Will "smart machines" steal our jobs? With America suffering through its sixth straight year of anemic job growth, worries about long-term unemployment are in vogue. Thinkers such as Tyler Cowen and Kevin Drum believe that the problems will get even worse in the coming decades: As information technology allows the automation of more and more middle-class jobs, fewer workers will be able to find work.
Massachusetts Institute of Technology professors Erik Brynjolfsson and Andrew McAfee have a more optimistic view, advanced in their new book, "The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies." This is an important book that has already landed on The New York Times bestseller list. Like Cowen and Drum, they predict that intelligent machines will increasingly displace many workers. But they believe that work will eventually shift to new jobs that technology creates. Workers will be displaced, not replaced.
Brynjolfsson and McAfee have the stronger argument. Indeed, economic data show that as information technology has increasingly taken over white-collar tasks in recent decades, it has not significantly replaced workers overall. While new technology has contributed to growing economic inequality, the evidence so far does not point to a jobless future.
The Second Machine Age
The MIT scholars make a convincing argument that we are about to see a whole new wave of disruptive information technologies. Major new advances in artificial intelligence are challenging common assumptions about what computers can and can’t do. Not long ago, experts thought that tasks such as driving a car, understanding speech and translating languages were in the distant future. But Google and other companies now have driverless cars on the road. Cellphones do a passable job understanding speech (Siri) and translating language (Google translate).
These changes have major implications for the future of work. Brynjolfsson and McAfee tell us “there’s never been a worse time to be a worker with only ‘ordinary’ skills and abilities to offer, because computers, robots, and other digital technologies are acquiring these skills and abilities at an extraordinary rate.” In the past, new technologies tended to automate blue-collar jobs. Now information technology has begun automating white-collar jobs, and the new technologies will increasingly automate even professional jobs. Already computers can diagnose breast cancer from x-rays and predict survival rates at least as well as radiologists.
The authors illustrate the economic impact of second machine age technologies with the example of TurboTax software for preparing tax returns. Millions of people now use this $49 program to prepare their income taxes, but the effect is to increase economic inequality. A small number of designers and engineers created this product; one of them has become a billionaire. But at the same time, “tens of thousands of income tax preparers now find their jobs and income threatened.”
But does this kind of disruption spell a future where most people can’t find work? The authors make a mainly theoretical argument that smart machines probably won't lead to permanent mass unemployment, given the right policies. Jobs don’t necessarily disappear when a new technology takes over some human tasks. This is because new technologies also increase demand for products and services. Consumers will demand more product if new technologies add desirable features or decrease the price, thanks to the smaller amount of labor needed to produce the product. Thus while less labor is needed to produce a unit of the product, the number of units that are sold increases. If demand increases sufficiently, the number of jobs will not decrease and might even increase.
For example, TurboTax has not eliminated tax preparers. In 2011, 56 percent of individual returns were done by paid preparers, about the same percentage as a decade earlier. While many taxpayers use TurboTax, the majority of taxpayers still demand the services of professional tax preparers. Tax preparers use software, too, and by working with the smart machines they provide improved services that consumers want.
Supply and demand
In the past, new technologies increased demand enough so that employment grew. For example, I've written about how, during the 19th century, machines took over much of the work of producing textiles. But the price of cloth dropped sharply, and people bought much more cloth. Enough cloth, in fact, so that textile employment grew.
Of course, as Brynjolfsson and McAfee note, past performance is no guarantee for the future. Today’s technology might be different and eliminate jobs on a mass scale. Nevertheless, the history is helpful because it shows why machines don’t necessarily eliminate jobs even when they take over human tasks: the key is demand. New technologies can increase demand, creating new jobs, but they do not always do so. During the twentieth century, textile technology continued to reduce the labor needed to produce a yard of cloth, but consumers no longer responded by buying much more cloth — there is only so much cloth people want to buy even at low prices. As a result, the number of textile jobs stopped growing and since the early 1970s has been declining. After 150 years of continually improving technology, the machines finally did replace textile workers.
But how will today’s technology affect demand? Brynjolfsson and McAfee argue that the answer depends on whether technology taps into major new sources of unmet consumer demand.
As long as there are unmet needs and wants in the world, unemployment is a loud warning that we simply aren’t thinking hard enough about what needs doing. We aren’t being creative enough about solving the problems we have using the freed-up time and energy of the people whose old jobs were automated away. We can do more to invent technologies and business models that augment and amplify the unique capabilities of humans to create new sources of value, instead of automating the ones that already exist.
