Virtual Roundtable: Technology and the Economy

Tuesday, January 25, 2022
Getty Images/Yuichiro Chino
Erik Brynjolfsson

Jerry Yang and Akiko Yamazaki Professor and Senior Fellow, Stanford Institute for Human-Centered AI; Director, Stanford Digital Economy Lab

Tyler Cowen

Holbert C. Harris Professor of Economics, George Mason University


Mary and David Boies Distinguished Fellow for U.S. Foreign Policy, Council on Foreign Relations


ROSE: Hi, everybody and welcome. My name is Gideon Rose. I'm the Mary and David Boise Distinguished Fellow in US Foreign Policy here at the Council on Foreign Relations. And this project is part of the Diamonstein-Spielvogel Project on the Future of Democracy.  This round table on the future on threats to American democracy is one that basically addresses the question of how we got to the place where we are now, without necessarily characterizing exactly where we are now, because the different variations are not particularly interesting. And we've explored everything from growing in equality to continued social stratification to rising partisanship to the impact of technology on politics and social media and things like that. And today we have two brilliant economists, Erik Brynjolfsson and Tyler Cowen, to talk with us about the role of technology on the economy. There clearly is a lot of discontent in American society, a feeling that the economy is not delivering lots of good jobs, lots of people at good wages in satisfying ways, and that this... Somehow there's a feeling might have been better in the past. And let's start by asking, how much have autonomous changes in technology as opposed to policy choices, as opposed to China, how much has simply advanced tech over time changed the nature of the workforce such that there are fewer good jobs now for large numbers of people at the lower ends of the scale, and even in the middle end of the scale than there were before? Erik, why don't you start with that? And Tyler, you can bounce off whatever Erik says and explain why he's wrong.

BRYNJOLFSSON: Or maybe right. Who knows? Thank you, Gideon. It's an honor to be here in this incredible group of people you assembled, and I appreciate the chance to chat with all of you and with Tyler. So I don't want to bury the lede. I think that the top line thing is that technology makes the pie bigger and it increases productivity... Not as much the past 10 years that I would have liked, but I'm actually pretty optimistic going forward that there are a bunch of technologies in the pipeline, especially around artificial intelligence where I'm focusing a lot, but also things like the mRNA vaccines, improvements in energy, and I could go on with a lot of other areas. But there's no economic law that everybody benefits equally, or even at all from that, and unfortunately for much of the past 20 to 30 years, there's been growing inequality, and a lot of people haven't participated in those gains. You see it economically, you also see it in things like deaths from despair, where for the first time, Americans are having shorter life expectancies, so it's not just a measurement issue that's part of the problem there. And those deaths from despair, our focus very much on high school educated men and women and not so much into college side, where life expectancy is still going up. There are a lot of reasons for that... That that's happening. Part of it is that technologies have replaced a lot of routine, repetitive tasks, and machines can do them now, and when humans compete with machines, their wages tend to go down. You can also use technology in ways that augment humans. You can also do things that take that growing pie that I described and share it more evenly. So there are policy choices that you can do that would make it so that we not only have higher productivity and more economic value, but also we have it distributed in a more even way. We just haven't made those choices.

ROSE: So what I hear you saying, Tyler, is that... Sorry, what I hear you saying, Erik, is that technology allows us to produce more stuff. And the question of how we organize the distribution, the production, and all the other things associated with that is ultimately a social or a policy choice that we can then make. And so the technology itself deserves credit for the upsides and some of the unfortunate tendencies other than that but we can correct the downsides with our own policy choices. Tyler, what do you think about that? 

COWEN: Here's how I see the winners and losers from this new world. If you have very good technical skills, odds are you will do great. If you have very good skills at management or marketing, I think you'll do very, very well, maybe even better. You don't necessarily have to understand the technology, but if you can mobilize and organize the human beings who do, again, that's a very high return for you. That's a scarce skill. And then if you just own some assets that are valuable, like, say quality land in major cities. So those are three categories or groups that will do very well. That's a lot of Americans, but it's probably less than half the country, by my informal estimate. So there's a problem that results. My preferred solutions to that problem... They're only partial, but it's to get more people into those three categories: Technical skills, management/marketing skills, or just owning some valuable assets. I very much favor having a social welfare state, but those are one-off transfers. Over time, you actually want to have people being more productive in some way for them to keep pace, in terms of socially, status, connections, networks. So I don't think redistribution is really the answer here though we should have some to make it manageable. The key question is, how many people can we get into those groups? 

ROSE: This is...

BRYNJOLFSSON: And let me just say... So just to clarify, one of the big roles for government is to help get people into those groups by changing the kinds of education we do. Also, I would say on things like healthcare, there are things we can do that make it easier for people to have the kind of healthy outcomes and make it more likely to be able to be successful economically as well.

ROSE: Okay, this is brilliant. And now, given you guys, I want to go back before we pick up from where you are just right here, I want you to go back and talk a little bit about how we got here over the last couple of centuries in the following ways. In a couple of minutes, through podded histories of what happened to the workforce when agriculture shifted over to industrialization, and then what happened in the 20th century when industrialization went through its various things. Who wants to do the two-minute podded history of the American labor force participation from 1800 to 2000? 

