#468 - How Progress Ends: A Dialogue with Carl Benedikt Frey

Hello, welcome to Converging Dialogues. This is Xavier Bonilla. On this episode, I'm, I'm really honored and privileged to, to bring you the conversation I had with Carl Benedikt Frey. Carl is the Dieter Schwarz Associate Professor of AI and Work at the Oxford Internet Institute and a fellow of Mansfield College, University of Oxford. He also directs the Future of Work program at the Oxford Martin School, where he also serves as Oxford Martin City Fellow. He has studied economics, history, and management. He has his PhD from Max Planck Institute for Innovation and Competition in 2011, and he founded the Oxford program on the future of work. And he's done much, much research on economics and the history of economics. And he is the author of the latest book, How Progress Ends: Technology, Innovation, and the Fate of Nations. Probably one of my favorite books I read last year. We had this conversation right at the tail end of 2025, and I was just super thrilled to talk to him about all of these ideas that he puts forward in the book. So we talk about, why decentralized systems create an environment for innovation. He's, as he says in the conversation, he's not the first person to notice this, but the way he talks about it and how he really elucidates certain points and brings things out about it, I thought was just tremendous. We go back in time in history. We talk about ancient China's centralization efforts. We talk about Europe's competitive state during the Enlightenment. We talk about the US and decentralization and how that was, you know, a huge part of how the United States has made so much innovation for so many things, especially in the 20th century and still at present. We talk about modern regulations. We, of course, talk about artificial intelligence and where it's headed for our future and many other topics. Again, I loved his book and he was wonderful in conversation. I was so happy I could get You get to talk to him about these ideas and really try to understand the essence of, from a systems perspective, what makes things work, what builds innovation, both from a technological standpoint but from other standpoints within our society, and the challenges that we have facing us now and in the immediate future. And so, he's really got his pulse on a lot of really important things happening for our society. Yeah, I just can't recommend his book enough. As always, you can find this conversation and all other conversations at convergentalogues.substack.com. I'm also on YouTube, so get to the Substack, subscribe, share widely with your friends and on social media. And now I bring you Carl Benedikt Frey. I am here with Carl Benedikt Frey. Carl, thanks so much for Coming on the podcast. I'm looking forward to talking with you.

Good to be with you.

Yeah, yeah, absolutely. You've written a fantastic book. I thoroughly enjoyed it. It's called How Progress Ends: Technology, Innovation, and the Fate of Nations. Are they promoting this like an economic book, a history book? It's got a bit of everything in here. How are they promoting your book over there at Princeton?

It's a very good question. You have to ask them.

Yeah, it's crazy, right? Uh, well, it's great. It's a, it's a very, very, very good book. Uh, so before we get into it, why don't you just tell listeners kind of what your day job is, what you do professionally or academically, or all the, all the other relevant details?

I'm an economist by background, but I don't work in an economics department, which means I can actually write books. Um, and so that's part of what I do. I spent the past 5 years writing this book. Before that, I Spent 5 years writing my previous book called The Technology Trap. And aside that, I do what academics do, teaching, admissions, and trying to be a good father as well.

No, that's very, very nice. Yes, well, I'm sure you're a good dad, you're a good writer, a good thinker. So I'll parachute us down into the beginning of the book. The main thesis of the book is how central government management is good for, as you say, exploiting low-hanging technological fruit, but decentralized systems are better for innovation and new technologies. I've sort of heard some version of that or put together in other places, but for your major thesis, how is it different from the way others have made sort of similar observations?

So I don't think there's anything in this book that nobody has ever said before. But as you mentioned, it covers a lot of ground. And so hopefully, you know, almost everybody will learn something from reading it. So what you just said about centralization being helpful in exploiting and taking advantage of low-hanging technological fruit, whereas decentralization is more beneficial for innovation, I think parts of that argument argument said by Gershon Krohn and related to European catch-up in the 19th century, but then confined to the European experience in the 19th century. Barry Eichengreen has made a similar argument around European growth in the post-war period and that coordinated capitalism essentially running into diminishing returns starting in the 1970s, but not sort of making this decentralization-centralization distinction perhaps quite as explicit. I think the contribution of my book is trying to create hopefully a coherent framework to think about why some places can grow rapidly for some time, then stagnate, and then even decline or collapse, as in the case of the Soviet Union. I think Gershkian-Krone or the Eichengreen accounts don't deal too much with incumbent interests standing in the way of reform and that making it very hard to make the transition between centralization and decentralization. But others have covered that aspect in particular in some detail. So it's a book that brings a lot of strands together, I would say.

