How the rise of smart machines will affect the US economy and jobs: A Q&A with Erik Brynjolfsson and Andrew McAfee


The age of innovation and technological advancement is not over. That’s the reassuring news from The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies by MIT’s Erik Brynjolfsson and Andrew McAfee. Economic growth is not over, despite the slow recovery after the Great Recession. Not by a long shot.

But accelerating automation will cause huge, disruptive changes for the US labor market — in fact, we’re already seeing them — as workers “race against the machine.” The ability to work successfully with technology will become ever more valued, just as failure will be increasingly penalized. America will need to upgrade the education and skill of its labor force and help entrepreneurs create new business models that incorporate the unique capabilities of man and machine.

I recently chatted with Brynjolfsson and McAfee for my Ricochet Money & Politics Podcast on what the rise of the machines means for economic growth, job creation, and public policy such as immigration. Here is an edited transcript of our conversation:

In the book, you guys write that “we’re living in a time of astonishing progress with digital technologies, those that have computer hardware, software, and networks at their core.” That probably seems intuitively right to many people. But do we see that in the data? 

Brynjolfsson: Well, I think there are some big problems with the way we measure our productivity, as you know. We are missing more and more of the digital revolution. All the free goods, by definition, don’t show up in GDP. And so that’s a big chunk of it. And new goods are badly mismeasured as well, especially when they are introduced and then a decline in price. So I think we have some serious problem with the economic statistics.

McAfee: Your question is a really good one. One of the problems is that we don’t have a universally agreed upon measure for rates of innovation, for whether that is speeding up or slowing down. So we fall back on related measures, like productivity growth and GDP growth, but those aren’t perfect proxies in any case. And like Erik says, there’s reason to believe they’re getting less good over time.

So the reason we wrote sentences like that in the book and talked about this age of astonishing technological progress is that just in the past few years, we have seen some of the really long held goals or holy grails of computer science and AI and robotics being realized, and not just in the lab, but out there in the real world. So we’ve got autonomous cars that drive themselves in traffic, on American roads without mishaps. We’ve got a computer that is the world’s Jeopardy champion, doing really tough unstructured search in natural language processing. We’ve got things on our phones now that are pretty close to the “Star Trek” computer that understands what we want and can give it back to us. And you can just keep going down the list. Humanoid robots are real things now out there in the world, instead of just being stapled with science fiction.

So over and over again Erik and I saw these examples of real, just long held goals of the disciplines and science fiction technologies becoming reality. It’s very hard for us to look at that as a period of slow or uninspiring innovation.

I’ve also interviewed Tyler Cowen, who wrote “Average is Over” and “The Great Stagnation,” and he said that while he disagrees with you about the past, he agrees about the future. And we just had a podcast with former Fed economist over here at AEI, Steve Oliner, who’s also fairly optimistic about where innovation in going. But then there’s Robert Gordon, economist at Northwestern, who’s probably the biggest opponent publicly of the idea that we are seeing innovation accelerate. He just doesn’t see it in the numbers.

Brynjolfsson: I think that we’d agree that what matters is the value that we’re creating, not whether a particular metric moves – especially a metric like GDP, which often literally goes in the opposite direction of welfare. When things become free, that can often lead to a decrease in measured GDP, even though it leads to a big increase in welfare. Wikipedia is a perfect example of that. Or take the fact that most people now have, you know, a device that gives them turn-by-turn driving directions. It’s pretty much free with most smart phones. But a few years ago, people were paying hundreds of dollars for a GPS machine. So I think we have to be careful about overreliance on a metric that was never understood to be or shouldn’t be understood to be a welfare metric.

And let me just pick up on this question about, you know, economic growth more broadly. I mean, people do talk about, you know, secular stagnation and all. And I think it’s worth pointing out that that term was coined back around 1939 by Alvin Hansen, as we were in a really bad depression, coming out of a slump even worse than the second worst one, which is the one we’re just coming out of now. And it’s not surprising that people sort of get pessimistic about that future, when you’re in that. But careful research has suggested that the 1930s were actually among the most innovative decades ever, a whole host of technological marvels, and the next two or three decades, the ’40s, ’50s, and ’60s, were among the highest productivity growth decades ever.

So I think it’s probably not a good idea to extrapolate from productivity numbers in the middle of a recession that we know are flawed to what’s going to be happening in the future. I think a better approach is what Andy laid out is looking at the actual technologies, the actual changes that we’re seeing and understanding what the implications are of them.