BRYNJOLFSSON: I'll take a cut, because the first graph we have in the second machine age tries to do this a little bit. And for most of human history, there wasn't much economic growth and things were pretty miserable. Around the late 1700s, 1776, there were a number of things that happened. The wealth of nations was published. The Declaration of Independence. But perhaps most importantly, there was an improved steam engine that was developed, and that helped ignite the Industrial Revolution, and ever since then, productivity and economic growth has really taken off, growing exponentially at close to 2% per year. And when you compound that, that's why we sitting around here are 20 or 50 times wealthier than those people living back then, but it was... As I was just saying earlier, it was uneven. There were people who found that their jobs were less valuable, that machines were able to replace what they were doing, and their wages went down, even if on average wages went up over most of that time period. In each wave of general purpose technology, steam engine, electricity, and now computers and perhaps artificial intelligence, the real gains come from the add-on innovations, the way that you use those to change the way work gets done and build complementary innovations, and that often takes decades. What we saw in the case of electricity was that initially American factories were not any more productive as they electrified. It took about 20, 30, 40 years, and the main thing that drove the productivity game was the reinvention of work, the redesign of the way we do things. I think we're going through a similar period right now where some of the technologies and computerization and artificial intelligence, they've sown the seeds of some amazing productivity gains. I think the next 10 years could be some of the best ones we've ever had. But they don't come automatically. It's not going to be simply a matter of swapping out an existing person or process and putting in a machine. It's going to be more a matter of reinventing how we do work. And that takes a lot of creativity, the kind of people that Tyler mentioned, the people who can manage and re-organize work, who have the entrepreneurial insight to think of a new way an industry can be organized. Those are the ones that are going to create enormous value and often capture a big part of that as well.

ROSE: Okay, so Tyler, has that... Are we in just a constant process? Is what you just described, the shift to those groups being favored, is that just the latest twist on this longer-term process of different groups being affected by differential technologies of their era? And so is this anything new, or is it... How new is it? 

COWEN: The details are new. I very much would second Erik's narrative. If this were 1963, if you were a young male in good shape and only a high school degree, you actually would do in relative terms, much better back then than I think you would now. But I'd like to respond to a point Erik made about the role of government. I see government right now as in many ways a hindrance for people getting into the groups that are doing well. I mean, just to ask a simple question, how good a job has the government done with our last two years of K-12 schooling? To me, that seems pretty abysmal. And you have districts that refuse to measure learning loss or even admit there are any kinds of trade-offs. If you think of ownership of land as one way you might improve your lot in this growing society, nimbyism, which is governmental laws that's rampant through our country. So I'm all for government doing things to help, but I think the current margin we're at is getting government to stop doing things that hurt.

ROSE: Okay, so Tyler, let me follow that. So Erik's theory or basic approach... I'm going to characterize this just for the sake of discussion. You could use government and collective action positively to redress some of the collective problems that occur, and that can make things better. And you're saying... Those kind of interventions often underperform, don't actually do the things they're set out to do, and often sometimes are even counterproductive. What then can be done other than simply submit to market forces? 

COWEN: I think the United States federal government is reasonably competent at sending around large sums of money through the form of checks, so I understand full well that various child support plans have failed to pass political muster. But that to me is a project that at least in principle can succeed. Send people money. Money makes people better off. Our government has a long track record of sending people money and getting them the cash. But if I look at the last few years, I mean, this is a small thing, it doesn't matter, but I think it's indicative, our CDC just put out a statement telling Americans it was too dangerous for them to travel to Canada. That is literally insane. So when our government these days tries to do more active policy, I see the incentives, the dysfunction is at a higher level than I thought two and a half years ago, and I was broadly pretty Libertarian then. It seems much worse to me now. So there's a lot of remedies that sound good as a kind of parlor game, like, "Government should do this, government should do that." A lot of that is correct in the parlor game setting, but in terms of actual implementation, I worry it's going to be like the CDC telling us all not to go to Canada.

ROSE: Okay, so let me stick with you for a second to follow up here. That you said things have changed from 1960 to now, pretty significantly in a lot of these areas.

COWEN: In America, absolutely.

ROSE: Absolutely. And presumably, there may well be comparable changes over the next comparable stretch of time.

COWEN: Absolutely, highly expected.

ROSE: What will be the factors that drive those changes? 

COWEN: Here are to me two key questions. So we all have a rough sense of what the world is like now, and what skills you need to work in the tech sector, right? One possibility, but not a certainty, is simply as time passes, you'll need more and more tech skills. Another possibility is computers and artificial intelligence will become so good, they'll do a lot of the tech skills for you, and what will start to matter more are the people skills. Now I'm quite sure I don't know which of those scenarios is gonna come to pass, but I think it's a really important question. Another question which we're seeing acted out before us is work from a distance. How much will that mean that employers will just hire the best talent around the world? And maybe... Maybe. I use that word on purpose. It's American upper middle class white collar workers who will be hurt the most, or it could just be the office setting is so important, there'll be a lot of foreign hires work from a distance at various margins, just as we've had programmers from Bangalore for a long time, and it won't fundamentally change the country. Again, there I am not sure which scenario will happen.

ROSE: Is there any scenario in which more Americans of the currently disfavored classes will do better over the next generation or two? 

