Transcription
Welcome to the John F. Kennedy Jr. Forum here at the Institute of Politics. My name is Iris, and I'm a sophomore studying economics and social studies here at the college. And I'm also member of the JFK Jr. Forum Committee.
Before we begin, please note the two exit doors on the park and JFK street side of the forum. In the event of an emergency, walk to the exit closest to you and congregate at the JFK park across the street. Please also take a moment now to silence your cell phones.
And to kick off our program, join me in welcoming the Dean of the Harvard Kennedy School of Government, Jeremy Weinstein.
Good evening, everyone. Thank you for joining us tonight. Welcome to the John F. Kennedy Jr. Forum at the Harvard Kennedy School. I'm Jeremy Weinstein, the new Dean, and I'm thrilled to welcome you to this special event. I'm delighted to introduce this year's James M. and Kathleen Dean Kathleen D. Stone Lecture in Economic Inequality. We're fortunate at HKS that we can bring some of the world's leading thinkers on economic inequality and wealth concentration to Harvard through this lecture series. This is the fourth Stone Lecture. Previous Stone Lectures have been delivered by the economists Thomas Piketty, Emmanuel Saez, and most recently Joseph Stiglitz.
This lecture series, and indeed all the work of our Stone Program in wealth distribution, inequality, and social policy, would not be possible without the generous support of the James M. and Kathleen D. Stone Foundation. And we thank the foundation's chairman, Jim Stone, its president, Kathy Stone, and its executive director director, Sonia Pleasant, for their generous support.
I'm especially pleased to welcome Professor Daron Acemoglu to Harvard as tonight's speaker. He'll be formally introduced in a moment. But I'd like to offer one personal reflection by way of welcoming him. As a PhD student training in economics and political science in the 1990s, I was shaped in important ways by a paper written by Daron Acemoglu. And I know that this is a common experience for almost every scholar who has come of age in this generation. And we were not all shaped by the same paper. Given Daron's wide-ranging contributions, it was just as likely for a PhD student studying political institutions to be shaped as one studying fiscal crises. And for me, as a student of global poverty and development, it was a classic paper on the colonial origins of comparative development.
20 years later, when I found myself working on the impacts of new technologies on society, and I began to look for the most consequential papers, you can guess what I found. Papers by Daron on the economic impacts of automation. Papers by Daron on AI and the role of government policy in shaping technologies economic consequences. And then, of course, the book that he'll discuss tonight.
Now, I want to welcome to the stage Deirdre McCloskey, the director of our Stone Program, the Peter and Isabel Malkin Professor of Public Policy, and Professor of Sociology in the Faculty of Arts and Sciences, who will formally introduce this evening's speaker. Deirdre.
Well, thank you, Jeremy, and thank you all for being here for the 2024 Stone Lecture in Economic Inequality. As Jeremy said, my name is Deirdre McCloskey. I'm the Peter and Isabel Malkin Professor of Public Policy, Professor of Sociology, and director of the Stone Program in wealth distribution inequality and social policy. And I'm just thrilled to have this opportunity to thank the Stones for their generous support for the study of inequality at Harvard. Thanks to the Stone Foundation, we have a really exciting slate of programming and opportunities coming up for the next year. So, I hope you'll join us and sign up for our mailing list to stay abreast of all the activities that we have planned. You can do so online at inequality.hks.harvard.edu.
That said, it's my absolute honor to introduce the two eminent scholars who will be on stage with us tonight. First, our 2024 uh Stone Lecture will be delivered by Professor Acemoglu, as Jeremy said. Daron is the Institute Professor at MIT, and he's an elected fellow at the National Academy of Sciences, the American Philosophical Society, the British Academy of Sciences, the Turkish Academy of Sciences, the American Academy of Arts and Sciences, the list goes on and on, the Econometric Society, the European Society of Economics, and the Society of Labor Economists. His work covers areas including political economy, economic growth and economic development, technological change, inequality, labor economics, and the economics of networks. And his work has been recognized by many awards, including the inaugural Sherwin Rosen Award for contributions to labor economics, and the John Bates Clark Medal. He's co-authored uh many books, including uh the classic Why Nations Fail with James Robinson, and the book that he'll be drawing on tonight with Simon Johnson, Power and Progress. And we're just thrilled to be hearing from a scholar whose work has been pathbreaking in its recent insights into how technological change drives inequality, and the conditions under which technological advances can generate shared prosperity rather than benefiting only a narrow elite.
After Daron presents his lecture, we are very, very fortunate to have Professor Dani Rodrik join him on stage. Uh Dani will be moderating a question and answer session. Dani is the Ford Foundation Professor of International Political Economy here at the Kennedy School. And Dani's research uh covers topics including globalization, economic growth and development, and political economy. And like Daron, Dani has been recognized by many awards, including the inaugural Albert O. Hirschman Award from the Social Science Research Council, and the Princess of Asturias Award for Social Sciences. And Dani's prolific writings include classic books, including The Globalization Paradox, Economics Rules, and Straight Talk on Trade.
So, it's my real pleasure to thank Daron and Dani for their insights tonight, and to welcome Daron to the stage. Thank you, Daron.
Let me first check that this is Wow, wonderful. Uh Thank you very much for this invitation. It's a true honor to be here. Uh I cannot think of a better place uh than the Harvard Kennedy School under the auspices of the forum and the Stone Foundation to talk about issues of inequality and in particular how they're going to shape our future. And thank you very much for that wonderful introduction, Jeremy and Deirdre. And uh I'm really happy also that to be sharing the stage with my friend Dani Rodrik, who uh has been a trailblazer in on these topics as well. So, uh I'm looking forward to that conversation, too.
