term
consequences for the labor force do not follow inevitably from technological
changes.

Brynjolfsson himself says he’s not ready to conclude that economic progress and
employment have diverged for good. “I don’t know whether we can recover, but I
hope we can,” he says. But that, he suggests, will depend on recognizing the
problem and taking steps such as investing more in the training and education of
workers.

“We were lucky and steadily rising productivity raised all boats for much of the
20th century,” he says. “Many people, especially economists, jumped to the
conclusion that was just the way the world worked. I used to say that if we took
care of productivity, everything else would take care of itself; it was the single most
important economic statistic. But that’s no longer true.” He adds, “It’s one of the
dirty secrets of economics: technology progress does grow the economy and create
wealth, but there is no economic law that says everyone will benefit.” In other
words, in the race against the machine, some are likely to win while many others
lose.