673. The World’s Progress Has Stagnated & AI Might Not Change That with Carl Benedikt Frey

Professors FM.

You're listening to the Unsiloed podcast with Greg LeBlanc, produced by University FM. Unsiloed is a series of interdisciplinary conversations that inspire new ways of thinking about our world. So wherever you are, enjoy today's episode, and here's your host, Greg LeBlanc. Welcome to Unsiloed. This is Greg LeBlanc, and we're here today with Carl Benedikt Frey for the second time because of the issuance of this new book called How Progress Ends: Technology, Innovation, and the Fate of Nations, which follows up on your book Technology Trap, which we discussed last time. And I should mention that Carl is a professor of AI and work at University of Oxford at the Oxford Internet Institute and also the Oxford Martin School, also a fellow at Mansfield College and also at Lund University. Welcome, Carl.

Good to be back, Greg. Thanks for having me.

Well, we were just discussing how you are an economic historian who studies history all the way up to the present. And so I studied economic history because I was interested in understanding the present and looking to the past to provide some guidance because the past is in many ways a sequence of case studies, right? I think this is a little bit different than the way a typical historian looks at things where everything is unique. But if you're a social scientist, you're looking for patterns, you're trying to find models that work over different time periods in different places. And in this book, you talk about a dyad which has appeared and reappeared in so much of the literature, right? That between, I'm not going to say markets and hierarchies, it's more centralized and decentralized organization of economic activity and idea generation, and you liken them to the explore and exploit trade-off, which shows up in organizational behavior all the time. And if one were to ask, is it better to organize a society in a decentralized or a centralized way, you're going to say it depends, right? And so you flesh out the Hayekian insight around decentralization, but also the Chandlerian insight about centralization. And I think in many ways you seem to be channeling Gerschenkron, who I remember reading in grad school, and I think he's sort of faded off the map, but he was, I think, the original economic historian of catch-up and how if you are a second mover, it may make sense to use a different form of strategy than if you are the first mover as a country. So were you drawn to this primarily because you were thinking about the China versus US situation in the current moment? And then was that what motivated the desire to go back and reexamine Prussia and the US?

Well, I think that's certainly part of it. So a key theme of the book is trying to inform the present, and obviously every time it's different to some degree. The cost is always different from one thing, but in the end of the day, we're humans and we tend to react in similar ways. to similar challenges, although some episodes rhyme better than others. Another reason for it though was I'm an academic and I'm somewhat selfish in some degree. I think actually about what do I want to study myself and learn more about and think about, right? And so this tension had been on my mind for a long time, and I thought it was timely because exactly as you described, the emerging tension between China and the United States, and the view that Chinese state capitalism is superior to what we have or see in the United States and Europe today. And I think Chinese state capitalism, if you can even call it that, is grossly misunderstood for the most part. And so just grappling with that part in itself was an interesting experience.

Right. And I think the big debate now, everyone is concerned that China is going to catch up to the US and is going to be a formidable rival. But I mean, here in Silicon Valley, even though there are people who are panicking about DeepSeek and so forth, I think most people aren't particularly worried about China because they think that it's impossible to generate the kind of innovation that we get here through the centralized Chinese system that Xi Jinping has And there was a flowering moment in the early 21st century where China opened up and became more decentralized, and you saw the rise of Alibaba and everything. But the recent crackdowns would suggest that they're not going to be at the forefront of innovation.