Yeah, certainly. It was interesting. I had no No, any knowledge of anything going into it. I just started it and read it and I kind of, it all hit me. I didn't read any real blurbs. I didn't read anything. I didn't get the previews before. I didn't, I just kind of dove right into it. So it was really cool to see your argument unfold. And I think the thing that's really wonderful about, you know, what you're, what you're doing is, is that it's very much international, right? It's very, it's very global. Um, some people focus on one country or one period. So you're doing this big sweep of different places, different, uh, Empires or countries. And so you start with 1,000 years ago, you go all the way back to China 1,000 years ago. And then apparently, in case folks didn't know, China was a powerhouse. They're a powerhouse now, and they were 1,000 years ago. There's many dynasties. Tell us about this story, why you chose here as a sort of kind of starting point. And I guess what, what was distinct about there and how there's some things, you know, some of the lessons we learned from it. I mean, you, you mentioned that there was, uh, this space where they were trying to centralize, they were trying to create this modern bureaucratic state, um, but it, it wasn't sustainable long term. How is this like a first example of one of the points you're making?

Part of the reason for starting there is where we are today. And so today we are very much to some degree in a similar position where China pursuing more state-led model, Europe and the United States pursuing a more market-based model, but perhaps looking more to learn from actually what is happening in China rather than the other way around, which is a Perhaps a bit paradoxical. But so I think that we can learn something by actually starting there. And what we learned through the Shire Days experiment— experience is that you can spearhead technological development for some time. But if what you're primarily doing is using technology to centralize decision-making, for example, through soil mapping, being better able to understand what farmers produce to levy taxes, to invest in a uniform writing system and develop population registers to be able to track people around. You can certainly create some growth by doing that, but you're going to reduce the space for exploration and incentives for people to take risks. If Galileo had been born in China, it's actually quite likely that he would have been a civil servant rather than a scientist, because that was the ladder for social mobility in China. So the ability of Chinese rulers to track and trace what people produced meant that they had less reason or less need to rely on people with local knowledge and to delegate decision-making to local levels so they can centralize. And then using those technologies to build elaborate bureaucracies through meritocracy Bureaucratic recruitment drew in a lot of the talent into the government bureaucracy, and also, I think it's fair to say, reduced diversity of thought. In Europe, that was certainly true that there were from time to time many spaces where the diversity of thought was limited. But when Louis XIV evoked the Edict of Nantes and made Protestants Protestantism illegal in France, the Huguenots could migrate to other places in Europe where their skills were highly valued and where they contributed to innovation and to economic growth. The same is true for many intellectuals that often moved around. And in China, some people migrated too. But when they migrated to places like Japan, for example, They had very little impact on discourse within China itself. Hardly anybody knew about them until perhaps much later. And so I think that's quite a distinctive pattern of economic development, which shows how you can take the lead and use technology and innovation to build up bureaucratic capacity and harvest some low-hanging technological fruit, but that eventually runs into diminishing returns and you need innovation or you need decentralization for innovation.

I have 2 follow-ups here, which will probably carry throughout. Is it something about the way that the system works that doesn't allow this to happen for a long time? Is it something about centralized systems in and of themselves, or is it something about, uh, how it's run or how it's, you know, alternatively, could you imagine a centralized system that does create the space or the environment where people can innovate, where they can do things long-term? Um, so I'm just curious, is it a systems thing? Uh, is it a kind of person management thing? Uh, or— Or is it just the time thing? Is it just, yeah, that's gonna work for, you know, 15, 20 years maybe, but it's not meant to be forever. You need to have some kind of disruption to that. I think in a lot of the ways, at least in the modern period, everyone kind of expects that some element of centralization is really helpful. It keeps things in check, it regulates things, et cetera. But you're— I guess the big point here is, is that might be nice for a little bit, But long-term, you're going to need some decentralization. You're going to need some innovation. What do you kind of make of these kinds of points here?

So I think it's a systems thing. So if you take the Soviet Union, which was the most centralized economy the world had ever seen, if you were an aircraft engineer in the Soviet Union, you developed a new engine, which you needed funding for, you could go to the Red Army. If they declined, you probably had 2 or 3 other options. If they declined, your idea would die with you. That's quite different from 20th or 21st century America, where Bessemer Venture famously declined to invest in Google back in 1999. They probably regretted it.

