Loading summary
Transcript24 lines
- [00:00]
David Autor
So I give you a simple answer to the question, have we learned anything? And the answer is no. I mean, our political system has learned nothing. In fact, to the Greece learn anything, it's learned the wrong lessons.
- [00:10]
Jasia Munk
And now the good fight with Jasia Monk. Today's conversation is all about the past, present, and future of work. What kinds of opportunities are available to average people in the job market today? Was the economic pessimism which characterized public discourse in the 2010s assuming that returns to labor were going to be much lower than returns to capital justified, or were we actually overly pessimistic in our reading of that situation? Has the China shock, the shock from China entering the World Trade Organization lastingly led to economic carnage in the American Midwest? Or has a lot of the American workforce recovered since then? And perhaps most importantly, how is artificial intelligence going to play into all of these questions? Is artificial intelligence a systematically different technology from ones that humans have faced before, or are all the kind of mechanisms that we've seen in the past, like the process of job reinstatement, where old jobs go away but new jobs take their place, going to play out? Well, the best person to speak about this is David Autor. He's the Ford professor of Economics at MIT and perhaps the world's most prominent labor economist. We talked about all of these questions in great detail, and in the last part of this conversation, we also considered how we can ensure a better future for ordinary workers. David has recently argued in a paper with Darren Atsamoglu and Simon Johnson that the future of the labor market doesn't just depend on the way in which AI develops. It depends on the policy choices we make now, which can ensure that new technologies will complement rather than replace human skills. To learn what kind of policies are necessary to assure that more positive outcome, and to listen to the full conversation without ads and without interruptions, and to support our work, Please go to writing.yashamonk.com Listen, become a paying subscriber and set up your private podcast feed. Go to writing.com listen. David Autour, welcome to the podcast.
- [02:57]
David Autor
Thank you very much. Pleasure to be here.
- [02:59]
Jasia Munk
I really look forward to this conversation. I was thinking, in preparation for this conversation, about how the broader shape of economic conversation has changed over the course of the last 15 or 20 or so years. And it strikes me that there was a period of deep pessimism in the 2010s, which came from Thomas Piketty's work about rising inequality, which came from Banko Milanovic's famous elephant curve, which seemed to suggest that the very rich globally gaining most from globalization, but a lot of a kind of global middle class isn't gaining that much. But it also came from your work about the China shock and about the kind of declining middle of the American job market. And 10 or 15 years on it feels, and all of this is before we get to the subject of AI like there's a little bit more optimism. You know, Piketty's work has been critiqued quite widely. Milanovic has updated his chart and it seems to be much more positive. It seems to show much more broad based based gains from globalization. And you published an interesting paper a few years ago saying that at least since the COVID pandemic, sort of less affluent American workers have actually done comparatively better than more affluent American workers. Was the pessimism of 2010s a mistake or what of it remains?
- [04:19]
David Autor
I don't think it was a mistake, but it's good that there's positive news as well. I mean something we learned something. The work that I did with Gordon Hanson, David Dorner on the China trade shock really showed, showed how scarring rapid labor market change can be when it's loss of critical sectors or loss of kind of career jobs. What has been very positive since around 2015, at least in the United States has been relatively robust wage growth in the bottom half of the income wage distribution. And this has been pretty pervasive. And it started before, but then it really took off after then. And in the United States a lot of what was going on before the pandemic actually was rising minimum wage laws across many states. We didn't actually see much wage compression outside of states that didn't raise their minimum wages. But then starting with the pandemic actually it was just across the board tightening and that's been really dramatic. So that's positive. And I think a lot of that has to do with running tight monetary policy and having demographic tightening, real labor scarcity that has contributed to that. Now I simultaneously think there's many things that remain concerning even before we get to the present era in terms of a lot of the job growth among low non college workers is not in stable career jobs. Much of it is in know kind of hourly service jobs that are comparatively low paid and not very economically secure and don't have high lifetime returns to specialized skills and expertise. Certainly you know, growing income concentration has been with us this entire time. And I think the sort of economic, the China trade shock in its the form that David and Gordon and I studied it, you know, was a very concentrated period of time. And that ran its course in its way. But the competition from China now is actually much more significant and not just about kind of jobs making commodity products, but really in kind of the core leadership sectors of technology that have both civilian and military applications and will affect the prosperity of the United States, of Western countries, of democracies very broadly.
- [07:00]
Jasia Munk
So let's hit a couple of those points. First, we were talking about how scarring the China shock was. There's two kinds of elements here. One is that economists say that there's gains from free trade and of course there's winners and losers. But that's okay because since there's an aggregate gain, you can redistribute some of those gains from the winners to the losers. But in practice, of course, redistributing those gains to the losers is very hard and then tends not to happen. The second point is that even if in some material sense for losers might be made whole, even if we did find mechanisms whereby the guy with a union job in a car factory, who's now unemployed, is given sufficiently enough money that he doesn't have a shortfall in income, which seems unlikely, there's probably a psychic scar, there's probably a sense of no longer being necessary and so on that remains. To what extent do you think that those genuine policy mistakes which explain the China shock and made it as bad as it was with all the political downstream consequences it had, and to what extent was it inevitable? To what extent was it good and necessary to get China into the global economy? Would forms of automation perhaps have taken some of those jobs if the China shock hadn't? You know, with the benefit of hindsight, were the real mistakes made there or were some of those disruptions inevitable?
