How Liberal Democracy Can Survive an Age of Spiraling Crises

0:00I'm Dan Kurtz-Falen, and this is the Foreign Affairs Interview. Once you start thinking of new technology, just like a gale or a storm that is out of our control, then the best thing we can do is we can try to adapt to it. We build better skills, we build better protection mechanisms, etc. But if we are choosing the direction of AI, then an obvious question is, why not meet halfway? Why not, rather than just like push AI in some crazy direction and have humans adjust to it when it's not really that easy, why don't we try to steer AI in a direction that's actually quite useful for humans? I'm Deputy Editor Kanesh Tharoor.

0:43Dan is away this week. The world has reached various inflection points, or so we're often told. Advanced technology, such as artificial intelligence, promises to transform our way of life. In geopolitics, the growing competition between China and the United States heralds an uncertain new era. And within many democracies, the old assumptions, the undergirded politics, are in doubt. Liberalism appears to be in disarray and illiberal forces on the rise. Few scholars are grappling with the many dimensions of the current moment, quite like Darren Asimoglu is. The world is in the throes of a pervasive crisis, he wrote in Foreign Affairs in 2023, a crisis characterized by widening economic inequalities and a breakdown in public trust.

1:31Asimoglu is a Nobel Prize-winning economist, but his research and writing has long strayed beyond the conventional bounds of his discipline. He has written famously in the best-selling book Why Nations Fail about how institutions determine the success of countries. He has explored how technological advances have transformed or indeed failed to transform societies. And more recently, he has turned his attention to the crisis facing liberal democracy, one accentuated by economic alienation and the threat of technological change. I spoke with Asimoglu about a stormy world of overlapping crises and about how the ship of liberal democracy might be steered back on course.

2:23Dharan, thank you so much for joining us today. My pleasure, Kanishk. Thank you for having me. We have a lot of ground to cover, but let's zoom in before we zoom out. You know, you've received a lot of attention recently for your thinking around artificial intelligence. I think a lot of this was springing from a paper you published earlier this year in 2025. You know, it's very common now to hear big claims about how AI will transform the world or indeed AI will destroy the world. But you have a more deflationary take. In that paper, you wrote that AI will contribute only 1 % to U.S. GDP growth over the coming decade.

3:02That's 0.1 % a year. Why do you think the impact will be so limited? And indeed, why do you think the impact of AI has been so overstated? What I wanted to do with that paper was draw out the implications that very basic economic theory has about what the productivity gains from automation-based AI would be. And just introduce a note of caution because these automation-based technologies, you know, are pretty impressive in some ways. but they generally don't lead to huge improvements in productivity because, you know, what you're doing is you're replacing workers with machines and getting some 20, 30 % improvements in costs.

3:47But that's not like revolutionary. And many of these automation technologies need to be integrated into organizations, which is not easy. So those were the two pieces that I put together. And And the numbers came out to be quite modest. Now, do I have much confidence that it's 1 % and not 2 %? No. Do I have any confidence that it's 1%, not 50 %? Yeah. With that in mind, then, what do you make of the extraordinary amount of investment, attention, focus that the biggest companies, governments, areas of civil society, you know, the brightest minds are now directing toward AI? Well, I mean, I think it's a very promising technology.

4:30And even if the spread is slow, there are reasons for paying a lot of attention. So I tell all of my students and everybody that, you know, you should learn AI. And I have also no doubt that for better or worse, AI is going to have fundamental impacts on our society, on our economy, if not in three or four years, but in definitely 10 to 15, 20 years. But I think a lot of the investment right now is driven by a combination of hype and a racing dynamic, by which I mean that many companies are locked into a race that they think is going to be determining in that whoever takes the lead now will have a huge advantageous position in a couple of years time it is impossible in my opinion to understand the investments that open ai or google or anthropic are making in ai without recognizing this winner take all perception that they have.

5:40In particular, it's not possible to rationalize those with the profit opportunities that AI currently presents. OpenAI is generating some decent income because they are able to make some money from consumers using ChatGPT, but it's a very small amount relative to their huge investments. The only way you can justify that sort of investment is either if you think that almost overnight, you're going to have super high productivity enhancing improvements so that they can start selling AI or AI-based software and services to the majority of the companies in the United States, or because they think that they're going to win the race.