I think a key insight of the book is that there are many different Chinas, right? And so unlike the Soviet Union, which was managed centrally from Moscow, every industry from steel to rail was managed centrally. In China, provincial governors actually have quite a significant degree of autonomy, or at least have had up until recently. And so from the opening in 1978 onwards, different provinces pursued different strategies. So Beijing started pursuing more top-down industrial policy. In the south, economies began opening up as special economic zones. And then Deng Xiaoping saw the results of that firsthand during his famous southern tour in 1992. And decided to pursue that path more extensively. And so there was a lot of experimentation going on at the local level. There was not a detailed plan for everything. And I think CCP members were surprised by many of the outcomes and have indeed admitted so themselves. And so when we compare China to the United States or to Europe, I think it is important to remember that competition is very much a part of the Chinese model, right? And so you have on the one hand provincial governors competing for positions inside the Communist Party by meeting growth targets, and that often being an escalator for promotion. You have most of the economy which is in private hands, many foreign invested companies, those firms on balance tend to be the most innovative ones. And then industrial policy in China can actually often be pro-competitive because you have all these local initiatives competing against each other. And so that's very different from Europe where we see now with German rearmament, Rheinmetall is gaining most of the contracts. In fact, it's very hard for a startup to compete for any of those contracts because you have to be profitable to be eligible for any of those contracts in the first instance, which excludes most startups.

Right.

And so when Europe thinks of industrial policy, it's often plowing resources into incumbents. In China, competition is very much part of the industrial policy landscape. Although I think there are open questions around industrial policy becoming more centralized around Beijing in recent years, there are also a lot of questions about the Chinese economy in general. It's certainly been successful in some key industries like solar, like electric vehicles, even pharmaceuticals, where 10 years ago everybody would point to that sector and say, look, here's an example of the Chinese can't innovate. Name one drug that they come up with. That is clearly no longer true. And so you would struggle to make the case that there is no innovation in China, but in a way, It's also not that surprising, right? It's a billion-plus people that are highly skilled and they have some freedom and autonomy and they're innovating. Just like it would be very surprising if a firm like Google or Amazon doesn't do any innovation because of the amount of talent they employ at the end of the day, right? And so the question is, is that Chinese system a better system or not? Well, argue no, and we can go into that, but it certainly has some key ingredient components that are conducive for innovation and much more so than the Soviet Union ever was.

Well, while there might be some debate now, it seems like there wasn't a debate in the past, right? Because there was a time when China was clearly ahead of Europe in terms of its economic development and in terms of state capacity. And so forth. And also in terms of the innovations that were being generated. And then of course the torch was passed, and I don't think there was any debate or dispute, right, that Europe took the lead. And so this great reversal, I mean, we— I've talked about it in some other podcasts, but I'd love to hear your retake on this great reversal because we do often forget how far ahead China was, right, not that long ago in the grand historical Absolutely.

Yeah. So China had a very significant technological lead over Europe before 1500, certainly, maybe even somewhat past then. And then Europe certainly races ahead during the period of the First Industrial Revolution.

I mean, I think people even forget that China was exploring the world. I mean, they were sending ships all over. I mean, they could have been the ones that discovered America if they had kept going on that trajectory.

Absolutely. Yeah. But I think a key difference is that China used technology to build a state bureaucracy. And so China, for the most part, developed and pursued technology that aided the state, whether that is soil mapping that better allowed elites, rulers to assess the potential value of what the farmer could produce and tax them accordingly, whether it's a printing press or a unified writing system that allowed them to build elaborate population registers of their subjects and communicate across the vast empire. And so China pursued an intensification of agriculture that Europe could not. In part also because of geographical reason that China had more consolidated land, could pursue more economies of scale there and intensify the use of technology. But that made, just like mass production made manufacturing output more measurable and more predictable, that intensification of agriculture made agriculture output more predictable and easier to tax. And then that rise of centralization enabled by technology, because without that technology, you can't monitor your subjects, you can't set taxes. And so if you can't do that, then you have to delegate more power to the local level and permit lords to be in charge of taxation, which essentially gave rise to the feudal system in Europe. And so that meant that European politics were much more decentralized even during the times of the Roman Empire, but certainly after its collapse. And that decentralization then actually opened the door for the Industrial Revolution, as Walter Scheidel has argued very eloquently in his book, escape from Rome. And there are several reasons for that. One is that intellectuals and innovators could escape from the grabbing hand of government or tyrannical rulers wanting to clamp down on intellectuals. So the likes of Voltaire, Descartes, and others moved around frequently for those reasons. It also meant that when Louis XIV revoked the Edict of Nantes and made Protestantism illegal in France, the Huguenots that were certainly a very skilled and thrifty population, they brought their skills to places like Germany and the Netherlands, Switzerland, and the United Kingdom. And that led to an upsurge in innovation in those places. And then you have autonomous cities, which can become hubs for institutional experimentation as well. And so none of that happens in China. There are no autonomous cities, no autonomous universities, and there is no scientific revolution the way that it takes off in Europe. In fact, as Europe goes through its scientific revolution, China goes through a literary inquisition, and that divergence then becomes even more prominent during the First Industrial Revolution.