COWEN: Well, I think the last waves of technological innovation have really been pretty good for the quite poor individuals. It's the middle class that's been hurt. But a lot of things are much cheaper. Go to Walmart, go to a dollar store. Internet, yes, you need to buy the smartphone or get a good connection. But once you have it, there's so much cheap, fun entertainment, learning instruction than there was before. So I think of the last wave, including the China shock, but more of it coming from automation, hurting the US middle class. Maybe my modal prediction will be the next wave will mostly hurt the US upper middle class, and the middle class will make somewhat of a comeback, but through the indignity of some upper and middle class workers balancing down into it. I think the poor have been doing pretty well from recent technologies, or if not, the problem is not the technology.

ROSE: Erik, over to you, to the question about what will drive the changes over the next 60-70 years? 

BRYNJOLFSSON: 60-70. Well, number one has to be technology. It certainly has been in the past, and going forward, I see in the pipeline even more remarkable... I mean, when I talk about these general purpose technologies, it's very hard for me to think of anything more fundamental than intelligence. And if we are on the verge of having machine intelligence, that's about as revolutionary as it gets, and people, if they're around centuries from now, will look back at our generation when this is happening, as being at a real interesting turning point.

ROSE: Are we at the singularity, or approaching a singularity or is that something...

BRYNJOLFSSON: Not a singularity. You can have pretty revolutionary change without it being a singularity. So maybe someday. I'm not ruling out that at some point, we will get to that point. That's not what we're facing right now, our generation, I don't think. I just want to say something a little bit... But I agree very much with Tyler, that there's been remarkable decline and dysfunction in government. Actually, in a lot of other areas of America as well. It's just sort of sad to see what we used to be able to do 50 or 70 years ago, in terms of the building the roads and bridges, the R&D investment and the K-12, that have all gotten so much worse lately. Not just the CDC, but the cost of building a subway or the education dysfunction that Tyler mentioned. But my instinct is to say, "Well, how can we fix that? Because we know that it can do better," and not to abandon it and say, "Okay, well, I guess we're going to give up on that."

My colleague here at Stanford, Rick Hanushek at the Hoover Institution, estimated that having a K-12 teacher in the top quartile of the distribution of performance versus the bottom quartile adds about $400000 per year in value to each class of 20, 30 students. So there's room to do enormously better than what we're doing right now. For a number of reasons, we're not doing that in K-12. Our roads and bridges cost much more to produce. So I think if we could try to fix that, if we could figure out... And I'm not a political scientist, but when I look at it, the dysfunction has gotten worse as we've eroded the capability. The CDC used to be acknowledged, not that long ago, as by far the world's best entity. They would never look to WHO for advice. Everyone looked to the CDC. I think in some ways, these programs have been systematically gutted, or there's been some other issues that have been going on. Their politicization that have tremendously eroded it. Tyler uses this phrase, state capacity libertarianism, that I think we need more state capacity. We should not be abandoning it.

ROSE: So let me... The question of... Okay, the best way to get into the kind of stuff I'm talking about... Both of you, how have your views on the stuff we've just been talking about changed during the course of this century? If I had asked you those same questions 20 years ago, would you have given the same answers? 

BRYNJOLFSSON: No. I mean, as I said, I've seen a remarkable decline in...

ROSE: No, not about the government. I mean about the economics, the technology, and the economy stuff, and the drivers of things. One of the things that's interesting to me is the production knowledge. We talked about the science in the pandemic. At the same time, we've had some of the greatest advances in modern science with the vaccines, in the speed and the quality and the impact. We've also had some of the greatest screw-ups in governmental failure and public messaging failure, and we've had at the same time the spread of such skepticism and ignorance and weaponized disinformation that we now have, even in this time of great scientific progress, less trust in science and experts than ever before. So it seems like a very weird mix of things, and I find it interesting how it changed.

Our views about things like economic policy are based in large part on our theories about the role different factors make. Exactly the kind of stuff we're talking about. What can be variable? What can change? And I'm struck by the fact that things like out of the blue comets, whether they are in the movies, or whether they're Al-Qaeda, or whether they're Donald Trump, seem to come and disrupt our politics and life in ways that we didn't predict, and that the experts then find themselves, like stock market analysts, retrospectively fitting their analysis and do better. I guess my question to you guys is, what have you learned that you no longer feel is right, but you now feel, "This is my... " Or have you had the same views on these things throughout your careers? 

COWEN: Here are some views I've revised over the last 20 years. 20 years ago, I was not sure Moore's Law would continue. I wasn't sure it wouldn't, but look, we didn't know. Mostly it has. It may be slowing now, but we've had close to 20 years of Moore's Law. That's made a huge difference. I've updated many views. Internet tech, far more powerful than my stochastic prediction would have been 20 years ago. I think now with the vaccines, but not only, we're on the verge of massive breakthroughs in biomedicine, mostly stemming in the broad sense from easier access to computation for scientists, and the democratization of scientific communication using in part the internet. So I think that will have a run of decades. I might be too old to benefit from it, but it will change our world in fundamental ways and 20 years ago... Again, I thought it might happen, but I would say now, we definitely see that it is happening. And the bureaucratization of our country has gone... I expected it to go badly, actually, but it's gone somewhat worse than I was expecting. Those would be some of my major revisions.

ROSE: Erik, what about you? 

