Skip to Content

When Machines Do Your Job

Researcher Andrew McAfee says advances in computing and artificial intelligence could create a more unequal society.

Are American workers losing their jobs to machines?

That was the question posed by Race Against the Machine, an influential e-book published last October by MIT business school researchers Erik Brynjolfsson and Andrew McAfee. The pair looked at troubling U.S. employment numbers—which have declined since the recession of 2008-2009 even as economic output has risen—and concluded that computer technology was partly to blame.

Advances in hardware and software mean it’s possible to automate more white-collar jobs, and to do so more quickly than in the past. Think of the airline staffers whose job checking in passengers has been taken by self-service kiosks. While more productivity is a positive, wealth is becoming more concentrated, and more middle-class workers are getting left behind.

What does it mean to have “technological unemployment” even amidst apparent digital plenty? Technology Review spoke to McAfee at the Center for Digital Business, part of the MIT Sloan School of Management, where as principal research scientist he studies new employment trends and definitions of the workplace.

TR: What’s your definition of automation?

McAfee: The obvious definition is one fewer job than there used to be, with the same amount of output. A tax preparer can get automated away by software like TurboTax, and just not find work anymore. An assembly line worker could be flat-out automated away by a robot on the assembly line. There is a closely related phenomenon, which is the massive increases in productivity brought on by digital technology. An example is the legal discovery process. By one estimate we heard, one lawyer is now as productive as 500 used to be. You might not lay off 500 lawyers, but the next time you might hire a few people and some software to read documents.

Where do you see automation leading to the loss of jobs?

Others have done work showing that if you are a “routine cognitive worker” following instructions or doing a structured mental task, you have been under a lot of downward wage pressure for a while now. I think that is largely a technology story. Payroll clerks, travel agents—we don’t have as many of them as we used to. We don’t have as many people working in manufacturing, even though manufacturing is a growing industry.

What was the response you received to Race Against the Machine?

People accepted that technology was really accelerating and that there were going to be labor-force consequences. The broader discussion was between optimism and pessimism. Does it feel like we are heading into the kind of economy and society that we want, or the kind of economy and society that we don’t? A lot of people who commented said, “Look, if these guys are anywhere near right, we are heading into an economy that is going to be dire for a lot of people.”

What does the economy that we don’t want look like?

The spread between the haves and the have-nots continues to grow, and more importantly, the absolute standard of living of the people at the middle and the bottom goes down. That is the economy that I don’t want to head into.

What is the optimistic view?

Erik Brynjolfsson came up with a great phrase: “digital Athens.” The Athenian citizens had lives of leisure; they got to participate in democracy and create art. That was largely because they had slaves to do the work. Okay, I don’t want human slaves, but in a very, very automated and digitally productive economy you don’t need to work as much, as hard, with as many people, to get the fruits of the economy. So the optimistic version is that we finally have more hours in our week freed up from toil and drudgery.

Do you see evidence for a digital Athens on the street, in the real economy?

No. What we are seeing—and this was pretty much unanticipated—is that the people at the top of the skill, wage, and income distribution are working more hours. We have this preference for doing more work. The people who have a lot of leisure—I think in too many cases it’s involuntary. It’s unemployment or underemployment. That is not my version of digital Athens.

Which is further advanced, the automation of intellectual work or of physical tasks?

The automation of knowledge work is way, way farther along. It’s really hard to get computers to do things that your four-year-old can do, like walk across the room and pick up a pen, and recognize it as a pen. So the physical world presents a lot of challenges to digital technologies.

But it feels to me as if we are starting to turn a corner. The data available to help a robot is big data, and it’s exploding. The sensors have been progressing along a Moore’s Law trajectory. And the physical pieces of a robot, the actuators and so on, have gotten a lot better too. So it seems the ingredients are all in place for the robots to start getting into the economy.

How should businesses react to the trend toward more automation?

I think the companies that succeed going forward are the ones that figure out what mix of human and digital labor is going to be the right mix. And I think that that proper mix is going to involve more, and more types of, digital labor than we are using right now.

What is your advice to the individual, or to the parent educating a child?

To the parent, make sure your kid’s education is geared toward things that machines appear not to be very good at. Computers are still lousy at programming computers. Computers are still bad at figuring out what questions need to be answered. I would encourage every kid these days to buckle down and do a double major, one in the liberal arts and one in the college of sciences.

Despite the glum view of changes in the labor market, you’ve used the word “cornucopia” to describe the results of innovation. That sounds very encouraging. What do you mean by that?

We have access to amazing digital resources. And a lot of it is all-you-can-drink, no matter what your income level is. Wikipedia is distributed to the masses. Warren Buffett doesn’t have any more Google than I have, or the unemployed person has. When I see that there are five billion mobile-phone subscriptions in the world—well, hey, that is cornucopia. It is important not to lose sight of that.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.