- [08:21]
David Autor
Wow, okay. There are so many questions in there. I feel like I should be like Claude here, like giving myself a to do list of in responding to all those things. So first of all, let me address first point about the nature of the losses. And I think you're completely right to distinguish between pecuniary losses, lower income and broader psychic scars about identity, about status, about purpose. And I think both of those are first order and many people who are losing manufacturing jobs in the China trade exposed areas. And just in case your listeners aren't aware, the China trade shock, you know, largely accompanying China's succession to WTO in 2001, you know, led to a huge contraction of US manufacturing. You know, about 4 million jobs were lost in the course of seven years. Not all that is due to the China trade shock. But an important part of it was, and that's not a huge number of jobs in the scale of an economy of 155 or 160 million workers. But it was very, very geographically concentrated in certain industries in certain places. And the people who lost that work, there was no, they couldn't just go to other manufacturing work. It wasn't, didn't exist. And so either they kind of hung on or they left the labor force or some joined, you know, typically much lower paid service employment. So the economic losses were, you know, were for first order the pecuniary losses. But then there's a greater notion of identity. What, you know, what is, you know, you know, for the people who, in that, in that work there, they didn't usually have, you know, high levels of education, broadly portable skills across sectors. Recent work by my colleague Amy Finkelstein and her co authors finds that actually manufacturing job loss, both during the time trade shock and during NAFTA was associated with higher excess mortality among non elderly men. And that was not true for non manufacturing job loss. Now it's an important question, why was it so? Why did that occur? But one reason is those were high wage jobs for non college men that had steady hours, long term employment prospects. They were often the anchor of a kind of a type of family structure and there was nothing comparable available. I don't mean to sound nostalgic, I think the data support that. But yeah, the other point is economists like to think of this and sort of they just invoke what we call the second welfare theorem that says well look, first you expand the pie and then you divide the pie and there's no tension between those two objectives. You can always just get richer and then make everyone better off. But that assumes that there is a like, even if that were going to occur, which it never does, that assumes that you can just compensate it with money. But I don't think that's true. You cannot give someone back their injured self esteem by writing them a check. And I think people understand this. People do not love the notion of being compensated for losses. They much would rather they would strongly prefer, they prefer unions, they prefer minimum wages, they prefer things that make their jobs pay well, not things that give them a crappy job. And, and a little, what do you want to call it, kind of a consolation prize for what they've lost now. So let me go on to your second question. Was this inevitable and was it necessary? So inevitably US manufacturing employment in labor intensive sectors would have declined eventually. These were kind of legacy activities in much of Europe they didn't exist anyway. But the US is a pretty low wage country actually at the bottom for such an advanced country. So those would have faded over the course of a couple decades. However, they would not have cratered in the time period they did. And the thing is, in labor markets, it's not just how much change you want, but how fast you want to get there that has such significant consequences. The labor market has a natural ebb and flow. A couple percentage of people retire each year. People enter and they enter new occupations and industries. Things that are growing, they don't enter, things that are contracting. So, you know, you can accommodate, you know, a 10 percentage point change over, you know, 10 years or so. But if it happens all overnight, it's, that's much more challenging because it's not like people are going to retire. Right. And it's not like young people can change careers partway through. So it really does matter how fast things change. So even if we accept the idea that, you know, we had to do this, and I'm not saying we didn't, we could have done it quite differently. We could have done it at a more gradual rate of change. And the Chinese trade shock, where China's merchandise surplus with respect, or US merchandise deficit as a share of GDP was up around 2 or 3 percentage points, just an enormous number at that time. And so that meant that manufacturing had declined very, very rapidly. If people were going to find full reemployment, they would have to go into services. And that was just, you know, the labor market, turns out, was really just not geared to change that quickly. So yeah, I think we could have, I think we bungled it in a variety of ways. Even conditional on the idea that it was a necessary step. We could have slowed it down, we could have controlled China's or tried to regulate China's currency manipulation. And we could have also had compensatory policies in place. We had trade adjustment insurance. It's trade adjustment assistance. It was very limited and it was directed towards training. We now know that wage insurance policies which the Obama administration experimented with were more effective, but we weren't really. The political and economic belief at the time was there was nothing to worry about, so we don't need to do anything about it. Finally, the question of. Let me ask your fourth question. Feel free to edit all this out. Was it necessary? Right. I think that's a much, much harder political calculation. I think the answer it was, it would have, I think it was hard to foresee at that time how Things would turn out. And I think it, you know, if, I think it was perfectly reasonable to say, look, China is a rising power. It's an, it's a, you know, it's a, it's a rapidly advancing country and it's, you know, the world's most populous country and we should not try to keep it out of the world trading systems. We should, you know, we should help it to reform and to open. Now, the fact that that didn't occur in exactly the way we expected, that China became more autocratic eventually and became, you know, less and less tolerant of, you know, of free expression, of democratic norms and became more of an adversary was, you know, I don't think we could have forecast that. In fact, if we were in the same position now and we had not and we had prevented China from joining the wto, we might be, look, now China has turned against us. It's non democratic, it's autocratic, it's our adversary. If we'd only let into wto, it wouldn't have turned out that way. So I think you're asking a bit much for people to have known that. But I think we, conditional on the decision, we bungled the policy badly.