Uh I want to talk about the future of shared prosperity because I believe that unless we strike the right sort of institutional and technological trajectory to create the type of economic growth that's going to benefit a broad cross-section of society, we would not be just creating a sad society, but we would be creating a very unstable one. So, shared prosperity, I think, is in some sense in the 21st century a sine qua non for all sorts of things that we take almost for granted. And in some ways, there are signs that today we should be in a better situation for creating at least the foundations of shared prosperity than any time before, because many people view our epoch as the age of innovation. According to a number of measures, this seems to be borne out. You see a new widget, a new app, a new telephone, a new uh electronic every day. And in fact, various statistics bear out that especially in the more important sectors of the economy that provide input and information to the rest, there has been a true explosion in innovation. So, the green line there that I'm showing is the total patents filed in the United States. It increased by about fourfold from the 1980 to today. Even more important for those who believe that we are in the midst of a transformative change, a large part of that change is driven by electronics, computers, and information technologies. And today, we are, at least in some parts of the country and some parts of the world, we are in the midst of a complete excitement about AI and generative AI.
But the questions that are sometimes not sufficiently emphasized are really at the heart of the shared prosperity implications of this. Who will benefit from the this innovation? What will these innovative efforts focus on? And what will they imply for shared prosperity? In particular, I think there is a danger that more and more people are seeing that AI advances, even if they materialize, may end up benefiting a rather narrow segment of society, a technological elite, some business leaders, and some workers with very specialized skills that are useful for AI, and not so much the rest of society.
Against this, both in public debates and in economics, there are some ideas that in fact, there are automatic forces that should guarantee something like shared prosperity from these advances. And the uh the heart of that is what Simon Johnson and I call in our book Power and Progress, the productivity bandwagon. It's an implication of pretty much all of the models and all of the approaches that, for example, those of you have taken economics courses in your master's or PhD programs would have seen, that technological advances improve our capabilities. That leads to higher productivity, meaning we can produce more with the same amount of resources, especially labor. And as a result of that, workers also benefit, and in particular, wages increase. And if wages increase, that's a very important foundation of shared prosperity because most of us earn our living in in in the labor market as wage and salary earners. So, that's a very important step towards those gains being shared.
However, it turns out that there are both uh there are some preconditions or assumptions in this productivity bandwagon. In particular, in what I have shown that is the link between the green and the red circle up there presumes that as productivity increases, employers are going to have a desire to employ more workers and pay them more, and labor market institutions are such that those forces are indeed going to lead to higher wages at the end. If you are a believer in the productivity bandwagon, that's great. You'll have a you'll sleep much better at night. And if you want to find examples in history, you can find many. There are many episodes in which indeed there are technological breakthroughs, those increase productivity, and uh there is some sort of shared prosperity forces that are operative. But, if you're more skeptical, sorry for your sleep, but you can also find many examples where productivity increases, technological breakthroughs happen, but nothing beneficial takes place for workers. And that's for two different reasons. The first one I'm going to talk about on this slide, which is related to that labor market institutions. Even if it turns out to be the case that labor becomes more valuable with technology, there are other ways in which employers can have access to labor and can motivate or force that labor to work. So, that's the institutional or the power piece, and that's why the book that Simon and I wrote is called power and progress. Power is a central part of that.
So, two examples there. One uh is is a very technological breakthrough for the medieval economy, windmills. They uh spread quite rapidly and changed the production process, increasing productivity by more than tenfold in some tasks, but the lot of peasants and farm workers did not improve in large part because this was uh this technology was embedded in a coercive labor market where uh peasants were had obligations, and they did not have the freedom to go and work for other people. And in fact, even those uh important technologies were monopolized by the very wealthy, the landowners or the or the or the high clergy. As a result of that, the gains from this very important technological change became very unequally distributed. The one on the right is an even more transformative technology, especially with in the speed in which it transformed both a regional economy and the world economy. It's uh it's the uh cotton gin. Uh the US South in the middle of the 18th century was an economic backwater, but it was generally recognized that it had climatic conditions and perhaps other possibilities that would enable cotton production. But, the cotton that could be grown in the US South could not be cleaned with existing technologies. So, the cotton gin changed that at one fell swoop, and it made the US South uh the largest exporter of cotton and the dynamo for the world economy, which uh was built on textiles at the time, especially British economy. But, the workers who produced cotton in cotton plantations and other uh parts of the economy were enslaved black people, and it's not a surprise that in that unequal relationship, of course, they did not benefit. In fact, they were moved to the deep South where working conditions worsened, uh and there is no evidence that their real incomes or any part of their comfort improved.