Now they do. Now they do. Now they're like, oh, what were we doing? Damn it.

With the benefit of hindsight, we're all smarter. But it also illustrates that Google was not a safe bet at the time. AltaVista and Yahoo were dominating search. And so to know if something will catch on or not, somebody needs to take the risk of investing in it and trying it. And the beauty with a more decentralized system is that you can have a range of people taking very different bets and explore different technological trajectories. And so I think that's a feature of the system. Now, that being said, you know, some organizations Permit more exploration than others, and so what you see during the period of the computer revolution, for example, is that organizations that were more decentralized did better because they permitted people at the lower levels of the organization who were actually closer to this technology, who were using it, who were experienced. Experimenting, they did better in productivity terms during the computer revolution. And so when technology is in flux and you need to experiment, even within organisations, you can delegate more to lower levels of the organisation. You can permit more autonomy. But I think in the end of the day, you do have reporting structures in place. You will have more veto points in large firms. You will have more consensus building. And so, and you will find yourself more invested in one particular technological pathway which you're associated with. And so I think part of the reason that Google, despite developing the transformer, was slow to act upon it and launch generative AI the way that OpenAI did, was that they were concerned about that having a negative effect on revenues which they derive from search and advertising. And, you know, Google has since then closed the gap and by some metrics now even overtaken OpenAI again with their latest release. And so you can see that you can pivot and you can, you know, provide space for experimentation, but it still depended on OpenAI taking the initial risk. And once the genie was out of the bottle, Google then rapidly followed. And so we see throughout history that the leaders in bicycles, none of them became leaders in automotive. The leaders in EVs are not the legacy automotive companies. None of the traditional media companies led the way in social media. And I can go on. And so I think there is something systemic there.

It is very interesting. Again, it's one of those things you don't, at least for me, I don't see the kind of bigger picture there. And then when you kind of lay it out that way, it's like, oh yeah, it is. It doesn't have to be always a continuous thing. You can have other people that pivot or they're able to have new innovators, which kind of goes to your point. In the book, you talk about Europe's competitive state system thriving under no central government during the Enlightenment. Before you get there, you mentioned Scheidel, who's been on the podcast. He's great. We talked about history, kind of, again, more 30,000-foot view. I know he's done a lot of stuff with Roman history and stuff. But, you know, he had mentioned— you state that he mentioned the collapse of Rome was crucial for this competitive state that emerged out of Europe. And a lot of the times people have made some parallels with how, you know, the US has done things with Rome, for better or for worse. But how did Rome and then Europe become— by becoming more decentralized, help propel it to where we know its kind of modern period? And again, of course, this is over hundreds of years, but— How do you see that picture kind of laid out?

Yes, I cite Schuydel extensively in the book and numerous other scholars. He's great.

He's very nice.

He's fantastic. And there are many people that I do cite that I'm not going to be able to credit here, so please read the book. And so going back to the points we discussed earlier about intellectuals and talents moving around, that was a key part of why European fragmentation was really important to industrialisation. And because of that competition that then emerged between European states and cities, many began to actually invest in knowledge societies and provide various incentives for inventors and talent to come to their place. And so it made governance by necessity becoming more pro-technology. And Britain is perhaps the clearest example of this, because wages in Britain were very high, and so the ruling classes realized that in order to be able to compete with places where labor is cheaper in global trade, they needed technology. They needed mechanized factory in order to cut costs. And that meant that they had to get obstacles to technological change out of the way. And the way they did that was in large part by clamping down on the craft guilds, which had regulated production in Europe since the 12th century and nurtured innovation to some degree because they transformed skills from one generation to another. But they also opposed any technology that threatened their own skills and incomes, and the Industrial Revolution the technologies that made the Industrial Revolution did exactly that. And then, because distances in Europe were not as long as in China, for example, it also meant that local guilds soon found increased competition coming from other places, especially as the road network developed with the turnpikes in Britain that undercut the political clout of the guilds. And so competition brought by this necessity of innovating, this necessity of technological progress for national survival, and In numerous cases, that led to productive reform. That is not to say that war is good for progress, as is sometimes sort of crudely summarized today. It's certainly not. But in some cases, competition can lead to productive reform, and sometimes even defeat can lead to. productive reform. And Prussian losses against Napoleon in Jena in 1806, which paved the way for the Stein-Hardenberg reforms that abolished the guilds and gave greater occupational freedom, secured private property rights, et cetera, is one such example.