BRYNJOLFSSON: Well, mine's frighteningly close to Tyler's. In the broad scope, I've been sort of optimistic about the technology. The specific instantiation about what's happening with artificial intelligence or mRNA vaccines or fusion power, I couldn't have predicted those specific ones, but I saw a lot of pent-up opportunity and the fundamentals have been there, and I think by and large that has panned out in ways, and there's still a lot of things there. On the government and the political and the societal dysfunction, that's been worse than... Although I wasn't super optimistic before. I actually wrote a paper about what we called cyberbalkanization in 1996, if I remember correctly, where I looked at how the Internet could lead to different groups finding each other, and maybe some of them learn about wine or political science and learn from each other, or maybe some of them learn less valuable or less productive types of ways of working together that led to dysfunction. So that's certainly happened as well.

ROSE: Will the technological advances... This is to both of you... That we can expect, and the building on the ones that come from future developments themselves, will they be enough to save us from the human problems that we do in using them, in the same way that the vaccines will ultimately solve us from the idiots who don't want to get vaccinated, or the... Will the energy technology development save us from global warming, because policy choices aren't going to? 

BRYNJOLFSSON: Well, actually, the energy ones, I'm somewhat optimistic about. Maybe some of the vaccines. But the broader question, I think it's more or less the opposite, that the more powerful the technologies become, the more important our values become, the more important our governance becomes, because most technologies can be used in ways that are very beneficial or that are very destructive. If you don't have a technology that can make much of a dent in the world, it doesn't matter that much what your values are and what your governance is, or what your systems are. But if you have a technology that can create vast abundance and freedom, or you can have one that can be dystopian and destructive, then it matters a lot. And right now unfortunately, I think that, as we said, the governance is working worse, and the more powerful the technologies become the bigger the problem that is.

ROSE: Oh, that's an interesting answer. Okay, Tyler, over to you for that question.

COWEN: If I look at the broader span of human history, it seems to me that either being conquered or attacked is by far the major big problem societies face. So all the new tech is wonderful, but look, those are also weapons. So simply cyber-attacks as one example, but far from the only example. Imagine drone attacks on American citizens. I don't feel confident giving you predictions, but I see generally in history, new technologies get used as weapons. This will be a massive challenge. I would say on average, I worry much less about particular disruptions in the social and economic fabric. There are very real costs to those who suffer them, but the big picture issue is, can we actually continue as a country and maintain public order? I never take that for granted. And the better new technology is, the more of a challenge it will be. You know, I'm not at all sure lab-leak hypothesis is correct. I would put it below 50%. It's not impossible. Serious people believe it. Even if it didn't happen this time, it could happen in a far worse way, and you go on down the list for all the wonderful things we've done, there are corresponding national security dangers, and those are by far my biggest worries.

ROSE: Okay. That's a nice cheery thought. Let me take you to... Okay, we see in the tech sector not some libertarian paradise of Silicon Valley in the early '90s or whatever, if there ever was that. But a giant oligopolistic sector with elephants protecting their turf, beating up each other and whatever, and trampling over the rest of us. Is the tech sector operating according to free market principles or oligopolistic... What is the way to understand the role of the tech companies in our society and economy today? 

COWEN: A lot of the words you used, I'm not sure what you meant by them.

ROSE: That's what my economist friends say.


COWEN: Our tech sector gives us a lot of products very cheaply. You have large players, but they're large because they're giving consumers what they want. I think it's a much better arrangement than our media or intellectual class will admit. I think most ordinary Americans see this. It's why the companies got so large, and I'm far, far less worried about those issues than other intellectuals I know. Far less worried.

ROSE: Erik, what about you? 

BRYNJOLFSSON: Well, Tyler's absolutely right that there's been tremendous consumer surplus created by these technologies. It doesn't show up in GDP. We created a new measure we call GDP-B, because we want to measure the benefits, not the cost. So when things have zero price, like many of these digital goods... People spend over six hours a day on consuming digital goods on average. Most of that doesn't show up in our GDP, and so that's a huge benefit. We need new metrics to measure it better. But I don't think that the reason they became big is simply because they create a lot of consumer surplus. There's an underlying economics from the digital world, where you have high fixed costs and low marginal costs, where you have network effects, and where you have other economies of scale on the knowledge production side. All of those tend to lead towards winner-take-all or winner-take-most markets that are pretty different than the atomistic competition that Adam Smith described in the Wealth of Nations that I mentioned earlier, for agriculture or a lot of small manufacturing.

And so when you have these different underlying economics, you tend not to have the kind of competition that you get from having thousands or millions of competitors. Competition is great because it leads to most of the benefits being swept to consumers, and it leads to a lot of other benefits in society, in terms of decentralization of power. So if you have underlying economics that tend to be more towards concentration, then you can't just sit back and rely on the first fundamental welfare theorem of economics that everything is going to pan out. You have to think more carefully how you can get some of the benefits from the underlying technologies, and it won't happen quite as automatically as it did when you had more atomistic competitors.

ROSE: Are we in another period like a century and change ago, in which there's a legitimate case that given the trend towards concentration you just talked about, we need a new kind of trust busting? 

BRYNJOLFSSON: We need to rethink our antitrust policy. Unfortunately, it's not as simple as just breaking things up, because most of the benefits do come from these scale economies. So if you divided up, I don't know, Facebook, and you had, like, okay, one-quarter of the people on each network and they couldn't talk to each other, that wouldn't be very beneficial. Now, maybe there are things you can do by forcing interoperability. I remember... I don't know, it was probably about a decade ago or more, phone companies wouldn't let you take the telephone number from one company to another. They said it was just too hard, too complicated, their servers would melt or something. Ultimately, the government had to require them to make those phone numbers portable because the private incentives were not matched to the public incentives, and there may be things like that that we can do to require information sharing or portability, but it's not going to be as simple as what we did with the phone number portability. It's a pretty messy problem, and unfortunately, for the past couple of hundred years, we've benefited from some economics where market competition naturally led to a lot of gains... Widely dispersed gains. Market competition doesn't necessarily work as automatically in a world where you have only one or two major competitors in each industry.