6:29And then once they win the race, they're going to dominate the next 20 years. I think it's a combination of the former, but there I'm not really optimistic because no company has produced very profitable production-facing, production-targeted applications yet. So it's really the latter that there are in this winner-takes-all race. In the thinking and the conversation around AI and technological innovation, we often sort of invariably run into this metaphor that sort of treats technological change like a disembodied force of nature. You know, it's like it's a storm or it's a tidal wave and it's up to us to build sturdier boats or to raise our houses on stilts.

7:16But in, you know, I think your last book, Power in Progress, which you co-wrote with Simon Johnson, you remind us that, of course, technology is not a force of nature. It's something that's mediated by states and societies and human beings, and that they shape how technology is adopted, right? When you look back at past innovations, you know, whether it's the plow, the steam engine, the telephone, and so forth, What can we learn about the way those technologies were adopted when we're thinking about AI today? Well, there are many important issues that you're raising. Thank you. I think the first one is that sort of this imagery of AI as an unstoppable gale, of course, contributes to the hype.

7:59So we should really note that. The second is the view that you hear all the time that AI is so different from the rest of technologies that humanity has grappled with that we have little to learn from the past. Then there's a third rhetoric that you hear sometimes, which is that in the past, we've done very well and everybody has benefited from new technologies. So therefore, we can learn something very high level from the past that technological change is going to be ultimately beneficial for everybody unless this time is different. So you see the tensions between these two narratives, but they're often made by similar people or the same people.

8:50But in fact, I disagree with both of them because, first of all, there is a lot to learn from history. Nothing is identical to AI, but we can learn about the mistakes we've made. We can learn how new technologies spread, about who wins and who loses and the political economy of it. But secondly, history actually shows that the adoption of new technologies has been a much more painful process, both because it's been very slow, and second, it has created huge winners and losers, especially when it has happened within a non-democratic setting in which a few companies or a few people could impose it on the rest of society by hook or crook.

9:34there were many downsides to that and workers often failed to derive any benefits. The industrial revolution in Britain is often given an example of, look, oh, we are so lucky. Yes, indeed, today we are very lucky, but it's like 300 years after. The first 80 or so years of the British industrial revolution, real wages fell, working hours lengthened, working conditions became horrible. so it wasn't like a pretty picture the satanic mills the satanic mills exactly you know life expectancy in manchester was 29 years old at birth amazing i mean can you imagine that today in the poorest country in africa it's like 62 or 63 so history has a lot of lessons for us we just have to apply them the right way because ai is obviously much more pervasive it's going to affect many more sectors things are going somewhat faster quite a bit faster than the factory technology ever advanced.

10:30But we can learn, first of all, that the birth pains of any technology are going to be more than what you might first expect. So weaving machines were amazing. They were really revolutionary. But it took them 30 years to spread sufficiently in the UK. And in that process, and once they spread, they also reduced the wages of weavers by two-thirds. in real terms. So it was a really big shock. So we shouldn't expect AI, if we're going to draw an analogy to weaving machines, to spread that fast. Same thing from electricity, same thing from computers. There will be exceptions. I think AI is spreading very fast as a consumer technology because it's entertaining and it's well marketed and all the hype helps.

11:18But many more people are using it just for asking advice about where to go for a restaurant than in the production process. And second, again, unless there is democratic oversight and the right sort of institutional adjustments and the right steering of AI technology, it might create big winners and losers. And that's really important because the promise that many people in the industry make is that this is going to be a really transformative technology from which we will all benefit. and again history puts a lot of caution onto that and the gale a metaphor that you started with has one more implication which i also like to take issue with once you start thinking of new technology just like a gale or a storm that is out of our control then the best thing we can do is we can try to adapt to it we build better skills we will better protection mechanisms etc But if we are choosing the direction of AI, then an obvious question is why not meet halfway?