Right. And of course, you talk about the rise of the bureaucracy, or the Mandarinate, as we call it in China, and it seemed like anyone who had ambition or anybody who was intellectually inclined, that was sort of the only career available to you was the bureaucracy. And it was indeed meritocratic, right? So it wasn't corrupt in the way that we might suspect, but it really kind of sucked all of the intellectual potential of the country up into these bureaucratic pursuits, which included the pursuit of— some kind of rarefied knowledge, right? Which wasn't particularly practical.

Yeah, that's an important point that I also should have made is that the incentives were such that to study for the civil service examination, you had to memorize the Confucian classics. There was no emphasis on science, technology, and engineering. And it was the way that you could move up in the social hierarchy. It was the best escalator for social mobility there was. And so if Galileo had been born in China, he would probably have been a bureaucrat rather than a scientist, and such were the incentives. And I think it's interesting, and I think it's a point worthwhile making, that really long-lasting legacy of a meritocratic civil service is still there in China today. And it's still there in East Asia more broadly, and it's still there in the countries that we think of as being the most successful in industrial policy. Because one thing that you do really need if you're going to pursue industrial policy is the meritocratic civil service that is relatively shielded from political pressures, that is competent enough to implement policy and get things done. And so If you look at places like South Korea, like Japan, like China today and ask the question, why did they have these growth miracles that we tend to associate with industrial policy and not Latin America, for example, that also pursued industrial policy? I think there are several reasons for that, but that's certainly one very important reason.

Well, as someone who's studied economic history, I mean, I think in the Anglo world we spend a lot of time on England, of course, and maybe a little bit less time on France, but Germany seems to be a blind spot, right? And Germany went from being fairly backward, or Prussia went from a fairly backward agricultural economy to one that ultimately could compete head-to-head against England by the early 20th century. So it offers a very different model of development. I was wondering if you could talk a bit about that and a bit of going back to Frederick, but certainly the Hartenberg-Stein reforms.

Yeah. So I think an important distinction with Prussia and Britain is that Prussia essentially develops a state before it industrializes. And so the Prussian state shapes the industrial revolution that takes off in Prussia to a much greater degree than the state ever did in Britain. And it does so in a number of ways, and it is primarily driven by geopolitical competition. And so what sets off the Stein-Hardenberg reforms that you mentioned are significant Prussian losses at the Battle of Jena in 1806, where they get beaten by Napoleon really badly. And the drive to modernize really takes off after that. And so part of that package is certainly liberal reform, and so secure private property rights, abolishing the guilds and the craft guilds is a really important part of the package, as had also been the case in Britain, as had also been the case in France. And this happens earlier in Prussia than in Bavaria and Baden-Württemberg, in part because the state is sufficiently strong to actually override the guilds, whereas in Bavaria and Baden-Württemberg, The state relies more on them for tax collection, among other things. And so overriding vested interests, I think, is a really important and underestimated role of the state that you see both in Britain and in Prussia, because the Luddite riots are fiercely suppressed by the government.

You need a strong state. To actually liberalize.

