Will They Take Our Jobs?

MIT economist Andrew McAfee on driverless cars, wireless fishermen, and the second machine age


Massachusetts Institute of Technology economist Andrew McAfee hasn't been replaced by a robot just yet. The following interview was conducted between two humans. Neither of the humans needed to bother to remember what was said, however: We recorded the conversation on an iPhone app, essentially outsourcing memory to a computer. Pre-interview research was conducted with the aid of Google-no humans required there either, just well-crafted algorithms pointing in the direction of McAfee's blog, popular TED Talks about automation and unemployment, and Amazon author pages. But the resulting MP3 file? It was transcribed by a human intern. Accuracy was important, and commercially available voice-to-text programs just aren't good enough yet. Which is a bit disappointing. Especially for the intern.

The rapidly shifting interplay between tasks that humans still do and tasks we delegate to our automated servants should feel like a familiar progression. In the first machine age-the Industrial Revolution-we replaced human brawn with steam power. But in so doing, we wound up creating more demand for labor: We needed people to tend to the increasingly complex machines, and to staff entire new industries that arose once humans were freed from the burden of lifting heavy stuff.

We continue to contract out our need for brawn to machines, say McAfee and his co-author Erik Brynjolfsson, but we have also started replacing human brains with processing power. In their 2014 book, The Second Machine Age, the economists describe a new world driven by the relentless doubling of computer processing capacity, known as Moore's Law. McAfee, who has a Ph.D. from Harvard Business School, cheerily anticipates a fresh profusion of consumer goods from this machine age, similar to the glut produced by the last one. But he says he's no Candide-like many, he predicts that this time there will be no compensating boom in demand for human labor and he's worried about the social and economic effects of widespread unemployment. Is he right? Are things really different this time around?

In January, Managing Editor Katherine Mangu-Ward spoke with McAfee about the economics of the robot revolution.

reason: You rode in the Google driverless car. Tell me about it.

Andrew McAfee: The experience went from terrifying to passionately interesting to boring in the space of one ride.

reason: Why was that?

McAfee: When the guy who was driving the car hit the big red button and took his hands off the wheel on the highway, that was a white-fingernail moment.

reason: Is there literally a big cartoon red button?

McAfee: There's honestly a big cartoon red button on the dashboard.

reason: That's delightful. So he hits the button-

McAfee: -and takes his hands and feet off the controls, and we're going at highway speeds in a completely self-guided car. That was a little scary. Very quickly that passed, and then it became super interesting, because I felt like an astronaut. I'm having this really uncommon experience, and after a while, it sunk in that I was in a car that was obeying all relevant statutes, not weaving, not seizing opportunities in the right-hand lane, going down the road at 55 miles per hour. I mean this as the highest compliment: It was a godawful boring ride.

reason: What are some places in everyday life where people may be undervaluing the extent to which the robots or machines have already taken our jobs or taken over our lives?

McAfee: I won't say "taken our jobs," because I still have one. A lot of these changes don't keep screaming at you. They happen kind of gradually. They're bit by bit, but then you look up and you're living your life pretty differently than you did a few years ago.

For me, professionally, if I could sit down and look at what I was doing a decade ago or 15 years ago, I think it'd be night-and-day different. When I sit down to start writing something or to learn something, I basically have 30 tabs open on my browser. I'm searching for a little stat, or I pull up a number from the St. Louis Fed that's got this great data repository. I don't go to the library; I don't fire off requests to research librarians. I use a research assistant for some things, but not for "hunt down this fact for me," simply because it's easier and quicker for me to do it myself. When you've got the world's knowledge at your fingertips all the time and you're supposed to be doing knowledge work, it really does change the way things happen.

reason: And how do you think that applies-if it does-to people who are doing a different kind of job? There are some guys in reason's office right now assembling a million new desk chairs. It looks like their jobs aren't very different. Am I wrong?

McAfee: I think that part of their job is probably not very different, but how they got their day's schedule, how they communicate with the head office, how they alert them that the job is done, the extent to which they're monitored-maybe their truck has a GPS device in it so headquarters knows where they are-I think those things are actually pretty big changes.

In long-haul trucking, for example, the industry has actually transformed itself, and trucking companies started owning trucks again instead of giving them to subcontractors, mainly because they could monitor the drivers so carefully that they didn't have to rely on the fact that people take better care of their own equipment.

Let me give you one from my nonprofessional life. I moved to New York City for the first half of 2015. Let's assume that I didn't have any friends here.

reason: Should we make such a sad assumption?

McAfee: (Laughs) No, luckily I've got a lot of people to hang out with and to show me around. But let's say I didn't have any of that but I was still interested in finding a good café to go hang out at, at exploring different parts of the city, at getting around efficiently.