12:29Why not, rather than just push AI in some crazy direction and have humans adjust to it when it's not really that easy, why don't we try to steer AI in a direction that's actually quite useful for humans, exactly as the skills that exist today would indicate? Of course, we have to build new skills. Of course, we have to upskill our population. Of course, we have to make sure that people understand AI and understand how to communicate with AI tools. But that doesn't mean that we should just think of AI as an autonomous force, which will go where it will go. And then humans just have to adjust to it either by putting up regulations or changing their complete psychology so that they are happy by staying at home and living on UBI.

13:09So it seems like a really strange way of thinking how society should adjust. I get a little skeptical, I suppose, of the sort of breathless presentism that often seems to accompany AI and the notions of its major transformations. I'm glad you do. Well, and I often, you know, I think this is where your historical view, I find it helpful. And I often wonder, you know, what would it have been like, you know, to be around in another period in the 19th century, you know, as all sorts of these new and frankly, miraculous innovations were spreading, electricity, the light bulb and so forth, the telephone.

13:50Do you think, I mean, this is a slightly whimsical question, but do you think people then, when they were looking to the future, were probably even more disoriented than we are now? That's what I would imagine. Because, you know, before the modern factory system, which really mechanized quite a lot of what humans did by hand, before that came on the scene, few people would have really forecasted that there will be those machines that act semi-autonomously in just a decade. Whereas if you think of current AI models, there's still no match to what, you know, science fiction writers wrote in the 1950s and 1960s.

14:36So this is not as unexpected. Okay, fine. Not many people were expecting this to come out of deep learning and next-world predictions. So there are some details of it that are very surprising. But the overall ambition and the overall shape that the world may look like with powerful AI isn't as much of a shock as probably the dark satanic mills coming up everywhere or machines work being just operated by the labor of children were to Victorian English or British. I wanted to return to the idea of a race that you brought up before. It's very common now to see in our pages and elsewhere a kind of panic that the United States may not innovate as fast as its competitors.

15:24It may not achieve AGI first, and that would be really bad for the country and so forth. But others make the case, and here I'm thinking about a recent book by the political scientist Jeffrey Ding called Technology and the Rise of Great Powers, that what matters is not so much who innovates first, but who distributes technology best, that we should be focusing on the diffusion of technology as much as, if not maybe more than, questions of innovation. What do you make of that case? Well, there is some validity to that, but I think there is more to be said. First of all, I sometimes joke that if China did not exist, the tech industry in the U.S.

16:05would have had to invent it. It is their trump card against any regulation. and that case is made on a very flimsy argument which is that we are locked into this existential race to AGI which will get realized within the next few years and as soon as AGI is achieved then we will move to the ASI artificial super intelligence stage whoever gets to AGI will have a huge advantage to get to the ASI, and then whoever gets to the ASI will dominate the world. So that argument has so many assumptions, none of which is empirically bolstered or validated, and it has many pernicious implications. We have so many global problems from climate change to AI regulation to taxation and regulation of powerful multinational companies, pandemics, nuclear non-proliferation, all of these require global cooperation, especially between China and the United States, to some extent between developing countries and United States and China, but also Europe.

17:19And none of that seems likely when we emphasize this racing aspect, because we think, or lawmakers and tech leaders think, that this is a zero-sum game and we should do our best to destroy China rather than cooperate with China because of this race to AGI, race to artificial superintelligence. So it really narrows the space for any kind of international collaboration. And it is very dangerous. I mean, in the United States right now, we are in this very polarized environment. The only bipartisan topic is China bashing. And don't get me wrong, I've been a longtime critic of the Chinese regime. I think it's very depressive.

18:03It is not going to be able to achieve the developmental milestones that people think it will. And it has also used AI in a very peridicious way with surveillance monitoring and data scooping. But still, we have to work with China. That's the other major global power. And we cannot confront pandemics or climate change or any global problem without some sort of collaboration with China. So the racing mentality is really dangerous.

18:58and focus on export controls and tariffs and the like. What's your perception at the moment of the relative trajectories of both countries? Well, look, I would be lying if I said I am optimistic about where the U.S. is going, and we can talk more about that if you want in a second. But let's focus on China first. I think there are certain things that China gets right, and there are many things that China gets wrong, and there are some things that China is lucky in, which are helping it in this current conjecture. What it gets right is that infrastructure spending is very important, especially digital infrastructure.