But GPS and iPhones aren’t exactly the combustion engine. These are interesting technologies, but they’re not sort of the big general-purpose technologies that are going to radically change our lives, are they?

Brynjolfsson: I just have to disagree fundamentally with that. Probably two of the biggest changes in human history were right on the cusp of, you know, true machine intelligence, and then a digital network that connects billions of brains together, so they can not only draw on the collected knowledge of humanity, but they can also contribute to it. As general-purpose technologies go, the ability to loosen the limits of our – of the human brain, of cognitive tasks, is right up there with anything from the first machine age.

McAfee: James, you talk about general-purpose technologies and their ability to radically change our world and it’s absolutely true that both the steam engine and the internal combustion engine are almost universally considered to be examples of these fundamentally important general-purpose technologies. Most economists agree that information technology and the digital stuff that Erik and I talk about also belongs on that list.

So let me paint some, I think, very, very plausible scenarios, likely scenarios about how our lives might change in the future thanks to the kind of stuff we’re seeing. We’re going to have autonomous vehicle easily within our lifetimes. Think how that’s going to open up possibilities and change people’s lives. Families can send their kids around to soccer practice after school without having to have a parent involved. The blind can go run errands and live much more free lives. If that’s not an example of a life changing technology, I honestly struggle to think what is.

The FDA has improved the first digital retinal implants. We are soon going to be restoring sight to the blind. Again, I struggle to think of more fundamental innovations than that. And IBM’s Watson is not just a Jeopardy champion. It’s now going to med school. IBM has announced that they’re putting Watson technology up in the cloud; serve that down through the smart phones that are going to be available, as Erik says, to billions of people, honestly billions of people around the world within just a few more years. And you have the world’s best diagnostician available to the majority of the world’s human population. Again, if that’s not an impressive change for our societies, our lives, and our economies, then I’m out – I’m out of answers.

We have seen productivity and growth statistics slow. Will we see a return to those old, higher rates? Will we see acceleration?

Brynjolfsson: I think a lot of that stuff is going to be difficult to measure in the productivity statistics, but absolutely, we named this the second machine age in – directly to compare it to the first machine age and the steam engine and the ability to eliminate a lot of the limitations on muscle power. Doing the same thing for mental tasks and cognition is going to be at least as big an impact in the number of jobs that are affected, in the ability to grow the economy, we think will lead to a similar inflexion point in growth rates.

McAfee: Erik and I are most confident that even though our productivity measures are not perfect, that we’re going to see a substantial acceleration in measured productivity thanks to the kinds of innovations that we’ve been talking about. But I’m actually even more optimistic and more interested in the kinds of benefits and good things that accrue that are not captured by those measures because like we’ve been trying to point out, more and more stuff is outside the formal tools we have to count dollars and count hours of labor and do a ratio.

But you are sort of implicitly saying that the economy really is growing faster than what we think it is, right?

Brynjolfsson: It is growing faster than what we think it is. But there’s also a really important point we make in the book is that it’s not just the technology alone, but it’s the combination of the way the technology interacts with our skills, our organizations, and our institutions. And one of the things we’re concerned about is the way that they are lagging behind and we give examples in the book, as you may – as you know, that, for instance, when electricity was introduced, research indicates that it was up to 30 years before you saw significant productivity improvements because they didn’t make the complementary organizational innovations. We’re seeing similar lags in rescaling and rethinking our economy and that’s going – that’s one of the reasons that we think that some of the gains are still very much ahead of us.

McAfee: And if you mean by economy something a little bit broader than dollars, if you mean goods and services that people have access to, then absolutely, Erik and I think that there’s a ton of growth going on, everything from music to GPS systems to encyclopedias. We’ve seen an explosion in the availability of these things, even as the dollars associated with each of them have gone down. I absolutely believe our economy is becoming a much, much more abundant place because of technology.

Yet despite that, many people are fearful of what technology is doing to the job market — that perhaps a small slice of the population who are good with computers will have good jobs with high wage growth, and the rest will be physical therapists and high-end butlers. To what extent then, do you envision a future labor market that looks like Cowen’s “Average Is Over” scenario?