- [15:42]
Jasia Munk
Excellent answer to all four of my explicit and implicit questions. I'm going to try and be more disciplined in the questions I ask you going forward to pick up on one of the strands of a conversation. Do you think that we have learned from the mistakes of the China shock in such a way that if we suddenly faced completely hypothetically a bunch of jobs disappearing because of a technological shock, let's say we would be much smarter at thinking about how we can minimize the social and then the downstream political costs of that? Or do you think that it is just incredibly hard to do that even with the learnings that we might take from that shock, in part because of questions of rational choice where the people who've just lost their jobs aren't necessarily the people who can best advocate for their own interests in part because that requires significant redistribution, which is always hard to do, and because all kinds of other structural obstacles would probably stand in the way of implementing even any of the smart lessons we may have drawn from those bad decisions?
- [16:51]
David Autor
Yeah. So I give you a simple answer to the question, have we learned anything? And the answer is no. I mean, our political system has learned nothing. In fact, to the degree it's learned anything, it's learned the wrong lessons. And again, I'm speaking of the US in particular. Right? The US has eliminated the Trade Adjustment Assistance Program, the only thing we ever had in place to help people displaced by trade. Right. It's no longer funded by Congress. So it seems like if we even, you know, so that, that just like is a mind blowing. And then the only lessons we seem to have taken about trade policy is that we should sort of across the board harm our neighbors and our adversaries simultaneously and protect sectors that don't need protecting while failing to invest strategically in those that do. Right. Like it's crazy to think we should be making tube socks or assembling iPhones in the United States. Right? That's like, there's never going to be high paid work in that and we're just taxing ourselves by doing it. On the other hand, we should be thinking very strategically about sectors like semiconductors, about fusion, artificial intelligence, drones, electric vehicles, you know, power generation, robotics. And yet the United States is not doing that. And moreover, you know, we're not big enough, the US Is certainly not big enough to contend with China on all those fronts. But of course we used to have allies in, I don't know, Canada, if you've heard of it, in joke, in Europe. Right. You know, who collectively with the United States were an incredibly large economic strategic bloc, and yet we fractured those relations. So I think like so many things in the Trump era, it's exactly the right question and just a terrible answer to the question. And so it's, I don't want to say no one has learned anything. I think both our political process and the economics profession has, you know, sort of done some rethinking and said, oh wow, we shouldn't be so laissez faire on trade. And I agree that's like the first like, you know, hit me over the head with sledgehammer lesson. But then in terms of what to do about it, I think our responses have been, you know, counterproductive and unwise and shortsighted and harmful in the long run. Both the United States and to our allies.
- [19:11]
Jasia Munk
Obviously a lot of economic discussion now is about the effect that artificial intelligence is going to have. And you've written very interestingly about that. I want to get to that in a moment, but before we get to that, how would you characterize the pre AI state of the American economy and of the American job market more specifically? I don't know whether you want to take that question by saying what did it look like on the day that ChatGPT 3.5 was released? Or what does it look like today, net of the effects of AI or take the question whatever way you Will,
- [19:42]
David Autor
so I would say before COVID like up from kind of 2015 through 2020 through the spring of 2020, the US labor market was in great shape, actually. Wage growth was robust. Productivity growth was pretty strong. There was a lot of reduction in inequality among, you know, kind of normal mortal earners, not, you know, the people in the between the 10th and the 90th percentile of the income not at the top. Things are still growing explosively. You know, you can have mixed feelings about that. But in terms of the people who, I'm so mostly, you know, deeply concerned about people without college degrees, you know, people who are earning, you know, below at or below the median, things were looking very good. So the, you, you know, the US labor market was in great. And it, you know, if you want to say it this way, the Trump labor market was a great labor market, especially for blue collar workers. And that provided a lot of basis of support for Trump. And if you ask, why did so many Hispanic voters, why did so many black voters come out to support him in 2020 and then more so in 2024? Partly because things were going very well from the perspective of American workers who were outside of the league. So I don't think most Americans begin to appreciate that There are only two countries that have had economic miracles over the last 20 years, and they are the US and China. Right. And so the notion of sort of American carnage and so on is so misguided. And productivity growth in the United States has been so robust and unemployment has been so low. Wage growth has been relatively strong. There are lots of things to not be happy. I don't mean to imply it's a utopia, but it really was, it was overall going very well. Then, of course, we had the COVID pandemic. That was, you know, kind of, that was a huge, huge setback. But then the US Economy recovered very robustly, more rapidly than almost any other country. Now, we also had a lot of inflation, and, you know, that was partly because of COVID inevitably, partly because of, I think, Biden. The Biden administration, you know, way overstimulated. But that was coming down even before 2024, and we had managed to land that without a recession, which is itself remarkable. Some would call this the immaculate recovery. To recover from that type of