Yes, sometimes you certainly do, for the most part, in fact. And then beyond that, Prussia shapes industrialization in more direct ways. And so through a relatively meritocratic civil service, it creates a pool for recruitment also for private enterprise. And so firms like Siemens, as they emerge, begin to hire civil servants. To run these companies because they have managerial skills that are transferable. The state begins to seed educational institutions such as technical universities and business schools and play a very important role in that. Prussia under Bismarck even nationalizes the railroads, and even before then, they take a much more significant role in rolling out rail, which had essentially been a completely bottom-up matter in Britain. But there are then certain things that the state cannot make disappear, right? With, for example, cultural cleavages between some of these Bundesländer, which had previously been autonomous states, and cultural cleavages between Protestants and Catholics, for example. And so whereas Britain had a much more dynamic and harmonized market for technology, if you look at patent assignments, for example, in Germany, the effect of the border between the Bundesländer is basically as big as the effect of the border between Germany and France. And so these things do continue to matter, and they probably help explain why Britain continues to have an advantage in Older industries, but as technology becomes more science-based and as Prussia invests much more heavily in educations and develops new educational institutions where you know ministers of education like Friedrich Altman takes a really leading role in everything from developing what's known as the Max Planck Gesellschaften today. focused on cutting-edge research to improving elementary schooling, you don't have a counterpart to that imprint.

It makes sense if they're going to do catch-up and copy the technologies, then this centralized industrial policy might work. But Germany became quite innovative by the end of the 19th century, as you mentioned, right? Dyes and steel and a lot of other industrial areas. How does that flow from the centralization? I mean, was that really Due to some decentralized elements within the system that they'd established?

So I think a key difference when it comes to educational policy, which I didn't mention with Germany as well, is that Germany has a federal system, right? And so whereas Oxford and Cambridge basically have a monopoly on education inside Britain, and whereas educational resources in France overwhelmingly flow to Paris, you have the states competing against each other in higher education as well. And that competition even intensifies then with Max Planck Gesellschaften, for example. And so I don't want to push this parallel too far, but it has something related to the decentralized competition that you see in China as well, and the government playing a more active role in that. And so the state in Germany played a role in seeding many of these institutions. And then the competition between them is really important. But I think it's also important to remember that the German model would not have functioned without the liberalizing reforms that I mentioned earlier. So private property was no less secure in Germany than it was in Britain. The guilds were quashed in the same way. And so Germany was very proactive in removing impediments to industrialization. And so I think that's a really important part of the package too.

Well, now Russia in many ways had a similar starting point to Prussia, but they never managed to enact any of these reforms in any kind of serious way. There seemed to be some attempts, But these attempts never really panned out. What was the key difference between Prussia and Russia?

So key difference here is between despotic and bureaucratic power, right? And so you see this to some degree even today. Vladimir Putin has despotic power. He can remain in power as long as society is weaker than he is, but he doesn't have anything like the one-party state that the Chinese have and the meritocratic civil service to implement policy and get things done. And so Russian autocrats essentially survived by making sure that society was weak and divided. And so the tsars for a long time resisted the construction of factories, banned railroads, Wary that they would import the revolution, which was almost prophetic in the sense that 7 decades later, Lenin would actually arrive by train from Zurich. And so there was a reason that these technologies were resisted and the tsars were not irrational to do so. But then at the same time, they wanted to be able to compete with the European powers in the international arena. And you couldn't really do that without having modern industry, having capital investment going into new technology, and having engineers and talent that is capable of building railroads and running factories. But Tsarist Russia tries still to control private enterprise to a different degree. And so rather than incorporation by registration, which becomes the norm in Western Europe and the United States and much of the world. They have a very complicated system where essentially the autocrat needs to sign off on any business being formed. And they managed to attract the Nobel family from Sweden and others to build up Russian industry, and they created some pockets of progress. But for the most part, The state survived on repression and did not take on the same role as facilitator as you see in Weimar Germany.

Well, I mean, it seems like the visible hand or centralization can arise not just from the state, but from large private enterprises, right? And as they get larger and larger, they become somewhat state-like, right? And so you talk about how antitrust played a role in the US of, in many ways, preventing that process from following that course. Can you talk a bit about that? Because it seems like the policy in Germany was very different, where the government seemed to encourage consolidation rather than discourage it.