COWEN: Erik, I think you're setting up the straw man with this notion of atomistic competition. It's never been atomistic. We're on a Zoom call right now, like we kind of all figured this event would be on Zoom. In that sense, Zoom has a somewhat dominant position. But our audience, they could be listening to people on YouTube, they could be doing something on Facebook, they could be on Twitter, they could be on YouTube, they could be on Discord, they could be out there gaming. The number of competitors for our time and attention on the internet, in social media space to hear output, read output... It's a very high number, and I've just gone through a few of the options. So I don't think it's perfect competition. I don't think we've ever had that outside of some wheat farmers somewhere. I think it's working as well as we could plausibly expect it to. Look at TikTok. It's a new company, came out of nowhere, had a better product in the eyes of some people, drew eyeballs away from Facebook, other places. They're doing great.

ROSE: So Tyler, let me press you on that. I appreciate that. And does that world... Does that lead you to have a mentality that we don't have to think of every problem needing an interventionist solution, because sometimes the normal course of affairs will either be less problematic, or will take care of the problem for us? 

COWEN: That is in general my opinion about a lot of internet services, yes.

ROSE: And is that true for the broader trends in how a polity should react to a changing economy and technology the way we talked about before? Should the response of the polity collectively be, "Okay, these changes are happening over multiple generations. Let's just basically ride them out and see what happens."

COWEN: Well, that's not exactly my view. So as I said before, I'm extremely worried about cyber-attacks, which hit hospitals, schools, a lot of soft targets, and then they ransom your data for crypto. Or you might have some cyber attackers who just want to put down institutions, so I absolutely think we need more of a collective governmental response to that. That said, I'm not sure exactly what is the correct policy. So I'm not calling for laissez-faire. I'm saying, let's look at what are the real actual threats? And I think cyber-attacks are by far number one. Some notion that, well, someone can track whatever... People for better or worse, they don't care about their online privacy that much.

ROSE: Cyber-attacks by whom? 

COWEN: I wish they would care more. I don't think we're gonna change that. To me, cyber-attacks, by far the number one issue requiring government to do more.

ROSE: Cyber by whom? Cyber-attacks by whom? 

COWEN: Russia, Iran, North Korea, Trump's thing about the guy in the basement on the sofa or whatever. We have some idea who's doing them, but there are so many potential attackers. It's one reason it's a hard problem to solve.

ROSE: Got it. We have lots and lots of people here who have interesting things they might want to contribute and ask, so I will at this point ask our participants and members to join the conversation if they would like. Let's raise their hands, and I'll turn it over to our tech people, and if not, I will continue. I have lots of things that we can continue talking about soon.

SPEAKER: We will take our first question from Audrey Kurth-Cronin.

CRONIN: Hello, this is Audrey Kurth Cronin from American University. It's a great conversation thus far. I'd like to pick up on the theme that Tyler mentioned about the national security implications of some of the technologies. And looking back over the history of the Industrial Revolutions, there's a tendency always to leave out major wars when you look at, for example, the second Industrial Revolution from 1860 to 1914, and the third Industrial Revolution is always picked up in the 1960s or even the 1970s to the year 2000, leaving out the First World War, the Second World War, and all the advances that occurred with Alan Turing and computers and all of that during the war. So my question is why are military innovation and commercial innovation artificially so separated? And doesn't that increase the deliberate ignorance of the commercial tech sector to the national security implications of their technologies? 

ROSE: And a two finger tag-on to that. If the great argument against that is DARPA and the internet, why have there only been a few of those rather than a continuing current, big stream of those things? Either one. Erik or Tyler.

COWEN: I find it striking that, say, by the standards of the 1970s... And I'm not saying those are the correct standards. I get that. But by those standards, the '60s, the '70s, we are at war right now with China and Russia. They basically attack us every day, and we put up with it. I'm sure we attack them plenty. I get that. We may even be attacking them more than they attack us. But it's not obvious to me that this is a stable game, so to speak, in the Thomas Shelling sense, that you have more parties enter the market, some attacks at least are hard to trace. Once we treat attacks as not an act of war, there's steady escalation, and maybe voters will support the notion that somewhere you draw a line and call it an actual act of war.

I don't know. I don't know how that equilibrium looks, but what I see so far, I'm not entirely encouraged by it. And I agree, we talk about these things separately, but they're not. You can see so many newspaper stories. Crime rates are down. It's actually not true for the last two years, but pre-COVID, crime rates mostly down. But if you count cyber crime, I don't know how to aggregate that into the totals, but in some sectors, crime rates are radically up. And that's what I want people to take more seriously. And it's also important to get them to appreciate technology. Just as atomic bombs, I would say irrationally turned people against nuclear power, we face the risk that cyber and other kinds of attacks will turn people against highly beneficial forms of tech in the same manner.

ROSE: Okay, Erik, do you want to add? 