McAfee: It’s one plausible trajectory, but it’s far from the only one. And we absolutely don’t think that it’s carved in stone. One of the points that Erik and I try to make repeatedly in the latter chapters of our book is that we have choices, we have policy interventions, we have decisions we can make as a society that will change the distribution of the pie. We’re firmly convinced the pie is going to get bigger because of progress. The distribution becomes the issue. And we’re not determinists of either kind, either the more fatalistic kind like when I read Tyler, I find him a little fatalistic, or the happy, utopian kind that thinks that everything’s going to be fantastic for everybody. We’ve got choices to make. And the last sentence of our book is “technology is not destiny. We shape our destiny.”

I hope Tyler’s scenario is not the best case scenario. What is a better one?

Brynjolfsson: I think it’s great to lay out that as a possible scenario. And if we don’t take any action, then we will tend towards that direction. The reason Andy and I are hopeful optimists is that we think we can take action to do a lot, to re-skill people, to encourage more entrepreneurship that will invent and discover new occupations and jobs for some of those people, the 85 percent or whatever the number is that have their previous jobs automated.

We think that there’s a lot that could be done to encourage work through removing some of the impediments that put a wedge between what employers are willing to pay and what employees receive, flipping it around with things like an expanded earned income tax credit. And we think there’re ways to grow the economy through more investment in infrastructure and boosting immigration.

You put all those four categories together and we think that we could have something closer to, you know, in the ideal scenario, something closer to what we sometimes call digital Athens where there’s an abundance of goods and services and wealth, and there’s shared prosperity in accessing them. And robots do a lot of the more menial and routine work, freeing us up to be engaged in what most of us would consider more interesting kinds of tasks.

McAfee: Yeah, we should keep in mind the US economy is still adding jobs every month. The jobs report that just came out today had more than 100,000 new jobs added. So we are – we do still need more labor to grow our economy, that basic fact hasn’t changed.

To us that indicates that absolutely the right thing to do in the short term is find ways to grow the economy faster. And we use the imagery of an Econ 101 textbook for guidance about how to do that, get infrastructure up to first world standards, get some sensible immigration policies in place, retool our educational system so it’s turning out workers with the skills employers need today, create a favorable environment for entrepreneurship because entrepreneurship is a great engine of job creation, and a government should be doing funding and investment in research. That’s just Econ 101. If we do it, it would grow the economy faster and job and wage growth would pick up.

Let me ask you a question about immigration. I see the logic, given your scenario, of more high-skill immigration. But if you’re talking about an economy where a lot of the routine jobs, low skill jobs can be done either by computer or robot, what would be the argument for the US to have more low skill immigration? How does that sort of fit with your scenario? It seems like we would need less low skill immigration, that there’s a huge risk that we would be bringing in people for jobs that might be automated away in five or 10 years.

McAfee: I think it’s Brookings that estimated recently that illegal low skilled immigration to the US in recent years has basically been net zero because they realized that this isn’t a good employment climate for them. The idea that we’re going to continue to be flooded by low skilled immigrants – if the low skill people who are already here can’t find work, that stretches plausibility a little bit.

So if there’s less demand, just like as we saw during the recession, if there’s less demand for low-skill jobs here, people don’t come here. But what would you advise policymakers who want to, for instance, have a guest worker policy which would bring in low skill workers — particularly for agricultural jobs that might be automated away — and a certain percentage of those people are going to end up staying here? I think that’s the concern of some people who are even overall pro-immigration.

Brynjolfsson: Well, I think the case is certainly a lot stronger on the high skill immigration and that’s really more what we focus on. But to answer your question about lower skill immigration, I think the data show that both the – on average, the existing residents of the country, as well as the new people who come in, are made better off by that exchange. It’s sort of just a generalization of gains from trade. I’m not a fan of having a special category of guest workers. I think that we want to have people treated as equally as possibly when they’re in the country and a broader path to citizenship. But to be frank, that kind of goes beyond the focus of the “Second Machine Age” that we need to analyze.

McAfee: Neither Erik, nor I is an immigration policy expert at all. But we cite a couple of very carefully done studies that show that for example when Mariel boatlift dumped a bunch of very low skilled of people on the shores of Florida, that the natives in the Miami labor market didn’t suffer at all. The same thing happened when Russian Jews came to Israel in very large numbers. Really good research showed that the natives at all levels of the skill ladder did not suffer from that. We should take some confidence from that kind of research.

Hasn’t the history of automation always been that it’ll create jobs? Sure we’re lose some jobs, we’ll lose all those farmers, but we’re going to create all of these other new higher value occupations. We just have to make sure we educate our children. It sounds like what you’re saying is that could happen, but it’s going to be a lot harder.