19:38And of course, education is very important. And of course, preparing a large number of engineers that are well-trained, that's going to be critical. So just the differences in supporting industry and innovation with infrastructure in China and not doing so in the U.S. would be a very important point and could create some advantage for China. On the other hand, industrial policy is a double edged sword. It works sometimes, it doesn't work other times. The evidence is not completely dispositive here, but there are several papers that suggest that Chinese industrial policy comes with a lot of costs.

20:19It may have played an important role in kickstarting certain things, but it's not an unalloyed good for the Chinese private sector. Third, more generally in industrial policy, there's a lot of evidence showing that both in research, in academia, in other things, the top-down Chinese approach is very costly. It does not allow a competition of ideas. It gives a lot of implicit power to politically influential actors. So, for example, in a research paper I have with David Yang and Jia Zhu, we find that in Chinese academia, people distort their choice of topics very strongly in order to carry favor with whoever becomes dean or party commissar or other influential position holders in the academia.

21:17So those kinds of things are very, very costly. There is evidence showing that when there are public-private partnerships, the public part brings down productivity. So again, that's the reason why I'm saying that industrial policy in China has been a mixed blessing. On the other hand, Chinese digital industry, and especially AI industry, has benefited tremendously from access to huge amounts of data and from a cadre of very well-trained engineers. So U.S., and this is where the immigration debate comes in, U.S. has a real shortage of engineers. Even today when we are giving a lot of H-1Bs to engineers, and if you cut those H-1Bs, it would have even more of a shortage of engineers for the very simple reason that we've let our education system fail.

22:03China does not have a democratic education system. It's got a lot of indoctrination in it, but it has poured in money and it has supported both middle school and university degrees. And there are millions of engineers in China, and that's a big advantage. So if you look at, for example, DeepSeek, which shocked the world because it reached parity with the models that OpenAI had spent hundreds of millions of dollars developing, supposedly for much less cost, much very quickly and starting from essentially a zero background. If you look at the details, what you see is that DeepSeek did not innovate in terms of methods or new approaches to AI.

22:51They were using age-old methods that were invented by statisticians 20 years ago or approaches that Google and OpenAI had pioneered. But what they achieved was an engineering marvel. They were able to put these pieces together in a much more efficient form than American companies were able to do. And how did they do that? That's where they used their engineering prowess. And that's probably where they benefited from the fact that they were on a shoestring budget. So when OpenAI wants to solve a problem, they throw another$300 million of compute power to it, Whereas they had to use limited compute power and find more efficient ways of processing models and data.

23:36So that is an advantage for China. And the future of China-U.S. competition will be shaped not just by issues like Taiwan, but how these advantage-disadvantage pair will play out. So the disadvantage being China's top-down, less free environment for research and the advantage being the supply of engineers and better infrastructure. And then what would it take now for the U.S. to get back ahead in that race, in that competition? Well, I think right now we are in the midst of, to put it bluntly, destroying every valuable part of the U.S. economic engine. So if I were to say, what are the things that make the U.S.

24:23economically special? I would count four things among the top list. One is the U.S. higher education system, which was the best in the world. It's being destroyed right now. Grants are being cut. Some of the most experimental areas, including mRNA vaccines, are being completely shut down everywhere. Universities have been declared to be the enemy of the administration. It's a bloodbath. Did universities not make mistakes? Of course, they did. Were some of those mistakes infuriating? Sure. But, you know, the way to deal with that is not to destroy that huge advantage. Second, because of the aforementioned shortage of talented personnel, which is partly to blame on our crumbling education system, especially in K-12 education, the U.S.

25:30relies on foreign engineering talent. And if you look at the innovators in the United States, those people who get patents or found new companies or spearhead the most radical innovation, immigrants, often coming with technical backgrounds, are very strongly overrepresented. So politicizing the entire immigration debate is also damaging that. So, again, were there mistakes in the Biden administration on immigration? 100%. Huge mistakes. But none of that was really about whether you would block engineers and new talents coming in and create a disadvantage in the industry. Third big advantage, innovation.