So while technology might eliminate jobs in some older industries, as long as new technologies generate major new demand meeting new needs, the net effect does not mean permanent unemployment. Clearly some new technologies such as the driverless car will, indeed, address major unmet needs. In this process, specific jobs and specific occupations will be eliminated. This may increase economic inequality for a time. And the new opportunities will require new skills and new business models; these might be difficult and slow to develop. Nevertheless, this view of the future differs sharply from the predictions of a dystopia with permanent mass unemployment and ever-widening economic inequality.
The empirical case for optimism
Brynjolfsson and McAfee’s argument is theoretical, but we can identify significant evidence supporting their view. The strongest empirical support comes from a close look at how information technology is affecting jobs today. If one believes that machines will become roughly as intelligent as humans fairly soon, then we should see signs that the machines are already beginning to take over.
Indeed, some people, such as Kevin Drum contend that smart machines have been permanently replacing humans for some time. Yet the data show that the first wave of computer technology has displaced workers, not replaced them. It is true that computer technology has taken over many tasks formerly performed by humans, but this does not necessarily mean that it has eliminated jobs overall. Since the 1970s, automated teller machines have taken over tasks from bank tellers, word processing software has reduced the need for typists, voice mail has replaced switchboard operators, accounting and logistics software has taken over work of bookkeepers and clerks, and desktop publishing has taken over work of typographers and compositors. But these changes have also increased demand for the services affected, offsetting the job losses associated with automation.
According to data from the Bureau of Labor Statistics, there are more bank tellers now than before ATMs. This is because the number of bank branches increased dramatically — the lower cost of operating a branch allowed banks to meet the needs of many more consumers. In other cases, jobs were displaced to different occupations. There are fewer switchboard operators, but more receptionists. There are far fewer typographers today, but many more graphic designers, and some 18 million people now do graphics and design work on their computers. Desktop publishing and design software increased the demand for highly designed publications of all sorts.
So, while technology eliminated jobs for typographers, switchboard operators, and typists, it also created other new jobs. To see whether machines replaced workers overall, we need to look at the total number of jobs affected by technology, not just individual occupations. This table, using data from the Current Population Survey of the Bureau of Labor Statistics, shows job growth rates since the computer revolution began across broad occupational groups ranked by their use of computers:
Share of Workers Using Computers (2001)
Annual Job Growth Rate (1982 - 2012)
|Scientific, engineering and computer occupations|| |
|Administrative and sales occupations|| |
|Healthcare occupations|| |
|Manufacturing production and maintenance occupations|| |
|ALL OCCUPATIONS|| |
While the workforce as a whole grew 1.1 percent per year, the occupational groups that used more computers grew faster, not slower. In particular, administrative and sales occupations — the group that includes bank tellers, typists, bookkeepers, etc. — grew a bit faster than the workforce despite the effect of technology on many jobs within this group. In other words, the jobs were displaced to other occupations within the group; the new technology created offsetting increases in demand. This displacement has its costs: the new jobs require skills and education that many mid-wage workers lack and this contributed to growing economic inequality. But computer technology did not replace workers in these occupations overall.
That was not the case with manufacturing-related occupations where jobs really did disappear. Many of these jobs are in mature industries such as textiles where lower costs did little to increase demand. Also, many jobs in these industries were sent offshore and were not necessarily lost to technology. Nevertheless, thanks mainly to technology, manufacturing employment has grown more slowly than the labor force since the 1940s, leading to a declining share of jobs.
But much of this technology was not computers, although computer numerical control systems and robotics surely contributed recently. Still, computer use in these jobs is low, and a recent MIT study finds that manufacturing industries that use computers intensively did not experience job-reducing productivity growth during the last decade. The loss of manufacturing jobs provides little support for the idea that smart machines are permanently replacing humans in the the workforce overall.
Machines don’t determine our fate
The job losses in manufacturing will likely continue, but they can be offset by growth in other sectors as they have until the recent Great Recession. Since the recession, overall job growth has been anemic. While other factors are likely at work — including the largest downsizing of government employment in modern history — new technology does contribute to unemployment when it requires skills that workers don’t have.
Yet the weak recovery provides little reason to conclude that machines have completely taken over human tasks. As long as technology continues to address major unmet needs, machines do not determine our fate. Just because machines take over some human tasks, that does not mean the end of jobs. We do, however, need to figure out how workers can develop new skills and how entrepreneurs can create new business models to generate the new demand that will provide growing employment.
Brynjolfsson and McAfee recommend a series of sensible policy recommendations to encourage education and entrepreneurship. If we fail, it will be from a lack of imagination and will, not from the predetermined workings of technology.
James Bessen does research on technology and innovation at Boston University School of Law. He is the co-author of "Patent Failure: How Judges, Bureaucrats, and Lawyers Put Innovators at Risk." He is currently writing a new book about technology and jobs. You can follow him on Twitter.