Yeah, no, that's right. And so as businesses grow larger, often they became part of the political system because they're Influence on policy increases as they grow. And so they usually don't just sit back and take the rules that are handed out to them. They actively want to shape it. And so when the state is strong, it has better capacity to resist that for good and for bad. And in the case of United States, from the onset, wary of the fact that the United States was founded essentially as a revolution against centralization, centralization under the British crown. German-style, Prussian-style industrialization was essentially not an option in the United States from the onset. And so direct taxes are prohibited. Already in the Constitution. Income tax is only instituted in 1913. What is there from the onset are restrictions on any barriers to trade between states, and that creates a large harmonized market within the United States. And then you have secure private property rights and a patent system, which is relatively democratic. Compared to its European counterparts, because the fee for filing a patent is much lower in the United States, and that leads to much greater participation in innovation from the onset. And so America creates a market for technology that's really unmatched anywhere in the world. And so around 95% of patents around 1880 are taken out by individuals, and around 85% of all patents are traded in the market for technology. And so the railroad companies didn't invent the railroads, telegraph firms did not invent the telegraph. Essentially, it was individual inventors that did so, and they licensed their technologies to firms, which then took those technologies and developed products around them. And so even AT&T, DuPont before World War I, their R&D departments were essentially screening the market for patents rather than innovating internally at all. I mean, they basically had no in-house R&D. But as these companies then grow larger and as the capacity of the American state is quite constrained, this becomes increasingly challenging, not least because the federal government becomes very dependent on some of these companies in doing big infrastructure projects like rolling out rail. And so that in the American context happens through public-private partnerships where the federal government induces the construction through land grants and through loans, et cetera. But the construction is done through private firms, of which some become larger than the federal government itself. And then you have number of corruption scandals, not least around Crédit Immobilier, which makes the headlines and highlights the corruption that is going on as a result of these public-private partnerships. And because the political system in America at the time, without any meritocratic civil service, essentially means that politicians are trying to mobilize voters by promising them jobs rather than trying to get them interested in various policy issues. And so future President James Garfield, who also becomes ensnared in the Credit Mobilier scandal, he's famously assassinated because of somebody who thought he should have received the job as consul To Paris. And that then mobilizes the American public, leads to the Pendleton Act, which then finally institutes a meritocratic civil service. And that civil service then eventually gets the capacity to begin to regulate these giant companies. A big part of that is the Sherman Antitrust Act, which gives the federal government the tools to clamp down on anti-competitive practices, but crucially also means that size per se is not an offense. And so it means that firms can grow large as long as they do so organically and without anti-competitive measures. And that balancing act, I think, has been really important for American dynamism.

Well, I was wondering if we could talk about Asia a bit because I'm old enough to remember when there was a panic about the rise of Japan, right? In the early '90s, everyone thought that Japan was lapping the United States in terms of technology and in terms of industrial prowess. And it seems like Japan did this catch-up twice, right? I mean, once under the Meiji Revolution and then again after World War II. But they've sort of stagnated somewhat since. And it seems like the Koreans in the '90s had a reckoning where they disbanded, or at least kind of kneecapped the chaebols in a way that Japan never had this reckoning in the '90s. And so how did the catch-up model of industrial policy of those 2 countries bring them to the state that they're currently in? And Why this divergence in ultimate outcome in the recent decades?