inflation without going to recession is almost without precedent. So we are already coming back into pretty good shape in 2024, 2025, when Trump took office. And then since that time, we are just awash in incredible uncertainty between tariff policies, shifting energy sectors, whether oil, gas, Solar, coal, Multiple wars of choice, and then sort of political persecution of our leading technology companies, of our universities, of elected officials. So it's actually, I, you know, I mean, this is not an area in which I've expert. I'm expert. But Nick Bloom has done a ton of work about the cost of uncertainty itself. How do you invest when you don't know what prices are going to be, what tariffs are going to be? Right. Who you're going to be at war with? And so I think the US economy and labor market has been remarkably robust in the face of all that. But job growth has slowed enormously. And in fact, the current jobs report of the last a month was actually quite negative. Now, those fluctuate a lot, so I don't think one should read too much into that. But the irony is the US didn't recognize how good it had it. And in responding to this misperception of how good, in thinking it didn't have it good, it implemented a lot of policies that arguably made things much worse.
- [23:40]
Jasia Munk
Yeah, it is striking that there was that gap between some of the data and the perception. You know, one thing that I'm really struck by, and I've mentioned this before on the podcast, is just that, you know, China and to a lesser degree India and other very populous nations around the world have had phenomenal economic growth over the last decades, which obviously easier to do when you are coming from behind, but is a phenomenal success story.
- [24:04]
David Autor
China's growth over the last four decades is miraculous, and not just for China. It is the creation of the world's middle class for the first time. Right. China's growth has created prosperity not just in China, brought hundreds of millions of Chinese out of poverty, but also in Central and South America, in sub Saharan Africa. Right. It has been the best economic event in world economic history for poverty reduction worldwide. It's created challenges. I'm not saying it's all upside, but it is nothing. The world has never seen anything like it.
- [24:37]
Jasia Munk
And one of the impacts of that has actually been a significant reduction in global inequality. If you look at the global Gini coefficients come down significantly because hundreds of millions of people have been lifted into the global middle class. Now, what's striking about that is that obviously as a result, the share of global GDP taken up by China and to a less extent by India and other countries has risen very significantly. The share of global GDP that Europe has has declined very rapidly. And it's astonishing to what extent the United States has actually held on. It's seen some decline in its share of global gdp. But given the background condition of that incredible economic miracle in China and significant economic progress in other nations like India, it's astonishing how little America's share of global GDP has declined. And yet people did feel very unsettled, very negative, very pessimistic. What explains that disjuncture? Is it just that once you're very affluent as a nation, it becomes much harder to generate growth that actually translates into a tangible improvement in your life prospects? And this is the kind of disease of affluence that we're going to have forever? Is it to do with political changes, change of the technology, environment? What explains that gap?
- [25:53]
David Autor
Okay, so I need to be clear. You know, I'm now just going to be giving you an uninformed opinion in the sense that I'm an economist, I'm not a sociologist, I'm not a philosopher. And so I can only. Why do people's perceptions, why are they so negative given that there's so many outward indications that things are so positive, or were? I would say it's two things. One is, unfortunately, Americans are very ill informed about where we stand relative to others. So if they're told America used to be great and is no longer great, we're being taken advantage of by everyone else, we're the laughingstock of the world, we've given up all our power, we're just a bunch of snowflakes and so on. They begin to believe that things are very bad. That's one thing. The other thing that I think is more, which I think has more foundation, is there's a ton of economic insecurity in the United States that I don't think was true in the 1950s, in the first three post war decade, three post war decades of really rapid economic growth combined with very robust, even growth of people at all education levels, of just rising living standards. So the average in the United States is not very informative because very few people are near the average. The outcomes are so dispersed. So I do think many people felt for the majority of Americans who don't have four year college degrees, they were like, well, our horizons are shorter than they were 40 years ago. We don't have a secure pathway into the middle class. It's not obvious that our children will do better than we, we did. And so, you know, I think that that is real. And that's, you could say, look, the US did, you know, created a lot of wealth on average, but it didn't distribute it very evenly. Now some would say, right, well, that's how it got so wealthy is because it didn't try to squander it all on redistribution, welfare state and so on. I don't personally believe that, but someone could make that argument. I couldn't refute it immediately, but. So I think the US could have used its prosperity a little differently to create more economic security. And I think that would have resulted in less dissatisfaction. Just like I think in terms of the China trade shock. I don't mean to imply the Chinese trade shock certainly wasn't negative for everyone. It lowered lots of piece of prices. It may even been positive and aggregate for the United States. But the distributional consequences were so adverse that they made it very scarring. It would have been possible to use some of those resources more effectively. Even if you can't make people whole, you certainly could make them better off than they were without that assistance.