Brynjolfsson: Yeah, and that’s a great question and it’s a natural one because for a couple of hundred years it has worked out in a positive way like that. But as we point out in the book, there’s no theoretical reason that it has to continue working that way, that, you know, there’s no economic law that says that everyone’s going to benefit from technological project, even as the pie gets bigger. It’s possible for some people, even a majority of people, to be made worse off. That’s the theory.

The data suggests, unfortunately, that we may be in that kind of a scenario in recent years if we don’t make some changes in our policies. Andy and I refer to what we call the great decoupling that people like Jared Bernstein first pointed out that while productivity and GDP and a lot of other metrics like total wealth have continued to grow, the median income and employment have stop growing in tandem with them, starting, you know, around the 1990s. So there has been a decoupling of that employment growth and that median wage growth in a way that we hadn’t seen earlier.

McAfee: But, Jim, the scenario that you just outlined is another plausible scenario and it’s actually the one that I hear most often from the – I guess many of the smartest and best trained economists that I talk to, essentially say what you just said, which is we’ve seen this movie before. There have been huge waves of technological innovation that washed over the economy, led to a period of job displacement, labor force turmoil, lasted a while sometimes, but then we reach a new equilibrium that is essentially another full employment equilibrium. I’ve talked to Nobel Prize-winning economists who say that if we play our cards right, we’ll see that happen again. It’s another plausible scenario. I don’t see that. But it’s another plausible scenario, absolutely.

Brynjolfsson: The thing there is that those positive outcomes in the past also didn’t just happen automatically. There were some big changes that we did in our economy to help make them happen. I mean, it’s been said that the best idea that America ever had was mass universal education. And it’s not a coincidence that as America became a leader in education; it also led to wide shared prosperity and economic growth.

There’re a whole lot of people, tens of millions of people, who have worked in agriculture, who needed a very different set of skills to work in industry and services. And we worked hard to create those. A lot of entrepreneurs worked hard to invent some of those new industries. And the people themselves worked hard to gain those skills. Without those three groups working in those ways, we may not have had such a good outcome.

And what would be sort of the statistics or what would you see up there that would make you sort of rethink your thesis, that Robert Gordon at Northwestern is correct?

Brynjolfsson: Well, my biggest worry – actually, after what I’ve seen on the technology side, I feel quite confident in that and I have pretty strong priers about the direction of that. The biggest worry, though is something that Bob Gordon does point out, which is this growing inequality, and for that matter – so does Tyler and so do a lot of people – and you can imagine a scenario where a lot of these gains get captured by a very small group and that has all sorts of negative effects, not just on the economy, but on our politics, leading to a lot of social unrest. That kind of thing has happened before. And so that’s a scenario that is, unfortunately, very plausible and not a very happy one.

McAfee: And for me, if we don’t see productivity, even with the flaws in its measurement, if we don’t see productivity start to pick up in the next few years, that’s going to cause me to rethink some things and really scratch my head a lot harder. My deepest conviction on the economic statistics, as opposed to the technological ones, my deepest conviction on the economic statistics is that productivity growth is going to accelerate.

When you look 10 years, 20 years down the line, do you anticipate a big class of people who are working a lot less or aren’t working and therefore we need to have some sort of universal basic income?

McAfee: I think if we can redefine what work is and move away from this idea that it’s a, you know, 40-ish hour week job with a single employer, who gives you all kinds of benefits and health care, and expects 40 hours a week from you in return, if we can move away from that to something a lot more fluid with a lot more optionality built in, then I’m pretty confident that a decade out, maybe even beyond that, we still have an economy full of work

Brynjolfsson: I think that that’s very important that we try to work towards an economy that still has ways to keep people working and engaged in – because one of the things we learned in doing research for this was that it’s – that work gives a lot of people a sense of meaning. And that although you can replace the income, simply giving the money doesn’t necessarily lead to greater happiness. And there’re a lot of negative outcomes that happen when work leaves the community. Bob Putnam and others have pointed out you get family dissolution, increases in crime, teen pregnancy, drug use, all sorts of negative effects when work leaves the community.

So we’d like – most of our policies are aimed at keeping work as an important part of our economy, finding ways to get people engaged, whether it’s re-skilling them, or things like the expanded earned income tax credit to make it worthwhile to hire people. And it may be that people work fewer hours. That was what Keynes predicted in his classic essay “Economic Possibilities for our Grandchildren,” because they have a higher overall standard of living. And that would not necessarily be a bad thing, but we don’t think it’s probably a very good idea to have masses of people who can’t find any kind of work at all.