26:16The U.S. has been at the forefront of economic growth for the last 35 years because it's been the most innovative economy. That has come with some costs. Silicon Valley has made mistakes, created inequality, has created a beginning of a two-tier society. But let's not make any mistake that the U.S. economy wouldn't be where it is if it wasn't so innovative. But what is innovation? Innovation is a very forward-looking activity. People of all walks of life start things in their garage and sacrifice time and resources because they think that they have the talent to create something new. And once they create that new thing, at some point, other people are going to see it.

26:58They're going to raise funding. With that funding, they're going to be able to scale up. With that scale up, they're going to be able to get workers. And with that, they're going to be able to compete against established companies, some of them with hundreds of thousands of employees. When they run into conflict, the courts are going to side with whoever is right rather than crush them. And they're going to be able to reach global markets. All of this has horizons of 20, 30 years sometimes. Why do people have that sort of confidence? Well, the answer is very clear. They don't have that confidence in, you know, Turkey or in Mexico or in Brazil because U.S.

27:35traditionally has had legal institutions, court systems, procurement systems, broad property right protections that made it sensible for people to believe that their investments would bear fruit and they would enjoy those fruits, that the courts would defend their rights. and they would not be crushed by rivals. And you know, what is the Trump administration doing? It's completely destroying those institutions. Now, all institutions are much more biased. Courts are not impartial. Companies that are allied with the Trump administration get special treatment. So you should have, rationally, much more doubt when you're thinking of devoting your time and money in your garage.

28:14The other big secret sauce of the US economy me is that we have had tremendous advantage because we can borrow very cheaply from foreigners. This is why the financial crisis was spearheaded by U.S. banks and all the other countries did worse than the U.S. because investors turned around and put more money in the U.S. corporate bonds and treasuries and stock market because they thought U.S. is the safest place. same thing during covid we were able to borrow and invest much more than many other countries that's why covid is much softer for the u.s as a negative shock than other industrialized nations why are people willing to put their money in the united states again because of institutions would you want to put five million dollars into the chinese companies and then end up in chinese courts then not be able to turn that money back into whatever you want to do whenever you want to do it?

29:12No. I think, again, that's an institutional problem. And the U.S. institutions being under attack, and especially courts being now much less effective and much less able to play that role, I think people are going to start rebalancing their portfolios with very important negative effects on the future of the U.S. economy. So on all of these grounds, I'm very worried about what's going on. And I'm also worried that some of these things are going to be very hard to reverse, even if there is a different government in three years' time. So this is a very important thing. I'm not going to say China is going to win the race with the U.S.

29:48because there's so many factors, and I'm hoping that U.S. doesn't become stuck in this new equilibrium. But the costs are not to be belittled. Speaking of institutions, probably the book of yours that is still the most well-known is Why Nations Fail, in which you and your core authors make a long argument, a long historical argument for why countries succeed and why countries don't on the basis of their institutions. And you focus on how institutions help distribute power within a society and prevent its consolidation. And you contrast those societies that are inclusive institutions with those that have extractive ones.

30:29The extractive ones being ones that invariably lead to failure, inclusive ones being the ones that have a higher chance of leading to success and stability. Now, the example of China, especially after 1980, is often raised as an example of a potential sort of outlier in your thesis. Now, you have suggested just now that, you know, you think that China will face certain headwinds in the future, born out of its own system, and its inefficiencies and so forth. But suppose the Chinese model proves to be somewhat durable or more durable than you anticipate, would that present some kind of challenge to the thesis about institutions and their role?

31:09Yes, absolutely. 100%. So just a second ago, in my somewhat long answer to your previous question about U.S. economy, I essentially articulated an institutional view of the U.S. success and the current failures to the U.S. Now, if the United States were to fix its institutions and go back, you know, to where it was, say, in the 1970s or 1980s, which were not perfect, U.S. institutions always had imperfections, poverty, racial discrimination, much greater inequality because safety net type arrangements are weaker. But if we went to where we were, and despite that, China surged ahead and became a more innovative economy than the United States, that would be a challenge to the institutional thesis.