Now, these power-related reasons may appear distant, but they're not. I'm going to come back to it. And these examples may appear a little bit distant because the favorite examples of people who think that the productivity bandwagon is almost almost a law would be much more about industrial technology. So, let's turn to the Industrial Revolution. So, the Industrial Revolution is a very interesting uh episode for us because first of all, it's transformative. Secondly, it is something uh on which we are all uh beneficially dependent. Our comfort, our prosperity, our health is much better than people who lived 250 years ago in large part because of the process of applying, developing industrial technology and scientific knowledge, and that started with the Industrial Revolution sometime around the middle of the uh 18th century. And uh you might hear people uh journalists, uh not many scholars, but perhaps some scholars and some commentators saying, "Look, the Industrial Revolution proves that even disruptive technologies at the end bring benefits to everybody." Well, the real story is a little bit more complicated. It is not one of automatic bandwagon. On the contrary, the evidence is very clear that for about 90 years, the uh beginning of the Industrial Revolution did not bring any improvements in the living standards of workers. By all measures, uh workers' conditions significantly worsened. They started working longer hours, perhaps 20% longer hours. They were subject to much more discipline, and their overall real incomes did not increase much or at all. And in fact, some of the workers that were at the forefront of the new industrial machinery, for example, the handloom weavers in the textile sector have seen their real wages drop by about a third or by about two thirds to a one third of the level that they had at the end of the uh uh of the 18th century.
This also is somewhat related to power because uh some of the ways in which uh new business owners, capital owners could organize the factory would not have been possible if workers were organized and had greater voice. But, there is something deeper about that. It's also about the nature of technology. The early technologies of the Industrial Revolution were all focused on automation, meaning substituting machinery and services of machinery for tasks previously performed by workers. First, very large scale in spinning, then uh weaving, which is the one that I was talking about. Automation created both disruptive effects and also very major distributional changes in the British economy. To understand why this very radical change in technology did not bring anything like the uh uh uh like the productivity bandwagon, it's useful to go back to uh to my schematic representation on the left. Productivity rises, workers also benefit. Would that link always work if we had, for example, competitive labor markets? And the answer is no. Even when productivity rises, there is no guarantee that workers are going to benefit, and that's comes down to what economists would describe as the difference between average productivity and marginal productivity. And I could talk endlessly about average marginal and average and marginal productivity, and that's not what you want. So, let me illustrate it with a story, which is uh famous and is ascribed to many people, so I won't even uh give the name of who I think may have come up with the story first. But, it goes it says the uh factory of the future will have two employees, a man and a dog. The man is there to feed the dog, and the dog is there to make sure the man doesn't touch the equipment. So, the uh humorous part of the story is, of course, that the man and the dog are completely dispensable. You don't need them. So, this factory, or as we move towards this factory, average productivity is going to increase. We're going to produce more and more and more with the labor that we have. But, the marginal productivity, meaning what the laborer laborers contribute to production is trivial. In fact, you don't need the man and the dog or any other employee in this uh dystopian or utopian factory, depending on which point of view you look at it from. So, if indeed we have technologies that increase average productivity but sideline workers, then there is no guarantee that those are going to benefit workers with in terms of higher wages or they're going to be a foundation for shared prosperity. And what are the technologies that actually create that wedge between average productivity and the contribution of labor? It is exactly automation, things that sideline labor and take away tasks from labor.
Now, all of this might be history, too, because at the end of the day, those who have a more optimistic take on the Industrial Revolution are correct because at some point the wages started increasing, and by the end of the 19th century, real wages were already on their way to being much higher than they were in the 18th century in Britain. But, those lessons are not completely irrelevant because we have tensions in many countries today, especially in the US, that are reminiscent of those disruptive phases of the Industrial Revolution. So, in this picture, I am plotting the evolution of real wages, inflation-adjusted wages, for 10 demographic groups, men and women, and by education, going all the way uh from the top in dark blue, workers with a postgraduate degree, to those at the bottom in red, workers without a high school degree. The uh years that uh preceded the mid-1970s are actually a the best argument for showing that there is no law of greater inequality in market economies or anything like that. They show actually declining inequality and broadly shared prosperity. Real wages for all demographic groups increased by about 2 and 1/2 percentage points every year in inflation-adjusted term for all of these 10 demographic groups. Here I'm showing you from one data source starting in 1963, but you can go back to 1949 and you'll see the same pattern even in a more striking way.
But then from around 1980, you see a strikingly different pattern. The top and the bottom are diverging, but even more jarringly, the the bottom curves, like the orange and the red, are actually declining, meaning that the real earnings of workers, especially men, but to some degree also women, who have less than a four-year college is actually declining during this this time period, especially until the late 2010s. So, this is a picture of the opposite of shared prosperity. Not only are the gains not being shared, but some How about half of the population is not benefiting and about a third of the population is losing out quite quite decidedly. Now, there is also no law that this is what industrialized nations should go through. If you look at around the industrialized world, there's a complicated picture. So, here I'll show you Sweden experience. I don't want to dwell on it too much, but what the Swedish experience shows is that Now, I can't of course read the the color, but the the curve, I think it's the blue one that's declining there, is the real wage wage inequality between the top and the bottom. In fact, during this time period, real wage inequality is declining in Swedish within the Swedish labor force, but overall income inequality is increasing because of changes in hours and changes in capital income. So, global globalization is bringing some other changes in the Swedish economy, but at least the the same measure that I was looking at in the US is actually showing the opposite pattern in Sweden.