Yeah, that's a really important point and I think an important story, right? And so Meiji Japan catches up quite rapidly with the technological frontier in the 19th century, in large part because of symbiotic relationship between the government and the emerging Zaibatsu conglomerates. And after World War II, The Allied occupation authorities under MacArthur, they tried to break these Zaibatsu up, instituted anti-monopoly law, but that is very soon relaxed and the Zaibatsu conglomerates soon morph into these Keiretsu conglomerates, which are also a really important part of the Japanese miracle to the degree that in the 1980s when PBS Frontline runs a documentary about Japanese growth, Sharmila Johnson remarks that the Cold War was over and Japan won. Well, I guess it wouldn't have been in the '80s, it would've been around 1991. And so I think that nicely underlies the sentiment of the time because in the 1980s, many believed that what Japan, what Toyota had done to General Motors and Ford, Sony was doing to American firms in consumer electronics, America was losing the lead in semiconductors, et cetera. And obviously it didn't quite turn out the way as many people anticipated with the benefit in hindsight. And I think in large part for 2 reasons. And so one reason is that by integrating with the China circle, with mainland China, Taiwan, and Hong Kong, American firms essentially undercut Japanese production advantage because labor costs in China at the time were just still much, much lower than they were in Japan, even if production processes in Japan continued to be superior. And secondly, whereas these Keiretsu conglomerates are very good at scaling, at sharing technology between them or within these conglomerates, I should say, The barriers to entry that they erect are very high. So you see very little dynamism in the Japanese case, whereas the much more competitive American model has the advantage of facilitating entry. And so the lawsuit against IBM, for example, forces it to unbundle hardware and software. It opens up the market for firms like Microsoft to come in. The breakup of AT&T leads to an upsurge in innovation. It also means that when the National Science Foundation releases ARPANET to the public, it's not handed over to the local or the national carrier service, which might have bottlenecked it in its boardroom if it had seen no immediate commercial use for it. And there was very little immediate commercial use of ARPANET at the time. And it's also quite different, you made the contrast to Korea. And so The chaebol are very much structured the way the Keiretsu conglomerates are. But the key difference is that following the Asian financial crisis, antitrust becomes much more vigorously implemented and enforced in the Korean case. And what you see after that is an upsurge in innovation. primarily driven by non-shaible companies. And I think that helps explain the relative continued dynamism of Korea relative to Japan as well.

And you also talk a bit about the post-war scientific research that was sponsored by the US government, right? So in response to Sputnik, there was an initiative to promote research, but it was done very Differently, obviously, than it was in the Soviet Union. And it could have been done internally within the Department of Defense, but instead it was more or less farmed out to universities. Why is that important?

Well, so there are very different programs out there. And so there's no way of denying that the Manhattan Project was successful in delivering the bomb before Nazi Germany. The Rad Lab advanced radar technology, the Apollo program put a man on the moon. And so all of these projects achieved their missions, but the key question is, what did they create beyond that? And so if you take the Soviet Union, for example, they were very early in satellites, right? And so Sputnik famously raised alarm bells in America that the Soviet Union was taking the lead. But that didn't usher anything like commercial telecommunications sector in the Soviet Union. And so having a system where robust spillovers are possible is one part. Another is that if the mission is clear and can be clearly identified, then you can essentially grind it down from the top down. fashion. And so when you know what is to be done, the centralized explanation has relatively little value. But when you want to explore entirely new technology, such as ARPANET, for example, giving people the autonomy to explore and experiment is much more critical. And so Licklider, for example, he had famously already written this important paper before becoming an ARPA director. And as ARPA director, he had a lot of autonomy in terms of hiring the people he felt would be most important for the project, which was, by the way, true of the Manhattan Project and the Rad Lab too. He also acted within a decentralized system of American universities. And so there are a lot of places around the world that look at the ARPA model and say, hey, we should do that too. But I think it's very hard to do without the quality of American universities and the talent that you have there to draw upon. In part just because if you want to give somebody more autonomy, then they better be good at what they do. And you also have the decentralized competition that then feeds through ARPA as well, which I think explains why ARPA in particular ended up being so successful in spawning new technologies.

But look, I mean, the Soviet Union collapsed not just because it couldn't innovate, right? But also because it couldn't even copy, right? I mean, the system was not just dynamically inefficient, but it was statically inefficient. And some people have said that if the Gosplan had access to, say, AWS, right, you know, maybe they could have survived a bit longer. If Jeff Bezos was running the country, maybe they could still have some kind of central planning. I mean, does the modern computational capacity, does it make more centralized planning more feasible? If you think back to all those people who developed linear programming and everything back in the '50s and '60s over there in Russia, I mean, they would just salivate if they could have access to GCP or AWS or AI.