BRYNJOLFSSON: Well, I think that there's not as much of an artificial boundary... Well, first, I appreciate your question, Audrey, but I see that there's actually quite a bit of dual-use technology between the Defense Department and other areas, and areas in cyber and a lot of other AI, it's becoming more and more clear that that's the case. And that does lead to some problematic issues, when we want to cooperate with people internationally in the area of artificial intelligence. It has become a damper on a lot of research efforts because of the very real and legitimate concern that China in particular would use some of those technologies in ways that would be militarized and be harmful. So there isn't such a sharp line between military and non-military technologies, especially in some of these more advanced areas of artificial intelligence.

ROSE: Actually, you know, it's funny, I'll just jump in here. I think that we grew up, all of us, in an era in which security studies was one sphere and international political economy was another sphere, and I now think that was a complete artifact of an era in which we had a security relationship with poor countries, and an economic relationship with richer countries, and our friends were trading partners, and our enemies were poor and we could exclude from the economy. Now... And the fields grew up not even thinking about each other but they were whole, separate sub-fields. But now...

BRYNJOLFSSON: The whole Russian scientific establishment was pretty much separate from the United States in a way that the Chinese research establishment is not.

ROSE: But now that our chief trading partner... And now that the chief economic player and the chief security player are the same, we have no good intellectual tools to manage a relationship with both of those. Actually, Tyler, I'm curious. Do you think we have good intellectual tools to manage a relationship with another actor, China, that is both a major economic competitor and partner and a major security threat? We have never faced that.

COWEN: I think American academics are too compartmentalized often to think usefully about China. A lot of our universities at least were too dependent on Chinese students. There's a whole set of issues. Did universities feel the need, as they were legally obliged to do, report Chinese governmental grants? And they didn't. So there's been a lot of people and institutions asleep at the wheel because it was very convenient to let this issue go away. We need to do much better than just get people to report the monies they're supposed to report. We need to actually have America's leading thinkers proactively engaging with this in the same serious way that they did with World War II and the Cold War. I just think we're in a much worse position to do that now because of our own weaknesses in academic life.

ROSE: Okay. Let's go back to the queue.

SPEAKER: Our next question is from Fareed Zakaria.

ZAKARIA: Hey guys. I wanted to go back to an area of happy agreement between the two of you right at the start, which was that there isn't any real problem with this technological revolution, and it probably will look a lot like all the other ones and will not displace... Will not result in large-scale unemployment. And I guess I want to just press both of you on it, probably Tyler first, because you are more of a sunny, optimistic Libertarian on this issue. When I look at it, it does feel like, through some combination of technological change and globalization, it has become much harder for non-skilled labor, that is not college-educated workers, to have pricing power. Some of them get jobs fine and things like that. But when you look at Henry Ford, Henry Ford could only become a billionaire... Let's assume he was a billionaire in today's terms, by having products that were made by tens and tens, if not hundreds of thousands of workers, when you add in all the affiliated people who make a car. Elon Musk barely needs anyone. There is a disconnect between capital and labor, and you see it in the Feds data, where you see the number of people at the top 20% moving up versus the bottom 20% moving down even in this pandemic.

You see it when you look at obviously stagnant wages. You see it in Nicholas Eberstadt's work on staggeringly low labor participation, which has been declining. And I'm just wondering, what do you make of that reality? That when you take the top 20 companies in the world today, they barely employ anyone by historical standards? And when you looked at it 50 or 100 years ago, these places employed vast numbers of people in comparison. And so it feels to me like, yes, there are a lot of low quality, low wage jobs, and the wages are going up a little bit. But the core of what made 20th century capital stable after the Second World War was this vast number of jobs that were created, good jobs and good wages for non-college educated, blue-collar workers. Still 75% of Americans don't get college degrees.

So what is the prognosis for them? Tyler, you seem dangerously close to giving them the old Roman emperors formula of Bread and Circus by saying, "Well, you know, they get... As long as they have a good internet connection, they can watch lots of great, cheap entertainment, and the cost of food and clothing has gone down thanks to China." But how do they have pricing power as labor in a world in which Google needs to employ 50000 people, Facebook needs to employ 50000 people? I don't even know combined how many Tesla employees. So it feels to me there's a bigger problem here that you guys were happily assuming that, in the long run, it would all work out.

COWEN: Well, I'm worried, but I think my worry is slightly different than yours. So I think many things will get a lot cheaper for poor people, except for rent. So I'm very worried about their rent. I think we need to strongly de-regulate that sector. I'm not sure we will. But there is a definite remedy that we could do, if we wanted to. And a factor that protects those lower income classes, sometimes called unskilled labor, is they do a lot of face-to-face service jobs, and many of those cannot be automated, or cannot be automated soon. So I'm not sure they will see very heavy downward wage pressure over the next 20 years.

By definition, the poorer people are not doing the best. But I am, I would say, more worried about the middle class to upper middle class, which has more political power, which has higher expectations, and if you've ever played around with something like GPT3, or seen what artificial intelligence is possibly on the verge of being able to do, the notion that you have this new class of Americans who feel they are not unskilled labor, but in essence, in markets will be treated as unskilled labor, and de facto will actually be unskilled labor compared to the machines. Their potential to wreak political havoc frankly worries me more than the lower classes. I think the backlash from the middle to upper middle classes could do more to hurt poorer people than the machines themselves will do. I think the machines will help poorer people by making things cheaper. So I'm definitely worried about the political economy of all this, but I distribute my worries in a somewhat different way.