So, this all uh amounts to the same question. What does it take to get what we saw in the US in the 1940s, '50s, '60s, '70s, and what does it take or what can we do to avoid what we saw in the 1980s, '90s, 2000s? So, I think the best way to understand that is actually go a little bit earlier and look at the early phase early decades of the 20th century. And and I think those decades, especially the automobile industry, which was, you know, the heart of the US industrialization, provides a very succinct summary of what I see as the two bookends of shared prosperity. One is that even though the automobile sector and then the other parts of the industrial technol- industrial sector introduced a lot of new machinery, some of them automating work, some of them completely changing the production process, such as electrical machinery, especially decentralized sourced electrical machinery, and you can see how the car cars on the one of Henry Ford's factories, the Rouge plant, are being made with electrical machinery that did not exist 10 years before, you also see that workers are still playing a key role. Ford factories were famous not just for the electrical machinery, but also for the whole engineering approach, which made many new technical tasks central to the production process. The way that the conceptual framework that I like on this, the one that I developed with Pascual Restrepo and is also at the root of the sort of the intellectual approach of the book with Simon, is that what these technologies were doing were they were generating new labor-intensive tasks, more sophisticated tasks where workers could start specializing in, especially as some of the earlier tasks that they used to perform are taken over by automation. The other element of this was that the auto industry was also at the forefront of labor organization. Uh that's the picture of the United Auto Workers strike uh uh for the GM, but there were similar strikes against Ford. And these were some of the events that changed the extent to which labor had voice both in wages and working conditions, but also in some very important episodes over how technology would be used and whether technology would be used just for automation or whether it would be associated with the introduction of new tasks, new titles, new training opportunities for workers as well. Interestingly, these two are exactly also the factors that you see in the British Industrial Revolution. It wasn't something automatic that took place in the middle of the 19th century that turned around the British Industrial Revolution. It was unionization, unions that were banned and heavily prosecuted became legal in the last quarter of the 19th century. It was democratization, and it was a change in the direction of technological change.
So, why did did things go wrong against this background in the digital age? Well, here is a picture of auto factory today. You see more fancy and better color machinery, but what's remarkable is that you don't see the workers that are doing the jobs. So, in simplifying it, perhaps overly simplifying it, the more recent digital-based technologies did the automation just like the Ford factories, but they didn't do the new tasks for workers. As a result, you had an automation bias towards inequality. Now, that may not be everybody's first explanation, so let me show you one picture to just frame that discussion. This is from work that I did with Pascual Restrepo. And what it does is that it now zeros in onto more detailed demographic groups rather than just gender and education. Now, I'm showing gender, education, age, and ethnicity. On the horizontal on the vertical axis, I have the cumulative change in real wages uh between uh uh between 1980 and 2016-2017, so before the COVID crisis. And you can see the zero line there, and you can see that about half of the dots are below the zero line. That was the pattern that I mentioned, the non-shared prosperity, where about half of the population is not benefiting from the economic change and productivity growth that's taking place. You can also see from the color coding that those are the low education groups. The more interesting element here, the new element here, is the horizontal axis, which is our measure of what fraction of tasks that this demographic group, the demographic group in question, used to perform in 1980 that have since been automated. So, you can see that for some of the low education demographic groups, that's about 25 to even 30%, and there's a very strong relationship between that automation exposure or automation's encroachment on a particular type of labor's tasks and how that group fared. So, by and large, automation seems to play a very important role in explaining about 60 to 70% of the inequality differences across demographic groups since 1980.
But of course, automation is not the only thing. Automation was enabled and took place in an environment in which labor was already losing its power, and that was very important for some of this automation to go on so unchecked and without any countervailing trends. Two forces were particularly important in this, although other people may emphasize other things like deregulation, but one and I don't literally mean that it's Milton Friedman who's responsible, but I put his picture there. Uh Okay, perhaps I am saying that. I don't know. Uh But the idea that businesses should be run just in the interest of shareholders without any regard for other stakeholders, especially labor, became started becoming very dominant. Uh Milton Friedman was a very articulate uh proponent of that view and pushed it a lot. And and business schools played an important role in propagating it to many important managers who then became powerful as well. That gave both cover and an additional impetus to many managers to reduce costs, and for many of them reducing costs meant reducing wages. And the natural force that could have resisted that, labor unions, were already in decline way before 1980, but their decline got further accelerated by Ronald Reagan's very visible take action against the professional air traffic controllers, which many other employers uh emulated.
But these two are also not enough. There is another factor which is going to help me bring the whole thing back to AI, which is that actually the distribution of power, the interests of managers, and automation are in turn also influenced by what are the technologies that are being developed. And I think both AI today and digital technologies over the last three three and a half decades have been driven both by financial incentives, business incentives, market incentives, and and and and ideology of the form that I mentioned from Milton Friedman, but they have also been driven by a tech ideology, which goes back to this brilliant mathematician Alan Turing's idea that, you know, autonomous machine intelligence is both desirable and feasible, and as autonomous machine intelligence starts developing because the machines are just functioning like the brain or the mind, that means they can start taking over more and more tasks away from humans. This, of course, reached its apogee in the early AI, which very much corresponds to what today we call AGI, because, for example, in the Dartmouth project that christened the field of AI, what these brilliant mathematicians and other scientists believed was that breakthroughs that would immediately bring human-level intelligence were eminent. And that sort of conditioned the technology field in a strong way to push towards more and more automation because A, you can do it. You believe that these digital technologies are going to develop higher and higher capabilities, and B, that also becomes a self-reinforcing dynamic that as the tech sector, you become more powerful, the more you push technologies and you develop technologies that can take over the world, so to speak.