So I think there are different parts to that story, right? And so one is, okay, you have more information, you have more compute power, and in the limit with AI, if you have more and more data, more and more compute, the question becomes what out there that is tacit knowledge can you actually codify, right? Because not everything that is out there can be easily codified and therefore managed in a centralized way. The other part too that I think is just incentives, right? And so for example, if you were an aircraft engineer in the Soviet Union, you could go to the Red Army and ask for funding. If they declined, maybe 2 or 3 other options. If they declined, then basically you're Idea would die with you, right? And so you have these bottlenecks, and they are very hard to circumvent. And so during the computer revolution in the United States, for example, Bessemer Venture famously declined to invest in Google back in 1999. And that only illustrates that Google was not a safe bet at the time, right? Alta Vista and Yahoo. were dominating search. And the only way to know whether that would catch on was for somebody to invest in it and try it out, right? And without somebody running those trials, you will not have that kind of information. And so having the space to explore is really important regardless. And then often when you innovate in a system that doesn't really encourage it. You may even be punished for it, right? And so if you invent something new, it slows you down in the work that you're doing on a continued basis. You might even see your income fall as a result of that. Or if you become more productive, and then the targets that you may meet may simply increase, right? And so none of that is a result of whether we have more access to data and compute power. It's a question of incentives and the incentives to innovate in a broad-based manner in the Soviet Union didn't really exist.

Well, what do you think the impact of these technologies will be on optimal firm size, right? Because you could hear stories about how you can have a one-person unicorn, right? Because you can do so much. But then you also would think that being able to coordinate much easier would lead to bigger firms. Do we expect firms to get bigger or smaller? And then I guess a corollary to that is I think a lot of people thought that the internet would undermine the power of the state, but it seems like it's actually enhancing the power of at least the Chinese state and their ability to surveil the public?

So I think in principle, AI means that firms that set up with AI from the onset can be much leaner. But I think it's also true that you can make the same case with the personal computer, the internet, with the cloud, et cetera, right? And so really puzzling pattern is that We have had all of these technologies that should have reduced the cost of setting up a firm dramatically. And at the same time, we have seen a decline in business dynamism since the '90s and even in the technology sector since the 2000s, although there seems to be a sort of a reversal since the pandemic for a variety of reasons. And so what's pushing up against that? I don't think we really know. I think rules and regulations is probably part of that. And so compliance costs are pushing in the other direction. And so if complying costs increase significantly around AI as well, that will push in the other direction too. And so I think it depends really on our institutions. I think it's important that we get some rules and regulations around AI. But if it looks like what we have in the pharmaceutical industry, where essentially you have to partner with a large pharmaceutical company in order to get your drug through clinical trials, then we have a system which is conducive to large enterprises. So these institutions are important. I think when it comes to foundation models in particular, obviously there's a big discussion around whether— how big the barriers to entry are and whether AI will essentially, at the end of the day, turn out to be a commodity. It seems that the fixed costs are very high, the variable costs are very low, and the moat is not very significant. And so That leads me to think that you should not expect the same industry structure as we have in search, for example, or in social media. But it's possible that distribution will matter a lot. And so firms with distribution channels like Google, for example, might be the beneficiaries of this in that particular industry. But then in other industries using AI, you might see the opposite pattern. And so I think that's a long way of saying I don't know, but I think a big part of it is, you know, the rules and regulations that we put around this, because I think technology for the past 3 decades or so have actually, for the most part, favored smaller companies, favored entry. Perhaps not in social media or around platforms, but in other industries using those technologies. It's clear that we can have technology favoring small, but we still don't end up with entry.

Well, so you end the book with a prediction and a diagnosis, and you say that both China and the US are potentially headed for stagnation. And I think the argument for China's stagnation is somewhat more obvious, right? This notion that the party is cracking down, right? That Xi Jinping has said the job of the judges is to protect the party, right? And so this doesn't seem like an environment that's conducive to a lot of innovation. But it does seem, I mean, the US with all SpaceX IPO and Anthropic and OpenAI, it seems like we're actually experiencing quite a bit of innovation. So how do you justify that forecast, at least with respect to the US?