I would do that by trial and error before. I'd make a ton of mistakes. I personally would find it really stressful, because I hate being lost and I hate feeling stupid.

Those problems are basically gone for me. I've got an app called City Mapper on my phone. I'm pretty sure it was free. All I ever do is say, "I'm here in the Upper East Side, I'm going to meet a friend for dinner down here in the West Village. How do I get there?" And it says, "OK, you walk over to Lex and 63rd, hop on the F train, you'll take it six stops, and get off here." It's incredibly detailed information about how to navigate a very unfamiliar city, so I can get around about as well as somebody who's lived here a long time.

reason: Tell me what this has to do with the fisherman in Kerala.

McAfee: That is probably my all-time favorite and most heartening solid piece of research about what's going on. A guy named Robert Jensen got to observe the economic lives of some systems-level fishermen in Kerala, India, before and after they got mobile phones for the very first time ever.

These folks were living in an I.T. vacuum. They'd go out every day and do their fishing, and they'd come back in and have to pick which local market to go to, to try to sell their fish. And you can imagine all the inefficiencies that would result because you couldn't match supply with demand carefully. Some days they would do great. Some days they would do lousy. Some days they would have to throw their fish away because nobody would pay them anything for it. It was a terrible situation.

In a really beautifully designed study, Jensen got to watch what happened before and after cell phone towers went in at different points along the coast. So he had a bunch of different experiments, and he saw the same thing over and over and over again. Markets start to behave predictably and rationally immediately after the new technology becomes available. The first thing these people did was all go and buy a phone, because none of them are stupid, and they would use it to call ahead and say, "What's the price at this market? Should I go over here?" And you just watch the markets regularize and clear in a way they could never do before.

This is an example of what happens, what's happening over and over and over around the world, as these new technologies diffuse. We are greatly improving the lives of people in a lot of ways.

reason: The current education system is almost hilariously unsuited to this universe that you have just described. Tell me why everything is bad and how you can fix it.

McAfee: I think there are a lot of really extraordinarily hard-working people in education, and I don't presume for a second to have all the fixes. But one thing that our primary education system is doing a really good job of is preparing the kinds of workers that we needed 50 years ago in the height of the industrial era. They acquire a suite of skills: They can read, they can write, they can do math at some level. And more fundamentally, they're encouraged to follow instructions and to be obedient. You sit in the same place. You go through this orderly process. People in the front of the room talk to you. It's great training for industrial-era white-collar and blue-collar workers. It's pretty lousy training for the kind of thinking and the kind of people and workers that we're going to need as we move deeper into the second machine age.

reason: So what's better?

McAfee: I was a Montessori kid, and I'm incredibly grateful I was a Montessori kid, because my earliest education bore no relationship to that system I just described. It taught me the world was an interesting place and my job was to go poke at it.

reason: You've said that entrepreneurship is something we should encourage in American kids and welcome in our immigrants. Why that, specifically?

McAfee: I haven't seen a computer that could convince investors to put together a business plan or really spot an opportunity and figure out how to go after it. That still does feel to me like a human skill. But as we mentioned in the book, entrepreneurship, and in particular tech entrepreneurship, has been driven by immigrants to a wild degree, and the people who want to come to this country very often are the kind of tenacious, ambitious, hard-to-satisfy ones. These are exactly the kinds of folks that you want to come in if you're interested in entrepreneurship. So especially at the level of skilled immigration, I find that kind of the biggest policy no-brainer out there. Even at the low-skill levels, we're not displacing tons of native workers from jobs.

reason: "Income inequality," "coming apart," "two Americas." There are lots of names for the ways that rich people and poor people are economically separating, particularly in the labor market. You call it "the spread."

McAfee: There used to be a bunch of economic measures that all went up and down together, luckily primarily up. They did it in lockstep. They were really tightly coupled. And then, in recent years, we start to see these measures head in different directions and gaps opening up between them.

For example, one of the graphs we draw has four lines on it for the entire postwar period: GDP per capita, labor productivity, raw number of jobs, and median family or median household [income]. For decades after the end of the war, they were all going up, and they were all going up just super, super close together. Around 1980, the average median family income line starts to tail off. More recently, the job growth line starts to tail off. And the job growth line starts to tail off before even the great recession kicked in. Job growth was fairly anemic all throughout the 2000s.

We call that phenomenon "the great decoupling." It's an example of this spread. You see it in returns to labor vs. capital. You see it in these four lines. You see it when we look at wealth and income measures. Thomas Piketty certainly sees it [in Capital in the Twenty-First Century]. He just looks at a couple aspects of it and labels them inequality, but these are all manifestations of a pretty common phenomenon.

reason: People get very emotional about this topic. You can see that in the response to the Piketty book, and you can see it in lots of other peoples' writing, including Tyler Cowen's The Great Stagnation and Charles Murray's Coming Apart. So before we get to your solutions, give me your "So what?" Why does everyone care quite so much, given that the vast majority of people are doing better, there's just a differential in the gains that seems to be opening up.