Yeah, it's very interesting. You mentioned that Britain's civil service structures, as you're mentioning their industrialization and education reforms, It had these ripple effects. Germany was doing this as well. But I guess tell us how you set up these examples in France, Russia, and Japan about how, again, they used centralized institutions. So, you're talking about many of the positive or productive elements of the decentralized networks. But in these other states, France, Russia, and Japan, How did they again? They they there was some positive things about it: robust militaries, industrial policies, some social nets. But it doesn't seem like it is able to last or be sustainable. You know, you tell us how this was good in in the moments, but then it turns out that it starts to kind of fracture at a certain point.

So once Britain industrializes. European powers on the continent can essentially grow by transferring British technology across the English Channel, which they begin to do in numerous ways. And so progress on the continent is more plannable. You don't necessarily need, to the same degree at least, institutions that drive innovation, You need institutions to absorb, to implement, and to scale technology invented in Britain. France is a good example. It was closest to the technological frontier. It was closest to Britain, and therefore the scope for catch-up was not as extensive as in Russia. or Japan. And French elites did try to spur innovation by committee, and that didn't work out very well.

No.

It rather led to people awarding family and friends prizes in competitions That you know were not given on the basis of merit, but they managed to build you know a very elaborate railroad network, and they used the state to much larger degree to catch up in railroad than Britain had in Russia. On the other hand, state capacity was much reduced by the fact that the tsar involved himself in every aspect of government. And that led to a situation where the tsars could essentially only survive by society and the private sector being weak. Divided, and so in Russia you see concession system for the incorporation of new firms. There's no registration system, as you saw in the rest of Europe, and so the Tsars essentially had to sign off on the formation of new. Types of companies, which meant that you had few firms in Russia, and which meant that you had very little and limited competition. But Russia, through Sergei Witte, for example, did manage to also build an elaborate railroad network. And they, quite similar to the Soviet Union, later on prioritized heavy industry. But they pursued industrialisation without institutional modernisation, which soon ran into all sorts of problems. In Japan, they had the legacy of the Chinese meritocratic state. And when American gunboat diplomacy led to the Japanese realization that growth and innovation is imperative for survival, Japan had the capacities to set up the Bureau of Translation to focus on actually translating Western texts to Japanese because Japan didn't even have a word for railroad, for example, at the time. And so this was a very concerted effort to bring technological know-how to Japan. And in addition to that, the Meiji government sent various delegations to Europe and the United States to Scout the market for technologies and develop pilot projects that are then later on sold to emerging cybatsu conglomerates who then leverage scope and scale. And so what the example of France and Russia and Japan show. Is A, the closer you are to the technological frontier, the less you have to gain from centralization. And B, the capacity to catch up is not automatic by any means. You have to develop institutions that are capable of transferring and scaling. technology invented elsewhere.

And those are very important lessons, I think, to learn. And again, it's one of those things that in real time, probably not seeing it, but when you can kind of step back and look at it, you start to just see, oh wow, like there's all of these different patterns that are emerging. One of the things that's really fascinating here is, you know, we get across The pond. And so we see all these things that are happening in Asia for many years. We see all these things happening in Europe, and then we get to the United States. Um, yeah, tell us, tell us a story of, from this perspective of the United States. Um, you know, you mentioned that one of the main, you know, great features about the US is that it was built on the growth of many entrepreneurs. And you mentioned this happened with a weak decentralized government, feeble bureaucracy, and further that the US Constitution was a decentralized system of secure private property rights. I mean, do you— tying those 2 together, do you think that because, you know, the US Constitution has that kind of embedded in it that the US, I don't wanna say always, but is sort of built around this sort of, how do we have growth and innovation with a kind of balance between this kind of decentralized government, um, and, and, and to, to, to promote, uh, ways of creating things? What, what do you, what's the story here, I guess?

So American independence was a revolution against centralized government. And so centralization had a bad rep in the United States during this period of time, and still has today in many ways.

That's interesting.