But one problem is that it's not just inequality. Where is the productivity? So, if these technologies are so brilliant and we've having this innovation boom, where is the technology? Well, where is the productivity gains from? Well, the productivity gains are not easy to find in the aggregate or even in the sectoral data. Today, in fact, the US economy on the left is showing some of the slowest productivity growth as measured by economist favorite total factor productivity, and on the right, other industrialized nations are not that different. And in fact, I think that's largely because what's missing is productivity growth from automation because we are not using digital technologies the right way. Automation has been with us for two and a half two and a half centuries, and will be with us. There's nothing wrong with automation, but there is something wrong with excessive automation, and there is something wrong with just doing automation and not doing the other things that are complementary to workers or the new tasks. And I think what we have gotten into, and I think AI is at the the danger of exacerbating, is what Pascual Restrepo and I call so-so automation. You do more and more automation, but it doesn't really improve your productivity. You shift costs a little bit away from yourself to some of your workers, to some to other customers, but it's not a revolutionary productivity growth. And that's, I think, at least one possible explanation for the anemic productivity growth.
Now, of course, it would be very wrong to just focus on economic implications, because economic implications and political implications are synergistic symbiotic, and there are also very important fault lines in the political implications of AI. AI is an information technology, and as all information technologies, it can decentralize information or it can centralize information. There were hopes that it would decentralize information, it would start another fourth wave of democracy, but what we have seen is that often it is authoritarian governments that have incorporated AI into their strategy and used it for surveillance. And I think everybody recognizes the picture at the bottom is a credit social credit system where centralization of information is being used in a particular way by the Chinese government, but I think there is more parallel than first meets the eye between the use of centralization of information in the hands of the Chinese government and in the hands of tech companies. When tech companies decide for their objectives, which are very different from the Chinese government's objectives, about how they're going to centralize and how they're going to use that information, I don't think we can be completely sanguine and say, "Well, it's not the government, it's not the authority, so we're fine if it's Google or Facebook or Twitter that decides how that information is used, how that information is presented, and how that information is curated." And I think there are real dangers to foundations of democracy, including sort of consensus seeking and uh and and and and understanding nature and and implications of information when more and more information becomes in a centralized fashion with these AI technologies.
So, this paints a bleak picture, and in my last 2 minutes, I'm going to make the slightly more hopeful thing that this is not destiny. When you emphasize both power institutions and the direction of technology, what's at least at least superficially attractive is that those are all choices, with choices that we make collectively as human societies. But what's wrong with this picture is that it I think is a pretty uh accurate summary of how these choices are being made. They are being made right now by the most powerful tech leaders, and not really by society. When there are implicated are technologies whose implications are going to influence the livelihood of uh tens of millions of workers in the United States and billions of people around the world, well, I think it is too much to say it's okay for two competitive CEOs to battle it out and then see whoever going to win. And in the past, when there were things like this, the way that we somehow stumbled into dealing with them is by balancing the powers of large corporations, large powerful CEOs, and so on via countervailing powers. Civil society organizations, for example, Ralph Nader in the consumer protection space, unions, and also government regulators. Part of the problem is that those things have been much weakened today, but there is again no iron law that says once weakened, they cannot come back. If you look at US history, they periodically weakened and periodically come back.
But what I want to end with is to emphasize that I think it would be insufficient, and in my book, at least short-sighted to think that just this is just a power struggle. It's not just a power struggle. It's really, at least in my assessment, it's about directing redirecting technological change. You're not going to be able to change the effects of AI in a more beneficial direction by just saying, "Well, let's empower workers and regulators, and they're going to just make sure that the cake is divided more evenly." And and there are there were many activists, such as this person, Ted Nelson, who thought that personal computers could be used in a way to empower people and to weaken large organizations, and they were particularly worried about organizations such as IBM. Little did they know that IBM would be replaced by Microsoft, much bigger, much more powerful. Microsoft would be replaced by Google, and Google would be replaced by OpenAI. So, we are dealing with the same sort of issues that these activists were grappling with, but it's not just an activist issue. The vision of autonomous machine intelligence that has so animated Silicon Valley of late is not the only vision of technology. In fact, there is a division of technology, digital technology, and AI that was articulated as early as Alan Turing. One person who did that was Norbert Wiener, then MIT professor. And people who put that into practice were people like J.C.R. Licklider or Douglas Engelbart. And what's remarkable about this is that A, they understood that uh uh there was a philosophical uh approach that was fruitful of thinking of machines in the service of humans. What uh Simon and I call in our book machine usefulness rather than machine intelligence, but B, even more importantly, they gave us the demonstration. Some of the technologies that we absolutely depend on today, computer mouse, menu-driven computers, hyperlinks, hypertext, the internet, all came out of this agenda spearheaded by Engelbart, J.C.R. Licklider, and their students. It's only over the last 30, 35 years that this agenda has been completely eclipsed, and everybody has become crazed about artificial general intelligence and autonomous machine intelligence. And I think this is a much better vision, much better set of priorities for using technologies for the betterment of humans. And one question that people may have is this is all well and good. Perhaps Silicon Valley will do it, perhaps computer scientists will do it, but we as regulators, we as democratic citizens, we have no power in this. We cannot redirect technological change. And I think that's not correct. I'm sure Danny will have more to say on this. He has thought about this for decades as well, but there are many instances where you see a little bit of government policy, a little bit of citizen involvement can actually change the direction of technology in a major way, and the most recent one is in the energy sector. Uh uh for the electricity grid, renewable technologies were about 10 times as expensive as fossil fuels 20 years ago. With a little bit of pressure from civil society, a little bit of changes in consumer demand, and a little bit of subsidies, now they are actually cheaper for the electricity grid than fossil fuels. So, it is possible to redirect technological change in the energy sector. It is also possible to redirect AI. Thank you.