McAfee: Let's be careful about that. We are all doing better as consumers-as people who want access to goods and services, and who want more of them, who want higher variety, higher quality, lower prices, all those things. The bounty that comes out of capitalist systems, and in particular technologically driven ones, is just stupefying. It's pretty unbelievable, and I find that unambiguously good news.

The challenge comes when I look at things like the median American household income. Even after we adjust for inflation and for changes in family size, it's not that it's growing more slowly than it used to, it's actually lower than it was 50 years ago. For me, that's a decent answer to the "So what?" question, because the fact that it's real income means that it represents our best attempts to take into account the fact that flat-screen TVs cost less than they used to, that it is your actual purchasing power. It's not a precipitous decline, it's not that the middle class is starving in the streets, but it is a slow, steady decline.

That's part of the "So what?" answer. Another part is that there are some important categories of stuff that are not getting a lot cheaper over time. Higher education, health care, housing. Now, we can have a really active debate about why they're following different trajectories and whether we should head more toward libertarian-style market solutions for that. That's a really important, valid debate. It's a bit of a separate question from the fact that are these things getting more or less affordable to the American family at the 50th percentile, and in a lot of cases they're becoming less affordable.

It's also becoming more clear as we get the evidence that social mobility is not where we think it is. The economic circumstances of your birth seem to play a really large role in this country in determining your economic life trajectory, even more so than they do in a lot of these European social democracies that we like to disparage. The low mobility is also part of an answer to "So what?"

Charles Murray has documented that among lower-middle-class Americans, there's been, over the past half-century, a really alarming rise in a bunch of social ills: in drug use, in dropping out of the labor force, in not staying married, in children raised in single-parent homes, in incarceration rates. What's interesting to me is that all those go along with a really sharp decline in work, just being engaged in a job. Those social ills are almost nonexistent in upper-middle-class Americans, and those upper-middle-class Americans have been working pretty steadily through this period as well. Murray would disagree with the following: My very simple narrative there is that work is a really good thing to have as technology encroaches and takes away some of the classic lower-middle-class job opportunities. I think we see some social ills coming out of that.

And then the last part of the answer is that there's some pretty alarming data that among the lower rungs of the education and income ladder, health outcomes are heading in the wrong direction. Average life span, for some demographic groups, is actually going down recently in America after decades of pretty impressive gains.

I put all those things together, and I don't find it easy to be blasé about the spread.

reason: You've said nice things about work for work's own sake. But actually, people hate work, don't they? Most people hate their jobs, at least some of the time. So why do you want them to keep working?

McAfee: Among people who have looked pretty hard at this, there's a really broad consensus that when work-I won't say jobs-when work goes away from the community, relatively few good things happen and lots of bad things happen. And again, that list, that litany that Murray put together, is pretty telling to me. I don't want to pretend that if everybody had a job, all those things would magically go away, but I do believe that part of the reason that these ills creep in is idleness and not having the sense of purpose and dignity that comes along with the job. I don't think those are just empty things.

reason: So if work has these good effects, and we're concerned about culture and economies coming apart, and meanwhile McDonald's is automating order-taking and burger-flipping and Google is automating driving, then…what?

McAfee: I want to be clear: I don't demonize McDonald's and Google and all these other companies for trying to use technology and use automation. They're trying to keep their costs low. They're a business. They're not a social welfare organization. And they're doing it because they think they deliver better goods and services to all of us. So I'm not saying that companies should take one for the team somehow and just start bringing on lots of labor willy-nilly for the good of the community or the good of society.

But all these companies acting in their own interests are generating, I think, less labor demand than was the case previously. For about 200 years, we had this wonderful phenomenon where, as the capitalist engine progressed, it needed a ton of labor at all different levels of skill, and instead of dropping out, instead of mass unemployment, instead of mass starvation, we had the rise of a large stable prosperous middle class in country after country.

It feels to me like this time might finally be different. The data that I talked about are not just blips; they look like trends. And when I look at tech progress, I don't see it changing course.

Now, what do we do about it? I think we try as hard as possible to prove me wrong and to make this time just like all the other times, where even though there was a lot of tech progress, the average worker wound up with a better job and a higher wage.

reason: And how do we do that?