ROSE: So skilled labor will be fully vaccinated. The definition will change over time as new things come along. Oh wow, that's great. Okay, Erik, do you want to jump in on that one? To Fareed's...

BRYNJOLFSSON: Sure, sure. Fareed, I very much share your concerns. I'm sorry if I gave you the impression that I didn't. I would nuance it a bit. I don't think the concern is so much on employment, and I don't think we're going to see mass unemployment. When we look at the data, even with advanced machine learning, we have a paper with Tom Mitchell and Daniel Rock that tried to look at what jobs and tasks were suitable for machine learning, and there was no occupation where machine learning ran the table, although they can do many specific tasks. But... There's a huge but. The real issue is around wages, and as demand declines in certain categories, you do see a lot of downward pressure on wages. And as I think you've both said, median wages have been stagnating... Actually the past few years, they've gone up a bit, like the past few years. But if you look at, say, the 20 or 30 year time period, depending on how you adjust for inflation, the median worker has barely moved at all, and that's a function initially of less skilled workers in the early part of that time period, and more recently middle skilled workers getting hit pretty hard, and that will lead to a lot of political dysfunction as people see that they're not sharing in the benefits of these new technologies.

It's not necessarily there's going to be mass unemployment, but if the gains are not widely distributed, then a lot of people are going to turn against some of the technologies, not just because of Tyler's point about cyber, but because they're not sharing in the economic benefits. And we saw that already with international trade, which I think was a very parallel situation, where economists said, "Hey, international trade makes the pie bigger. There's gains from that. The country gets richer. Of course, there are winners and losers, but you can compensate that and make sure that everyone or almost everyone's better off." We did the first half of that bargain and we reneged on the second half of that bargain, and that got a lot of people angry. Maybe justifiably angry. And so there's been a big backlash, where both parties are putting up trade barriers and making it more difficult to get those gains from trade. It wouldn't surprise me if we start seeing more and more of a similar story with tech, and we are seeing it with a bit of a tech backlash.

ROSE: Like a neo-Ludditism? 

BRYNJOLFSSON: In many areas, yeah. There's a lot of things that people are upset about big tech for. And with good reason. And I think for tech folks to downplay those too much is going to ultimately backfire. I think we need to acknowledge them really clearly, and think about how we can make sure that they don't lead to misinformation and disinformation. They don't lead to people having their wages go down. They don't lead to some of the other concerns that people have. I think it's doable. I'm kind of a tech optimist, and I think that, as I said, the pie's getting bigger and we can use this to have more easy spread of truth and information, but it's far from automatic, and my main message is that none of these outcomes are baked in, that we can just sit back and watch it unfold in front of us and think, "Okay, it's great." I think it takes a lot of hard work, the reason that America was very successful in the middle of the 20th century was a bunch of people worked really hard to help guide it and steer it in ways, and prevent some of the bad outcomes. Put in place pollution regulations and did other things that they saw would benefit society more broadly. Somewhere along the way, the focus on me rather than we became very important, and the outcome has been that we have a lot of dysfunction, I think.

ROSE: Back to the queue.

SPEAKER: Our next question from Jan Lodal.

LODAL: Am I unmuted now? 

ROSE: Yes, you are.

LODAL: Hi. Thanks. Thank you guys very much. I wanted to go back to the question about cyber, 'cause I agree with Tyler. I think it's a very big deal. And I have for a long time been an advocate that it's actually fairly easy to fix that problem, if we just had the willingness. But of course, you guys are pointing out that's true for a lot of these problems. But in particular that if you simply change the internet a little bit, and with IPv6 it's pretty easy to require an authentication token in every packet, and then you build a better key management infrastructure to make sure you knew who was who. Not trivial, but workable. That for a pretty low cost, you could actually kind of eliminate cyber attacks. But for a long time... This was proposed when I was in the government 20 years ago by a bunch of us, and it got knocked out on political grounds. Got pretty far. To the president, actually. But it still seems to me to be the solution, and it seems to me that we have to do something different along the lines of what Tyler was warning if we really want to keep going in the direction we're going.

COWEN: I think it's more likely we just spend 20 years toughening up our soft targets, paying increasing ransoms along the way, maybe having some cities or states with power knocked out for long periods of time in major social and humanitarian problems. And at the end of some decades-long bitter process, learn piece by piece in a very decentralized way. I don't know enough to evaluate your proposal, but my sense is the political economy of our nation is not well geared to making things like that happen.

BRYNJOLFSSON: Yeah, I want to share both your guys' concern about that. I'd mentioned earlier, I've been on boards of directors where you get the most scary reports about the vulnerability of the company, and I don't know why peoples' hair is on fire trying to address that. I also agree that we could do so much more with public key cryptography and authentication. I remember back again in the '90s, I was teaching this to my MBA students, these technologies that have been developed soon will all be authenticated in all of our communications. I was terribly wrong, and it's not because the technology wasn't there a couple of decades ago. There are some other issues. I don't know whether there's some forces trying to prevent it from happening, or I don't totally understand why we don't have a more secure, authenticated infrastructure. It seems like it would be very beneficial. I wouldn't go as far as saying it would solve all the problems, but it would go a long way, and we could do a lot more in that direction. I hope that people who are on this call listening will take actions and try to help implement those kinds of technologies to give us an additional modicum of security.

ROSE: Back to the queue.

SPEAKER: We'll take our next question by Donna Tuths.