I just wanted to come over here. Okay, perfect. Thank you. I I I It's a great pleasure to to moderate this discussion. This was a wonderful talk. I I I think I'm uniquely qualified to be sitting on stage with Daron because having been born in the same country, in Turkey, and having sharing a common language, I can properly pronounce his name, uh which is Daron Acemoglu for all of you. And that's about probably the most important contribution I'm going to have to this discussion this evening. I'm a great fan of your book. I really admire not only the scope scope and it's it's it's it's ambition, but also how you're you're sticking your neck out in in a profession that is very much stuck in its own ways and you know, it's it's you know, you can you can you can be an outlier and be sort of you know, defy conventional wisdom but when you start attacking productivity as a concept, then you really have an entire economics profession, it seems against you unless you're of course Daron and you're writing of course the the articles that the economics profession is reading.
So the I have so many questions and and we have a very little time, but let me just start with something maybe a little bit um that might be one way of you know, summarizing your argument in a hostile way would be to say that you're rehabilitating Luddites who argue that sort of technology was bad because it was displacing people. Today we're actually talking in my class on industrial revolution. I talked about the etymology of the word sabotage, which was sort of you know, French workers who were throwing their clogs at the machines to to break them because they thought this technology was bad. So can you try can you sort of tell us why you're not a Luddite?
Perhaps I am. No, I'm kidding. Well, actually first of all, there are there is a disagreement about what Luddites were. So there are two schools of thought in history in British history. Sorry about that. One is that Luddites were completely anti-technology, anti-progress and they were just trying to defend the privileges of the labor aristocracy of many of which many of whom really did belong to that kind of labor aristocracy. A second view is that Luddites were a ragtag that had many different elements in them, but it sees in these sort of the beginning of social democracy and there were members of the Luddite movement that articulated ideas that weren't just against technology, that were about how technology was being used and who was controlling it. So if you go with the first interpretation, I am definitely not a Luddite because my take is not that technology is bad, but it's the technology's direction matters and technology's direction is a choice. It's a democratic choice, it's a choice by leaders, it's a choice by regulations and so on. And uh And I would even not say that according to this view, there is any simple argument that says you should resist technological change. What you should do is you should try to redirect technological change, which sometimes may involve slowing down certain applications of it. But if you take the second view of the Luddites, I think they were also trying to grapple towards this, but without of course the hindsight of 200 years of history and the tools that we have in political science and economics today.
I want to get back to the to the last slide that you showed as an example of directed technology and how a specific concerted public effort to to invest in renewables had dramatic effect on the cost of solar and wind and so forth. You did not mention that most of that was due to the industrial policies of a very authoritarian government, namely China. And you want to democratize technological change and the direction of innovation. You're a great believer in democracy. But how do you sort of square that with the fact that where these things have seemed to have been most effective, some would argue been in authoritarian settings as in China.
Uh So yes, I am a instinctive believer in democracy, but of course we have to be open to all sorts of empirical data. And I think by democracy by democratizing innovation, I certainly do not mean that we should have votes over what the direction of technology should be, but there should be a say of the citizenry in where priorities should be and what the objectives of certain major technological investments should be. Within that, I think I would still put my money on democracies for realizing that, but there may be some authoritarian regimes that also for a variety of reasons actually push technologies in that direction. That could be the case in AI, but in the case of AI, we see China is actually at the opposite end. For example, it's been the China has been as important a node in the surveillance technologies as it has been in solar panels. It exports it to about 60 countries around the world, most of them authoritarian. And and I think there is no guarantee from authoritarian regimes that their commitment to this are is going to be true and lasting, although you can say there's no guarantee from democratic regimes either. But in terms of the the the history of it is actually a little bit more complicated. First of all, China went big time into solar panels because of a German law. Germany created the first big program for subsidizing solar panels and Chinese companies took advantage of that in big time and that's how they started and then there was some sort of infant industry type dynamics. And then the second thing is I think you are absolutely right, solar panels solar panel costs declining in China has been an important part of that figure, but also innovation in the rest of the world has been very important as well. So if you look at during this period, I didn't show that picture, but there is a very clear trend break in the patterns between renewable technologies and fossil fuel technologies and some important breakthroughs were made in the United States, in Germany, in Switzerland and so on.
So what's what's going back to um So how we would redress some of the inequities that biased technological patterns create. What's wrong with the with the argument that the main the main issue is really the absence of appropriate redistributive institutions. I know you sort of you dismissed that, but I want you to expand a little bit on that. The argument could be made that um the unions for example, their first order impact really has been on redistributing the enterprise surplus that the redistribution to the tax system to the fiscal system and the welfare state is really has been has been what's been critical rather than their effect on redirecting innovation. And today, you know, of course a large part of the tech community would argue that the way that you handle these problems is not by mucking around with the direction of technology, but you know, through something like UBI, universal basic income. Let's give you know, everybody a basic living allowance. And of course if you're on the wealth, you would combine that with saying let's have a big wealth tax that's going to also finance that so that floor is relatively high. Why is that not the right way to go?