McAfee: I just want to underscore that. I think that’s absolutely right. As we were writing this book, I became a lot more obsessed with work than with money. We included the quote from Voltaire to open one of our chapters. “Work saves us from three great evils: boredom, vice, and need.” Need is going to be the easiest of the three to take care of.

Follow James Pethokoukis on Twitter at @JimPethokoukis, and AEIdeas at @AEIdeas.

5 thoughts on “How the rise of smart machines will affect the US economy and jobs: A Q&A with Erik Brynjolfsson and Andrew McAfee

  1. I would argue that boredom and vice are both possible outcomes in the absence of need. Witness the recent CBO result RE obamacare- 2.8 million will not need to work. The only motivator for some people to do something like acquire a job skill IS being hungry (or at least a fear of..). I doubt any of those damaging the buses which carry google workers from san Francisco to silicon valley were really interested in doing something to reduce “inequality”- for instance getting an education that would allow them to be in a job that offered such a perk. Entitlements and a lack of self motivation must accept at least some of the blame for “inequality” as defined by obama. As for immigration- most of the immigrants I have met or worked with here in California are more motivated than non-immigrants. Initially many have a problem with English, but at some point their performance in school or work becomes impressive.

  2. If Erik Brynjolfsson and Andrew McAfee are right, then the CPI or PCE deflator indexes are overstating inflation.

    We are producing goods and service that are not getting measured (I add on, we have cleaned up the environment a lot too. L.A. used to be a poison gas chamber, but smog has been reduced 95 percent in the last 40 years. The additional benefit does not show up in the CPI).

    So the Fed is trying to “beat inflation” and hold it to 2 percent on the PCE, except they have it at 1 percent.

    But since the PCE probably overstates inflation, is the Fed actually obtaining deflation?

    Very possibly. That might explain why monetary asphyxiation has been the norm in recent years.

  3. Fantastic article, about a topic that very few people even know about – but is coming upon us like a tsunami.

    Fundamental changes in transportation have always yielded fundamental changes to macroeconomies. From the wheel through the steam engine through aircraft, etc. We’re all well read in this. The autonomous automobile will change the economy in ways that go beyond letting junior go to soccer without an adult.

    The concept of cities will change. The need to be in close proximity with each other will change. When I can live 60 miles from my job, and my job is in Los Angeles, and I can get there in half an hour – any time of day? That’s fundamentally changing. And that’s what autonomous cars will do. Traffic will become a thing of the past, and safe speeds of well over 100MPH will become common.

    Cities will grow for cultural reasons, but “proximity to my job” will no longer be a driving factor for living in the city. Suburbs should see a renaissance.

    Impaired driving will become a non-issue, which will change our attitudes on nightlife.

    Further improvements in communications will further reduce the need for knowledge workers to physically be “at the office,” which will further change the calculus of urban/suburban living.

    But Tyler’s hypothesis remains. The one thing that nobody here is talking about are “people who cannot become skilled workers.” This isn’t Lake Woebegone. Not everyone is above average. As manual-labor and unskilled jobs are replaced ever more by automation, what does the lowest quartile of our population do? They aren’t smart enough – there, I said it – to become knowledge workers. Fast-food is now made on a totally automated production line, from order to fulfillment. The floors are cleaned by robots. Shelves are re-stocked by robots. Etc., etc.

    What do those people do in this utopian world? Even if there is plenty of pie to go around, Maslow’s hierarchy of needs is irrefutable: if they do not have a *purpose,* they will become an unruly mob.

    Do we go back to some form of Bread and Circuses? Do we let them eat cake? What if WE are “them?” At what point does the line move such that only the top 5% of all human beings are smart enough to contribute to new production, while everything else is automated?

    This could become a very scary future, indeed…

  4. The key question is how this effects low productivity sectors. Most of the reported productivity comes from manufacturing, farming, basic resources etc. and other sectors that have limited job growth. Most employment growth has come from sectors with minimal productivity growth; consumer services (i.e. restaurants) and government (i.e. education). I suspect that the impact on consumer services will be less since more involve physical effort rather than information. Anyone who has ever tried to scrub a toilet bowl with a GPS or cellphone can attest to this. The government sector is very information dense and already past due for application of technology. However I suspect that the vested interests (i.e. public sector unions) will be more interested in a strategy of more unproductive jobs rather than fewer automated jobs.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>