- [28:37]
Jasia Munk
So I feel like we've been working up to the topic and now I want to grab the Nassaus by the horn. That's a horrible mixed metaphor. What kind of technology is artificial intelligence? I mean, when we think about artificial intelligence in the context of other historic technologies that have had a major economic impact, what do you think is the right comparator? What kind of level of change and disruption are we talking about here?
- [29:04]
David Autor
Well, so let me try to describe what makes artificial intelligence distinct from other technologies. Right. So, you know, prior to the computer, we were in eras of mechanization where we had, you know, better tools, but, you know, that could do amazing things like, you know, chemical reactions and, you know, plows and so on. The computer was the first commercial symbolic processing tool, right? Something that could take information, stored symbols and instructions and act upon, process and analyze information. We never really had any tools for that other than our own minds and pens and papers. And that gave us enormous power to take repetitive, complicated, difficult tasks, everything from calculating, you know, space trajectories to playing chess and write them down as code and have increasingly dramatically inexpensive machines do them with perfect accuracy, with incredible speed. And that was an amazing, amazing breakthrough. And we got kind of, from the 1980s to 2020 was this era of computerized automation. And that was very displacing for people who did work that was often relatively expert but followed well understood rules and procedures. It was very complementary to professionals who took an important part of their task, of their work that was time consuming. Retrieving information, looking things up, calculating, and just made that super effective and inexpensive so they could focus on what they're really good at, which is making expert decisions. It was not especially helpful for people who were doing a lot of hands on and blue collar work where it wasn't very applicable. But everything that we computerized, we had to understand the rules of how to do it. So we had to understand it formally so we could write down the procedures, so that a non sentient, non improvisational, non problem solving machine could just do it by following the instructions. And that actually turns out to be a huge limitation on what we could program computers to do. Because as the philosopher Michael Polanyi said, we know more than we can tell. There are many, many things that we do that we don't know how we do. Right? We don't know how to, you know, we don't know the code to write a funny joke or to make a persuasive argument or to develop a hypothesis, or to ride a bicycle. Right. Or to cook a meal. Right. These are all things that people do based on tacit understanding that they learn inductively by doing them. They never write down the rules. And so we didn't have machines that could do those things because we didn't know the rules. And so AI overcomes that Polanyi paradox, because AI can learn things without us telling it. Right? It can know more than it can tell us at this point. Right? Because it learns inductively from examples, from data. It can learn from unstructured information. It can solve problems we don't know how to solve. It can do tasks that we would think of as creative if done by another person. So that moves technology into a whole new domain of cognitive and physical activity that we just didn't have machines that could go there. So it is really qualitatively different from traditional computing, right? It is, you know, an analogy I've used in other places. If you think of traditional computing as being like a orchestral musician that reads the notes and plays the instrument exactly in sync with the rest of the orchestra. Right. AI is much more like a jazz musician that can solo and improvise and extrapolate, and that's, that's incredibly powerful. And we're only beginning to figure out how to use that. Well, I mean, one thing that AI is not is a better, cheaper, faster version of something else, right. In many ways, AI is not as good at many tasks as traditional computing. You saw New York Times ran this story yesterday about don't use AI to do your taxes. It's not reliable, right? No one ever said, oh, I wouldn't use Excel on that problem. It'll probably hallucinate, right? That was not a thing
- [33:12]
Jasia Munk
so that's really fascinating, and I guess it makes me wonder about the terms in which we should think about the economic impact that artificial intelligence is going to have. For good reasons, social scientists like to look to the past for guidance about what might be around the next historical corner, while being very careful about making projections right. But when there's a new technology that arises, the natural thing to do is to look at, well, what happened when other big technologies came about in the past. And that may be an imperfect way of reasoning through the problem, but it's the best we have right now. What tends to have happened in the past is a phenomenon that economists call job reinstatement, right? Where some machine or other form of automation takes away a bunch of jobs that can now much more efficiently be done by this machine. And often, as we discussed in the previous part of a conversation, that can be deeply disruptive to the people who are directly affected by it. Often they are too old, too set in a particular range of skills that they've learned, perhaps particular way of life that was associated with that profession, when you're thinking of farmers and peasants and so on, to really accommodate to these new jobs. And that can have deep disruptive consequences. But their children, their children's children simply adjust to this new world, and they learn the new skills that are now necessary to do new kinds of tasks, many of which might not have existed beforehand. And therefore, over time, we haven't seen the phenomenon of mass scale job loss. Now, one of the background conditions for that was that there was always what I've come to call a kind of mental reservoir. There was always a kind of reservoir of tasks that machines couldn't do. First, because there was just really no way of automating cognitive tasks. And later, because to the extent that we could automate cognitive tasks, we could really only automate cognitive tasks that were kind of following the steps in an algorithm in the kind of way that a computer can substitute for. Well, now for the first time, we have a machine that does potentially rival us to some extent rivals us even today, and probably to a much larger extent is going to rival us in 20 or 10 or five or perhaps two or one years in being able to carry out all of those cognitive tasks. Is that going to make this metaphor from the past of job reinstatement obsolete? Or do you think that that is a mechanism which is sort of timeless and which is going to persist even as this new machine sort of penetrates deeply into an area of human skill? But up till now, only members of our species were able to carry out.