And so what the United States had going for it, but was by no means inevitable, was a large internal market, as many have pointed out. But that was only really enabled by a constitution that ensured tariff-free trade inside the United States. So the large homogeneous market that the US benefited for, for a long time was an institutional invention, if you like, going back to the Constitution. And what it had in addition to that was a system underwriting secure system of private property rights and a patent system with very low barriers to entry. And so the filing fee in the US was much, much lower than in Europe. The American patent system was second to none. And that meant that innovation in the United States was fairly democratized by the standards of the time, I should say, in the United States. What it did not have was the capacity to do large-scale industrial and infrastructure projects, which became more important, not least because distances within the US were and are still enormous by European standards. And so the railroads in America mattered much more than they did in Europe, but the American state had much less of an active role to play in that process relative to the continent in Europe because it didn't have the resources and the state capacity. So what it did to alleviate the deficit was building public-private partnerships, granting land to railroad companies that then laid out tracks, And then with the Railroad Act of 1861, it began to harmonize the system that had been built by then in a very fragmented way. But that meant that you had huge industrial firms emerging around the construction of the railroads. And you have, with the growing and more integrated market in the United States, other firms emerging to exploit the economies of scope and scale. And these companies became larger than the federal government itself.

Right.

And perhaps not surprisingly, they tried to bend the rules to their own benefit. They did what they could to stifle competition, and that threatened entry and innovation. And so the response to this came largely from the grassroots, where people demanded reform. And that led to the Pendleton Act, which gave rise to meritocratic civil service. And that civil service then used those powers to begin to regulate firms and enforce competition through the Sherman Antitrust Act to begin with. And so many of the building blocks of American technological dynamism in terms of this decentralized system of private property, a large internal market, a government capable of enforcing the rules of the game had emerged by the late 19th century. 19th century. And a big part of the attraction of talent and inventors moving to the United States was also that.

I guess 2 things here. The first thing is that you have to have people that want to innovate. and to create and for the greater good or for many people en masse, things like that. Do you see, aside from select companies or few companies at the moment, do you feel, not because of centralization, we've really the past 10 years have had less of that. But do you feel that, at least in the US, and maybe even in Western Europe as well, but do you feel in the US that we're not innovating as much as we used to, that there's less innovation? Or would you say that there's more innovation? We can talk about SpaceX, OpenAI, Google, You know, Apple to a certain extent. You know, okay, we got the big 5 or 6 companies that innovate and create, but as you kind of look at it all the way throughout, do you see less innovation or do you see more? How do you see it? And if it's less, what could be the causes for that?

So the real puzzle is that since the 1980s roughly, we've seen an enormous acceleration in patenting. And we've seen disappointing productivity growth. And so, you know, we see a little blip with ICT revolution, 1995 to 2004, which then sort of peters out. And so why do we see this upsurge in innovation and this decline in productivity? And one explanation is that each of those patents that we're filing since 1980 are less important than previous patents. So we're sort of doing quantity rather than quality, if you like. And we see this if you look at, you know, the share of truly sort of breakthroughs, disruptive patents which are highly correlated, or publications which are highly correlated with Nobel Prizes, for example. That's been going down. Research productivity has been going down. And so it suggests that something is happening there. And so in academia, for example, it's publish or perish. And so that system pushes you towards quantity rather than quality. If—

Another reason I'm not in academia full-time.

Right. It's a good reason. And so if I don't produce anything for the next 10 years, I work on one thing, I don't know how long it would take before my colleagues start to wonder what on earth is Carl doing? Well, he's at the office, but hasn't produced anything. Was he on social media? What is he doing with his time?

Right.

And so it's hard to do that. And we're not incentivized to spend 10, 20 years doing one thing. And so that might be part of it. I think it also has to do with declining entry. Declining competition, and so large firms they have scale. They tend to focus on what they're already doing and improving production processes around that, lowering cost. That's good for consumers. That's important as well. But they are less likely to develop new products, new kind of industries that drive growth over the long run. Right? And so think about it this way: If all we had done since 1800 was automation, we would have produced. Productive agriculture, we would have cheap textiles, that would be about it. We wouldn't have vaccines, antibiotics, computers, rockets, airplanes, et cetera. And so the fact that we are having less entry, fewer new firms, less dynamism, it doesn't just mean less innovation. It means that we're having a different kind of innovation. We're having more innovation more concentrated in larger firms, focusing more on production processes rather than driven by smaller firms, which develop new products and new lines of industry. And I think that's also part of it. And the question then becomes why we're having less entry. And so part of it is, I think, is Thus, we have more rules and regulations, which lead to compliance costs. And many of those rules and regulations are perfectly understandable and reasonable, don't get me wrong. But they nonetheless add compliance costs. And large firms find it easier to manage compliance costs. And so, you see this very clearly with the GDPR in Europe, for example, where smaller firms have suffered, whereas larger firms have essentially offset compliance costs by capturing a larger share of the market. And then in addition to that, you have incumbents pursuing anti-competitive strategies such as buying up promising startups just to shut them down, killer acquisitions. You see a revolving door between the US Patent and Trademark Office and incumbent firms whereby examiners are more likely to grant patents or low-quality patents to these firms and then take well-paying jobs at those particular companies. And so I think those aspects together probably explain quite a bit of what's going on.