Well, you should be suspicious if tech billionaires, the extreme left and economists meet on something. So if you know, I think that there's some validity to that. And the economics theorem that says that's the right thing to do sometimes goes under the name of Diamond-Mirrlees, don't mess with the production structure, don't change the firms' investments, innovation, intermediate use decisions, just use exposed taxes for fiscal redistribution. I think it has a valid logic and it's that logic is very strong if everything inside the firm and if everything in the market is functioning very well. I think what I'm pointing out and I didn't expand on this from this angle, but thank you for giving me the opportunity. What I'm saying in some sense is that when it comes to innovation, that doesn't work. Innovation is not something that the market economies are pretty good at innovation quantity, but the direction of innovation, there are many biases. And if the direction of innovation is biased, then just doing the exposed redistribution is not necessarily optimal from an economic point of view. But I think it's also quite clearly suboptimal from a social point of view. And I think I would have many problems with UBI. First, I don't believe that the political economy as we have it today would support anything like UBI or or like that that would really give a comfortable way of living for the majority of the people. But even if it did, I think what you would end up in that society is a very, very sad and very disruptive status hierarchy. You would be in a society in which there's a small fraction, perhaps 10% of people who are the makers, who generate all the wealth, the remaining 90% are the takers that just live off the crumbs of that 10%. I think that would be such an unequal society from a status social point of view that it would actually make even today's society seem healthy. But there's another problem with UBI and all of these things. I find it defeatist. So, if indeed you believe that there is a way of redirecting technological change and that technology can be used, as I was trying to argue at the end of the presentation, in a more human complementary way to create new tasks, create new human capabilities, saying, "Oh, well, we'll deal with this problem with UBI." is defeatist because it throws in the towel. So, that's part of the one of the reasons why I try to argue against UBI whenever I can.
Um I'm going to ask one other question, but I also want to let the audience know that there is going to be some time for questions, not a whole time, not a whole lot of time from the audience. There are four microphones, two downstairs and I think two upstairs, although I cannot see the ones upstairs. So, if you have any questions, you know, just prepare and just by walking up to the to the microphones there. But while people while we're doing that, so on on the question precisely of ex ante regulation, I think the the most serious problem that economists have with the ex idea of ex ante regulation, with the kinds of issues you're talking about, is whether you can identify labor friendly or human friendly
AI or or other technologies in an ex ante way. Cuz there's a big difference, it seems, with identifying technologies that are going to be friendly to the environment, that's going to reduce the cost of renewables or to take another domain, technologies that's going to help the defense effort, the military, and so forth. Whereas when you're talking about this domain, we're really talking about what kind of technologies will increase the demand for labor, and that's an economy-wide thing. You can't just locate it at the level of individual sectors or technologies or firms. How would you sort of have you thought about how you would actually teach a Kennedy School class on this and tell our students how you would design the policy?
Policy. Uh, no, no, that's an excellent question. It's a completely valid concern. But let me make three points. First of all, today, yes, you're right, we know how to subsidize green technologies because we developed a measurement framework. In 1950, 1960, we didn't have that measurement framework. We didn't understand what was causing climate change and and and how to measure that. So, so it's it took effort and expertise in order to develop that framework.
Uh, second, uh, I think there are telltale signs that at least are suggestive about what are the technologies that would be more useful in the production domain. Both my sort of more academic work and the general discussion that I gave suggested it's about new tasks. So, in fact, under relatively weak conditions, if you are creating new tasks, new things that labor can do, uh, those are going to improve labor standing, although which type of labor, there's some complications there, but labor relative to capital, there's going to be a positive effect and general equilibrium effects are very unlikely to undo that.
And then third, you know, I think as opposed to, say, for example, uh, fossil fuels versus green, where it's very clear that today you don't want to invest billions more dollars in fossil fuel. You want to do it in renewables or other alternatives that have low emissions. Uh, so, every dollar that goes misdirected is a big waste. It's not the case when it comes to automation and new tasks. What I emphasized, I mentioned it, perhaps I should have emphasized it even more, automation is not wrong, automation is not bad. You want a lot of automation, but you want a lot of new tasks to go along with it. So, if we target new tasks, if we have a federal agency that provides seed grants or we have competitions that aim for new tasks, and 20% of that gets misplaced and taken by automation, that's not a problem. We want automation, too. So, as long as we do create this positive kick to new tasks, I think that would already be socially beneficial.
Wonderful. So, given the time, I'm going to try something unusual, which is actually take one question each from the microphones downstairs and then allow you to sort of give quick questions. So, please introduce yourself and then a very quick question over there. And a very quick question.
Hello. Thank you so much for joining us today. Um, I wanted to ask a little bit about the power dynamics between regulators and people that you start to advance in your book, um, namely that regulators need not be technological experts in order to be helpful in this arena. You also simultaneously advance the argument that technology is creating a sort of panopticon, um, where regulators have too much oversight into regular people's lives. So, I'm wondering as young people and students are increasingly aware that there's a gap in knowledge between who we elect and the people who actually understand the technology and this added sort of panopticon divide, how can we start to trust our people in Congress, our elected officials again?
Well, the thought on that. Let's take that. Thank you so much for being here. My name is Roy Shpaltzman. I'm in Master of Public Policy in year two. And my question is regarding our our own Professor Ricardo Hausmann's theory of economic development that depends on the export of diverse sophisticated products. What are your thoughts about this theory?