- [36:05]
David Autor
Yeah. So first let me say that although we've been through multiple technological eras and not all the transitions are smooth and painless. So the first industrial revolution was an era where valuable artisanal skills were wiped out. People who did textiles and weaving and so on, those artisanal skills became valueless. Instead they were replaced by indentured children and unmarried women working in the so called the dark satanic mills in Blake's terms. And it took decades before new work began to appear that actually used the kind of expertise in numeracy and literacy of the next generation of workers. So there's no guarantee that it ever goes smoothly. And usually even in the best case, there are winners and losers. Just like in the case of the China trade shock where some people's expertise is devalued, new expertise is reinstated. So, and you know, we've seen a lot of this over the last four decades, right? Computerization has been decidedly non neutral for, you know, welfare of different skill groups. It's been great for educated adults. It's really been probably, you know, on average, you know, not so good for people who don't have college degrees because it's hollowed out the middle of their, you know, the higher paying jobs in production, operative, clerical, administrative support and moved a lot of people into services, you know, food service, cleaning, security, transportation, home healthcare. And although those services are socially valuable, that work is relatively inexpert. Most people can do it without much training or certification, meaning it won't pay well because there's just not scarcity. So AI changes the game again. And I guess there's sort of two, there's two things we want to think about simultaneously. One is, well, how does this change? What is human comparative advantage? How does it change what people think people should do and what things? That's the question saying, well, what will the machine make? Too cheap to meter. You wouldn't want to do this anymore because you can't compete with a machine that can do it instantly and almost zero cost. And what will it make more valuable because humans are really needed to do those remaining tasks. And there we don't have a very good answer to that question. Like for now we can say, look, there's lots and lots of physical hands on work that will require labor for even if we make tons of progress in robotics for a couple decades. And I am very confident saying, look, there will be tons of people working as doctors in medicine, I think in education, I think in the trades, I think, and I think in many, many Settings, humans will continue to provide a kind of a intermediary layer between sort of formal bodies of expertise and machine support and other people. And that we will look to other people to guide us in decision making in every high stakes domain. And I don't think that's going to go away, but that may mean there'll be fewer people who are more expert or in some other cases, it'll mean there are many more people who are less expert, right? And so technology can be very bifurcating in this way. Like useful, you know, examples like, you know, ride hailing, right? You know, so Lyft, Uber or whatever, right? That really changed the occupation of taxi and chauffeur drivers. Not just because it allowed a lot of people, more people to do it, but because there was, there were two components of being a taxi driver, right? One was driving the car and the other was knowing the routes. And now ride hailing meant you no longer need to know the routes, right? You could go drive in a city you'd never been in before and you could drive tax Uber there immediately with the right software. So that actually reduced the expertise requirements. It simplified the work, it allowed many more people to do it. It simultaneously created new opportunity and created unwelcome competition for incumbents. In other cases, you'll see just the reverse. That technology eliminates the simple part of the work and leaves the expert components remaining. I think many people would say that in their professions, right? The part they actually have to do is much harder because the grunt work is done for you and you have to focus on diagnostic skills or in software architecture or in contracting on solving the hard problems. So I think we have to ask, well, where will we re specialize? What will be the things that we do that are expert now that will be decontented? What were the things where human expertise will become more valuable? Because it's kind of the central piece remaining to put all this together. So I think that's a very, very hard question. Obviously many people are working on this, but I don't think it's sufficient to say this or that is exposed to automation, which is done all the time, right? Oh, this is exposed to AI. And the implication being that, oh, if it's exposed, it's at risk, right? It's going to shrivel up and die. But that's often not the case. Exposure could mean it gets simpler and more people do it at lower pay, as in ride hailing. Or it could mean that it actually becomes more specialized because the technology does the easy part and the People who remain become, you know, kind of, they do a more abstract, more cognitively demanding task. Like we've seen that happen. You know, accounting has gone from bookkeeping, right? To, you know, planning and, you know, and forecasting. And so that can happen simultaneously, right? In the professions, computers have made our work, taken away a lot of the simple work and made us more kind of focused on the high end of the work. So that's very hard to forecast. Now there's a second component of this, and I imagine many people have this intuition like, well, doesn't it just mean there'll be less work for people to do? And I don't want to be a Pollyanna and say, no, no, that can't happen, happen. Here's how that can happen. You know, Vasily Leontief, you know, famously said, well, humans are like horses, they'll soon be put out to pasture. And you say, well, obviously he would, you know, either he was wrong or he was very far ahead of the time. In other words, it hadn't happened yet. But why would we be put out to pasture? Well, you know, ask yourself, why aren't horses being used anymore? They're just productive as they ever were, right? They're, they're ready to go. Why don't we use them? And the answer is, there's no circumstances under which a horse can compete with a internal combustion engine or some other forms of transportation. It's fundamentally too expensive, right? You know, horses, you have to maintain them, right? You have to have, you have to have a stable. You have to, you know, have, you know, oats and you have to have land to graze them on and so on. And I don't know if horses graze. Actually, I know nothing about horses. Let's just make that clear.