McAfee: In the book we tried to concentrate on the really uncontroversial parts of the Econ 101 playbook. And you've got to go a long way outside of the mainstream economics profession before you'll find someone who'll say that the government should not be involved in building out infrastructure or primary education or basic research, because the private sector tends to undervalue and therefore underfund that kind of stuff. So our playbook consists of things like education reform and immigration reform and increased focus on entrepreneurship and doubling down on infrastructure and revitalizing basic research. To me, that's our best chance to create an economic environment that would let the happy pattern repeat itself and bring labor demand back. There's no way that labor demand is going to come without a lot more economic growth. Great. Let's do what we can to get the economic growth.

reason: In the next, say, five to 10 years, what are the first jobs to go?

McAfee: One of the quickest ones to me looks like different flavors of customer service reps, where they're using their language skills. They're using their pattern-matching skills. Our technologies are really, really good at both of those right now. They're going to get worlds better over the next five to 10 years, so people doing that kind of knowledge work, I think, are going to face some unemployment headwinds.

Depending on the regulatory environment, I think a highly functional, autonomous vehicle is easily in that timeframe, so we have a lot of people who drive for a living now who are going to be confronted by automation.

I think if a piece of technology is not already the world's best medical diagnostician, it easily will be in five or 10 years. Now, I don't know if, again, there are going to be regulatory policy changes that would allow that technology to diffuse. But if that happens, we've got a lot of people who diagnose us for a living who are going to be confronted by technology that does it better.

What then happens in these different fields is not that the employment goes down to absolutely zero. It's that it goes down to a pretty small number of very competent, pretty high-level people supported by a ton of automation.

reason: That's something that has happened in lots of other places already, right?

McAfee: Yeah. Longshoremen are the classic job where that happened in the 20th century, but the happy phenomenon is that other industries sprang up that, again, needed labor at all different skill levels. I'm encouraged by things like Uber and Airbnb and the rental economy that's giving average people a chance to earn some money. That's great. I hope it continues.

reason: There's this vogue for famous technophiles to freak out about artificial intelligence [A.I.]. We've got a statement from the Future of Life Institute signed by Stephen Hawking and Elon Musk, saying basically, "Everyone panic, the robots are going to kill us all." Are they right?

McAfee: This is just not high on my list of concerns at all. The best I ever heard it explained is that we are multiple Watson and Crick moments away from anything like a Terminator or a Matrix scenario.

I could be wrong about that. I could easily be wrong. In which case, oops. Because the interesting point they make is, "Look, even if it's a very low probability of that, and even if it's kind of a long way off in the future, we're talking about an existential risk." OK. It's easy to look at some of the recent advances and extrapolate them forward and say, "Holy Toledo."

reason: What do you think is the most "Holy Toledo"-inducing advance recently?

McAfee: The most telling demonstration for me was when the guys at Deep Mind Technologies told their system to learn to play classic '80s-vintage Atari video games. They didn't tell them the rules of the games, they didn't tell them what controls they had, they didn't try to tell them what was good or what was bad or advanced or "shoot that tank, but don't shoot that thing over there." All they said was to the system, "Your job is to maximize that number up there, which is called the score. Knock yourself out." For the majority of the games that they included, the system is now the world's best player.

reason: How did it do on Pong?

McAfee: You would never score a point against it on Pong.

reason: That's disappointing. That's a lot of time wasted by a lot of teenagers.

McAfee: Yeah. It's the world's best Battlezone player, and I played a lot of Battlezone. I'm not getting those hours back. (Laughs)

reason: What technologies are people currently undervaluing and what tech are people currently overvaluing?

McAfee: I think we're simultaneously overconcerned about A.I. progress in an existential sense and underconcerned about it in an economic sense. Because I do think that these advances are going to pretty quickly enter the business world, and I think they're going to accentuate all these phenomena that we talk about in our book.

I personally think 3D printing is extraordinarily cool, and it's going to help with our innovation work and our prototyping and stuff like that. There are people who believe it's going to massively disrupt office supply chains and the manufacturing industry and everything all around the world in some realistic timeframe. I don't see that.

reason: So you're telling me that the future of, "Computer, please make me a ray gun" is further off than I was hoping?

McAfee: That's actually going to-if you want to invest the time to put one of these things in your house and learn to use it and acquire the plans, you can print out your gun. People have done that. What I don't think is that all the gun manufacturers should say, "Oh man, all of our big centralized factories are now completely worthless."

reason: What is the "to be sure" paragraph you wish you had put in the book?

McAfee: Ask me that question in a few more years. Maybe the job market's going to spontaneously tighten back up and the middle class is going to get on a healthy trajectory again and this whole book is going to stand as another example of "Ha ha ha, see how terrible that timing was."

reason: Right.

McAfee: And I guess when the Terminator comes and knocks on my door, I'll say, "Gosh, I wish I'd been a little more guarded about the prospects for artificial intelligence."

reason: I think the Terminator's going to let you live, because you're convincing all of us to lower our defenses.

McAfee: That's true. I could be the quisling for the Terminators, right? I'll be their intermediary.