Okay, very diverse and very interesting questions. So, uh, you know, first of all, I definitely dread a world in which regulators have a lot of power. I wouldn't want a world in which regulators have too much power if there was an option that was different. But I think we have come to a stage where regulatory powers of the government in the United States and to some extent in the rest of the industrialized world is so weakened that you need to build sort of regulatory muscle. And I think it's quite feasible. You see that already that in the case of AI, five years ago, there was no lawmaker who actually understood even the basics of social media, smartphones, or AI. And today, at least their staffers, you know, they're not at the level of being top earners in OpenAI, but they are much more informed and then there is more informative policy debates on this. So, as in many things, I think it's going to be a delicate balance of power and we're going to get it wrong in some ways. Sometimes we're going to have too much regulations. Sometimes we're going to have enough regulation, but I think just the effort of the democratic process again starting to build regulatory muscle and regulatory priorities would be very useful. But of course, in a democratic society, it would be horrible if regulators were not accountable to voters and to politicians. And that brings you me to your uh second part of your question, for which I do not have any answer. How do we start trusting our politicians again? Good luck.
Uh, in terms of Ricardo's very interesting theory of development, you know, I think uh, you know, there are many parts of that that are not so directly related, but I think what's so what's what's relevant in some sense here is that I think his theory also emphasizes technological capabilities and the malleability of the technological capabilities. So, in some sense, one of the other sort of bedrock ideas that is important in my work and important in the book and it's the background of uh of what I talked about, and I hinted at it, but I should say it now more explicitly, I am arguing, as opposed to, for example, a sort of a naive reading of the current AI debate, which is there is a direction that we build, we get more processors, we get more data, we exploit scaling laws. It's sort of a linear progression. I'm saying technologies are much more scientific knowledge is much more malleable. You can develop them in many different ways. It depends what expertise you have, what priorities you have, what the market structure is. So, I think in general, theories like that, both in the context of economic I think are enriching for our debates.
I I I I think we'll abuse our time constraint just a little bit and just take two questions from um from up top. So, please, yes.
Hi, I'm Farid MPID. You will like this one, Professor. Do you envision a global governance structure that supports your argument in favor of a redirection of technology, empowering ILO, for example, or conditioning loans? Empowering, sorry, who? ILO. ILO. Yes, please.
Hi. So, I've been thinking a lot about consumers as workers. You talked about AI and automation, but I'm also thinking about data. People talk about data as the new oil, finance, railway. So, do you see a viable path for consumer data unions to act as a counterbalance against not just the concentration of power, but data among a few big tech companies? Um, I'm curious what your thoughts are on that. I know how you feel about UBI, so now I'm wondering about consumer data unions.
Yep, uh, in the last chapter of Power and Progress we discussed that and approvingly, uh, I think we do need data unions. And the argument is the following. Uh, so data is increasingly a very important part of the production process. Probably for the future it's going to be much more important than land. If I told you that, you know, all land is up for grabs, anybody can take land, and if Danny is using the land I can go and build on it as well, and nobody needs to do do the upkeep of land, everybody would go crazy. They'll say this is the most inefficient system. But we have created a system in which data is like that. It's up for grabs, nobody invests in it, etc. So, you need to change that. And the high-quality data, especially if you're going to try to do worker complementary technologies, high-quality data is very important. You know, worker complementary means, for example, uh, sorry, I'm taking a long ti- a little bit long to answer this question because it's very important for me. Uh, for example, it means we give, uh, information and capabilities to manual workers or to electricians or to plumbers to do their job better. But for that you need reliable high-quality data. Who's going to produce that data? So, you need to provide some sort of incentives, and I think the market system is a natural one. But individual ownership of data wouldn't work. You can recognize cats as well as I can. So, we're going to drive the prices of data down to zero if everybody can be, uh, sort of can compete against each other as individuals or they can be divided and ruled by tech companies. So, there is a very strong economic logic for data unions, and I think the correct infrastructure for data markets together with something like collective ownership of data could actually be both good for distribution of gains from AI, but also very good for generating high-quality data, which is an important input for the redirection of, uh, the AI direction. So, I think it's a very important. Thank you for raising that question.
Now, your question is a very important one. After every book I wrote, uh, I said, "Well, this book ignores a very important thing, which is the international dimension." And at one point I will come back to it, and I never come back to it. Uh, by using my, uh, freedom of moving to the next topic. So, I think the international dimension is super important when it comes to regulation of AI as well as it is for regulation of capital, regulation of globalization. It's very very difficult because we're we don't have agreement within a uh, within a country. How you going to do that between countries? How you going to find the common ground between China and the US? Uh, I I think it's not hopeless. I think China is always going to, uh, have a bias towards surveillance and control, but there are other places where China and the European Union could cooperate within limits. But I think all of this is going to need two things. One is international organizations. I think ILO would be very important for that from the worker rights perspective. The other one, even more important perhaps, is probably a new body that develops voice and perspectives from the developing world. There are about 4 billion people today who don't live in China, who don't live in the European Union, who don't live in United States or Canada, but who are whose lives are going to be completely affected as workers, as citizens, as dissidents by AI. Their voice is completely unheard. There is no medium for their voice to be heard. So, I think you we need something more than ILO, more than UN that really does that. But of course, uh, that's really a pipe dream because, you know, good luck, you know, having an organization in which India, Pakistan, Turkey, and Brazil sit and agree on something. But there's no stop to hoping.
Well, thank you, Daron. I'm afraid we have to to really stop here. Um, it's always, uh, a pleasure, uh, to listen to you. It's it's, uh, it's wonderful, if somewhat disorienting for me to find myself sort of to your right as I'm sitting now, sort of, um, increasingly, uh, the, uh, but, uh, yeah, I I I want sort of everybody, please, uh, join me in in thanking Daron for giving us a lot to think about.