- [42:53]
Jasia Munk
I don't know. I believe they graze, don't they? The mass graze. I think they do. Some listener will correct us.
- [42:58]
David Autor
Yes. So there really, there are cases where they can't be a competitive factor of production, right? Similarly, like, there's no wage at which a human could compete with a computer to do a set of calculations, right? There's just like you wouldn't. The amount of calories they would have to burn just to do that calculations couldn't support the food consumption that would be required rather than just do it with a microprocessor. So you could, it's possible to have set of circumstances where actually the cost of hiring a person actually just like, like the pure malfusing cost would not justify hiring them when you could get a machine to do it. It's not impossible. And one indication that that might could be occurring is that the labor share of national income is falling, right? It's been falling for a while. It's been falling for a couple of decades, but it's fallen from about 60 percentage points, maybe a little bit above by about maybe 8 percentage points in the United States. So that's the majority of every dollar first going to workers and the minority going to capitalists. And now it's getting closer to parity. So that is a possibility. I don't think that we're going to enter a world where labor has no value, where we have no more labor scarcity. I think that's a long way away. I think that's unlikely. But on the other hand, if labor's share of national income falls even by a substantial number, by 20 more percentage point strength, that's going to be pretty seismic.
- [44:23]
Jasia Munk
So I have a few follow ups on this. The first is about the example of horses, which is really interesting. The primary reason is that, as you're saying, keeping and maintaining a horse is just very expensive, which is why the number of horses has gone down radically in the world. I think there may be also a second reason. I recently visited an Amazon warehouse, which is a fascinating thing to do. And one of the things that becomes clear is that there's these very efficient robots which move produce around and store them and bring them back out. The person who wants to pick it out to put it in your box when it's needed. Sometimes these machines malfunction some kind of small way because some piece of lint gets attached to their wheels or whatever, and the human needs to come and clean that up. That is very costly because to avoid accidents, they have a special vest which makes all of the robots around stop until the human is out of that area and they can resume their operations. And so there's a significant cost to that human robot interaction. And so an additional reason why it may be bad to have horses in the street is that, you know, there's cars around and horses and cars don't mix very well. And that is sort of an additional reason. You know, the handover cost from machine to human and back may be so high that even in machine, even in tasks where humans have a small competitive advantage, it may be easier to push them out of a loop for that reason entirely. The other point I wanted to make is about judgment, right? You're assuming that in areas like medicine, for example, that human judgment will always be required. Now, on many benchmarks, Humans are AI bots are now very close to the judgment of humans on those decisions. They don't yet have sufficiently big context windows that they can take into account all of the patient's medical history and that they're really fully read into the situation and so probably still want high stakes decisions to be made by doctors. But quite plausibly, machines are going to have even more powerful models within a few years, so their acuity in judgment is going to be even higher. We're going to continue moving towards being able to solve those kind of external constraints, for example, by giving them larger context windows, by finding smarter ways for them to really understand more about the context context of the patient. And we already know that humans actually tend to trust AI bots quite highly. So, you know, why are we so sure that those most high stake medical decisions aren't going to be made by humans eventually? And then the real question I want to ask you, based on your academic expertise, is that, you know, it seems to me there's a world of Silicon Valley technologists who don't really understand the constraints of a real world and who think all of the jobs are going to disappear tomorrow. And that's deeply naive. On the other hand, I think that there is a strain in parts of economics where people are sort of like looking at studies of productivity gains from using ChatGPT 3.5 in highly constrained circumstances two years ago and projecting forward from that. And I think that tends to give you far too low estimates. But if we think that, you know, for 50 or 100 years there is a significant decline in the demand for high school labor. Right? If a world which we're entering is perhaps one where there's some jobs that are highly specialized, that are new, and there's a good number of jobs that are not very specialized for which we now need people. So there's a little bit of the lawyer who no longer has to do the grunt work and just makes the decisions phenomenal, and a little bit of rail right hail worker who actually has been deskilled in a certain kind of way, that is going to be a pretty bad outcome for most people. Right? Because, and this is a genuine question, you will know this better than me, but I'm imagining that if there's a small cast of people who are highly skilled and make these high stakes decisions, and then there's a big pool of not very differentiated labor that can do these relatively low skill jobs, and perhaps there's a little bit of slack in overall demand for workers, that would be enough to radically decrease the wages of a lot of people and lead to pretty deep economic disruptions as I'm imagining it. But you're the economist who can model this, and I'm just a political scientist who's speculating.