Well, first, I think even if you look at space innovation, the sort of real upsurge comes in the 1990s with privatization. And so since then, if you look at patenting, we haven't seen anything similar. And so I don't want to diminish the compliments of SpaceX, but I think it's a continuation of what happened in the '90s, not the new pattern. And clearly there's innovation happening in the United States, but it's not yet showing up in the productivity statistics. And so you can have exponential technologies without exponential growth. We've been having that in computing for some time now. I'd be very surprised if AI doesn't have any impact on productivity at all. I think there will be a productivity upsurge. I think the question is how long. or how sustained will it be and how significant? Because it matters if it's just a couple of years or it's like a decade for the computer revolution. That's nothing like what we saw from 1920 to 1970 with the sustained productivity growth of the Second Industrial Revolution. And so my best bet will be that we will see something a bit similar like the ICT revolution from '95, to 2004, but probably a little bit more muted because, you know, the ICT revolution automated more downtime, at least in scientific discovery. And so you get instant access to material rather than sitting around and waiting for a book to arrive for weeks or, you know, having to go down and search the library. You would get access. The AI, on the other hand, automates more of the production of the essay that you will write, but you still need to verify it, right?

Yeah.

And so to verify the output, you still need to be in the loop. You still need to understand the output, which means that you have to learn with the AI. And so the productivity gain there is the time saved minus verification, which you didn't have with the computer revolution. And so that would be my best bet. I will see a bit of a productivity upsurge, probably not as sustained. What would change? My prediction there is that if we begin to see AI being used to develop entirely new products that lead to new industries, then we can have something more similar to what we saw with the Second Industrial Revolution. Because I think new tasks and activities are critical, not just for jobs, but for sustained growth as well. And so if all we'd done since 1800 was automation, we would have productive agriculture and cheap textiles, and that would be about it. We wouldn't have antibiotics, vaccines, computers, So most progress comes from doing new and previously inconceivable things. And if we begin seeing more entry in AI applied to entirely new products and industries, then I think it can be much more sustained productivity growth. And so How Progress Ends refers to the pattern that what we've seen globally for the past 2 decades is productivity stagnation. And the point is not that that trend cannot be reversed, but it doesn't look Well, are there any policy recommendations that you would make in order to increase the likelihood that progress doesn't end?

I mean, is there a concern about consolidation of political and economic power that could undermine innovation in the US?

So one actual encouraging thing is that we are beginning to see IPOs again, right? So companies becoming firms in their own right. And so in the early days of the computer revolution, basically every major company from Apple Apple and Microsoft to Amazon and Google IPO'd and became firms in their own right. And then you see exit by acquisition, you know, YouTube, Instagram, WhatsApp being bought up. And now we're sort of maybe beginning to see a little bit more of IPOs as well, a bit more competition. And I think that could be a good thing for innovation. So reducing compliance costs of being a public company is probably a good thing. And more vigorously pursuing anti-competitive practices like killer acquisitions, whereby firms just buy up another company in order to shut it down. We seem to have a revolving door between patent offices and some of these incumbents whereby they grant low-quality patents and then begin to work for some of these incumbents. So clamping down on those practices I think will be really important as well. And then I think in the space of the ideas-producing sectors, giving people more autonomy to produce new and novel ideas, giving them better incentives to produce new and novel ideas. When you get a new powerful productivity tool, basically you can do one of two things. You can use it to drill more holes, or you can use it to dig deeper. And the incentive in academia, broadly speaking, is publish or perish, which means that people tend to try to pursue more projects. We see that since the computer revolution, people pursue more projects at any given time. What we also see is that those individuals are then less likely to actually produce a breakthrough because their attention is more thinly spread across multiple projects. And so reshaping our incentives to encourage more breakthrough innovation, creating new products and industries, I think will be really important.

Well, Carl, thanks so much for joining me. This new book, How Progress Ends, it's really fantastic synthesis of history, And commentary on current events. Thanks again.

Thank you, Greg.

It was a pleasure. Thank you for tuning in to the Unsiloed Podcast produced by University FM. If you enjoyed today's episode, please give us a 5-star rating and review in your favorite listening app.

To listen to our other episodes, please visit our website at www.unsiloedpodcast.com.