TUTHS: Hi, thank you so much. I want to ask a question, and I want to start with the premise, which I think you both agreed. Look, tech can be good or tech can be bad. And therefore it can have an upside and it can have a downside to it. And I think if we look at simple upsides, we look to cloud computing, which essentially is what allowed an unprecedented percent of this economy to go remote. Nobody would have imagined that, I think, in the past. But at the same time, you're talking about concerns about cyber security, etcetera. And so I think if we think it can be good or bad, then we have to agree that we need some entities to come in and ensure that we are actually benefiting from the good, and so whether we like government or we don't like government, I think we have to agree that they have a role here, and I guess my question to you is how much do you think technology fluency amongst our policy makers is holding us back? And I ask this question because even Tyler, with your point about people aren't outraged enough about data privacy, look, people were outraged about the Snowden affair, and yet they didn't seem to care that Google had been flying over their homes for the last several years... Not with the anonymized data. Publishing pictures and their address. People don't understand this stuff. So how much is that hampering us in order to kinda make progress here, and what can we do about it? 

COWEN: If I look at the ongoing attempts to regulate crypto. And we can all agree some version of this will happen, no matter what we may or may not prefer, I don't see nearly enough policy makers who even begin to understand what they're talking about, and I don't know how to fix that. And I think there's a more general, deeper problem. Our regulatory structures, they're pretty well set up to regulate intermediaries like banks, companies. They're very poorly set up to regulate software, which often is not transparent, and maybe that's the way you're framing the forthcoming problem. How will we figure out how to regulate software, per se? Very tough issue. I don't feel we're close to doing a good job on that.

BRYNJOLFSSON: And Donna, to one of your specific questions, do our leaders have enough technical expertise? Unequivocally, not even close. One of the saddest days I've had recently was when I went to testify in Congress about artificial intelligence. [chuckle] And not just the level of questions and what they want to talk about, but the lack of interest in genuinely understanding and grappling with what was going on was just really frightening and disappointing. And as we've been talking, these technologies are becoming more and more powerful and influential, and you can't be governing if you don't really understand them. So I was really, really worried. One ray of light that same day, I talked to the Federal Reserve Board, and they were very engaged, and comparatively were quite on the ball.

They read the papers and they were quite familiar with some of the issues. I don't know what that says about democracy or governance. Again, I'm not a political scientist. I don't understand. But to me, it's kind of a sharp contrast between two of our governing institutions, and I guess insofar as Congress has Article One power, it made me kind of worried that those are the ones that ultimately make the decision. We need to do a better job. I'll put in a plug for the Human-Centered AI Initiative that I joined here at Stanford. We have a program where we're trying to teach some of the congressional staff, and even some of the Congress people about these issues. But I think it's just a small step towards what needs to happen.

ROSE: Okay, well, I'll just ask you... Guys, I'll take one last question, and I'm going to wrap up. And Matt says, we've talked about American decline in this period. In the same period we've been talking about, China has just gone from nowhere to gangbusters, even to the point of where when I started my career, we were worrying about large numbers of poorly-clad and shod foot soldiers, and now we're worrying about Chinese AI going to take over the world. Is China an example of almost a reverse example of the US, which is that you can go from nowhere to the top of the game through your own efforts and writing surfing trends successfully, even as other countries screw things up? 

COWEN: That's a big and complicated question. I guess I'd like to close just by putting in my plug for optimism. All the problems the world may have... And we've talked about some of them, not all. The title of my next book, it's called Talent, due out in May. Talent. And if you ask the simple question, "Does the world today, our country today, have more talent than before?" It's not even close. We have much, much more talent, and that good news outweighs all the other bad news you've heard this hour. That's my final take on everything.

ROSE: Erik, your final take.

BRYNJOLFSSON: It's a good take. We do have hundreds of trillions of dollars worth of human capital is one way of measuring it, and that's cause for optimism, and amazing technologies. But as I said earlier, those can be used in ways that are unimaginably beneficial or in ways that are frighteningly destructive, and the next 10 years could be easily the best 10 years we've ever seen, but they could also be perhaps the worst. I don't think either outcome is baked in. Whether we're talking about China or the United States, to briefly tag your question. And so my concern is that we have to think harder about our values and how we want to direct what we want in society, and not be passive about what the outcomes are. I think the stakes are really, really high, and the fact that the human talent and the technological capabilities are so much greater now than they ever have been before means that our decisions are that much more consequential.

ROSE: On that note, Erik, thank you. Tyler, thank you. Members of our group, thank you. This has been a great discussion, and let's hope that the optimists are right. [chuckle] Take care, everybody.

BRYNJOLFSSON: Absolutely a pleasure.

COWEN: Thank you.

SPEAKER: For more event audio, subscribe on iTunes, or visit us at cfr.org.

Top Stories on CFR


This interactive examines how nationwide bans on menthol cigarettes and flavored cigars, as proposed by the Biden administration on April 28, 2022, could help shrink the racial gap on U.S. lung cancer death rates.


Sheila Smith, the John E. Merow senior fellow for Asia-Pacific studies at the Council, sits down with James M. Lindsay to discuss the reasoning behind Japan’s new defense strategy and the Japanese government’s decision to double defense spending.

United States

In addition to minority communities and those on the political left, far-right and white supremacist extremism threatens violence against institutions conservatives cherish as well, such as the U.S. military.