- [48:36]
David Autor
There's so much in that question, and these are all constructive speculations. And of course, I don't know the answers either. I do think you certainly raised the point that there may be many domains that we think you need humans involved that ultimately you don't. Right. And medical diagnosis might be one of them. Right. Where, like, the costs of error are so high and you have machines that could be trained on tons and tons of cases. Maybe they'll just be on it. You know, they'll be better. I think in many cases they'll work collaboratively. I don't think it'll be full automation for quite a while. And I think there will be a lot. But, okay, but, so, so that, but let me just, you know, take your question on its premise, right? It's quite possible there are areas where we have labor now and we just won't want it. It. You would just rather, you know, there was a while, like, where, you know, ATMs were complementary to bank tellers, right? Bank teller employment grew. Why? Well, banks began branching, right? Because they could do it at lower cost. They built more branches and they, and they hired tellers not just as, like, cashiers, but as sort of salesperson people to introduce you to other products like, you know, loans and investments and credit cards and so on. But now actually, teller employment is declining. People have said, hey, you know what? I, you know, I can just do it all online here. I don't need a person. You know, most of the people who stand in line to see bank tellers are relatively elderly individuals who have been doing that for a long time. They're used to that. So there may be a period where there's collaboration, then eventually there's just encroachment. Now, I think the other side of the thing we need to think about this is we say, look, you know, if all this is really happening, you know, we're getting very wealthy all of a sudden, right? That means that we're doing a lot of stuff very cheaply, right? That, I mean, that there's a productivity side implied by this. And so how does, you know. So let me put two thoughts together. One, you know, why. Why did this happen to horses so fast? Why has it not happened to people so far? Well, a few different reasons. One is, you know, people are much more flexible than Horses, right? We have a capacity to educate ourselves and do many, many, many things. Horses probably have less of that, right? So as my colleague, my co author, Ana Salmons likes to say, you know, people are not one trick, ponies two. Of course, people own capital, right? We own the machines. The horses don't own any capital. So, you know, we have, we get, we bear, get the fruits of a lot of those, you know, productivity benefits. And the third is, of course, we vote. So democracy also really buffers the effect or, you know, shape. Shapes how these things bear out. But when you say we could all be. We could end up jobless yet with high productivity, then I'm reminded of what the science fiction writer Ted Shong said. He says fears of AI and automation. They're not really fears of the technology, they're fears of capitalism. It's that we're afraid of not what the technology will do, but what the economic system will do once we have those capabilities, that if people don't need to hire workers, then the fruits of that productivity will not be distributed evenly. And I think that is a very significant concern. I think the biggest concern I have about the technology, at least from the labor market perspective, is the threat that it implies for democratic function. That in my view, the labor market, you know, is really, you know, arguably the, or certainly one of the most important social institutions, right? And it works hand in glove with democratic institutions because most people in democracy are both considered to be, you know, contributors through their labor and their taxes and claimants through their, you know, retirement and their education and so on and social insurance. And if all of a sudden labor were just devalued, right. Then most people would be claimants without being seen as contributors. And so sort of, I think the political economy of that is a nightmare, right? Just say, oh, we just have a few rich people and we'll just tax them and redistribute the rest like that. You know, historically that has not tended to work out. So I do think this is a very, you know, it's not that I see this as the most likely scenario, but it's sufficiently plausible that I wouldn't dismiss it and say we shouldn't worry about it. Right. And I do think people are very polarized in this. Many people are like you, we're doomed. There's lots of doomers and then there's lots of kind of utopians. And I don't think, you know, I think we should recognize a range of possibilities are plausible. Some of them quite good, some of them quite bad, many of them will co occur. By the way, we'll make some terrible mistakes even if we have some big gains. And I do think we should be worrying or not worrying is probably the right. I think we should be adopting policies that help us to ensure a better transition if such a transition should be needed.
- [53:20]
Jasia Munk
Needed so much for listening to this episode of the Good Fight and the rest of this conversation. I asked David about his recent paper with Daran Acemoglu and Simon Johnson. They argue that we have to determinist a view of the future of artificial intelligence. But we tend to assume that the future will depend simply on the natural tendency of where this technology goes, but that actually it'll depend on the kind of policy choices we make, and that we can shape the adoption of this technology in ways that will actually boost ordinary workers to listen to that part of the conversation to support the work we do here. To stop listening to those annoying jingle ads you will have gotten for the last 50 minutes or so, please go to writing.yashamonk.com Listen, become a paying subscriber and set up your private podcast feed on your favorite podcast app. Writing yashamonk.com listen.