Just going back to the main argument that you're making, and again, it's not this kind of— don't make this false dichotomy of black and white, but since 1980, The, you know, the how there's been lower types of production of things. Like, when I think about it in the US, we had 8 years of Reagan, pretty decentralized, 4 years of H.W. Bush, pretty decentralized. Even the 8 years of Clinton, what was his whole, you know, we're gonna shut down the role of government, the third way, blah, blah, blah. Okay, plus or minus 8 years there, decentralized. 8 years of Bush, pretty decentralized. I mean, you see my point here. We had a long time of doing this. I mean, really, I would say 8 years of Obama was probably the most regulated or centralized or attempts to, but even then, I mean, I don't think it was as draconian as other people on the right tried to make it be. It was pretty pro-business. And the same with Biden. I think it was, there was a lot of spending, but I don't think there was as many regulations of sorts. So I mean, when I look at the past 46 years, it doesn't seem like it's been, at least in the US, we've been stifling growth or innovation necessarily from the kind of federal or centralized component of it. What I do see is kind of what you were mentioning there at the end You do have a lot of big entities or big companies just kind of just swallowing up all of these other startups, all of these big mergers, all these things. I mean, wouldn't that be the case of sorts to say, I guess the line here is for this is, and we can pivot this too with the stuff we're talking about with AI nowadays, but. Where's that limit, right? Maybe it doesn't need to be central— centralized government, but where, where does that limit go where we're like, okay, you can, you can play in this space, but once you hit this ceiling, now you're breaking rules. Now you're doing stuff like in the Gilded Age or, you know, the robber barons, all these things. And again, this isn't a— this isn't an off-ramp to Bernie Sanders types of kinds of thinking or something like that. I'm not, I'm not trying to do that, but Yeah. I mean, there is some challenges here. If yes, we have innovation, we have really great things that a lot of these big companies are doing that we all use, right? Everyone's bought an Amazon Prime package. Most people have a Gmail account. Most people have some type of Apple product. Okay, that's fine. But at what point does it say maybe this is too much and it's not allowing other people to kind of What do you make of this?

So first of all, I would just note that I think there's, in the space of regulation, I think there's been more continuity between the precedents you're describing than may meet the eye. So I think there has been various policy differences for sure. There has been huge differences in rhetoric.

Certainly, yeah.

But I think even going back to Nixon and the Environmental Protection Agency and so on, which begins with him, there has been a steady rise in regulation. Many of them around environmental protection and occupational safety. Many of these rules and regulations are things that are by no means necessarily bad. So I'm not suggesting that we should scrap all of them. added to compliance costs for firms. And those compliance costs, I think, do favor larger companies. And I do think we need to be mindful of that. And so that's something that goes sort of all the way from Nixon up until present day. On the question of— sorry, I lost the thread on your last video.

The big kind of companies kind of swallowing up other ones, these mergers.

Right, right. Where do we draw the line? So, um, you know, I think first of all, we want to, uh, do what we can to deter anti-competitive behavior.

Mm-hmm.

And that deterrence might even mean some missteps. So not all, you know, antitrust intervention is productive. You could easily see that, you know, breaking up Meta just leads to, you know, a new social media giant emerging in the end because of network effects. But you could also see that it would send a signal, and I think it would make firms more cautious in their approach. And so I think you want to monitor killer acquisitions, for example, is something that you want to monitor very closely. In merger reviews, you may want to look more to valuation than just turnover. And broadly speaking, I think what you want to do is you want to tip the scales. So, if you look at the early days of the computer revolution, from Microsoft and Apple to Google and Amazon, they all IPO'd and became firms in their own right. If you look from the mid-2000s onwards, WhatsApp, Instagram, YouTube, etc., All exited via acquisition. And so what we want to do is that we want to push the scales back so that we incentivize firms to go public and become firms in their own right rather than growing through consolidation. And so deterrence Trying to tip the scales with regard to consolidation, I think is the way to go. You know, I'm not you know a great fan of the consumer welfare standard, but at least this is a standard. And so we do need to have a standard that means that we're you know enforcing these rules to you know particular criteria. And so we don't want to move into a situation where antitrust is becoming overly. Politicized and lawsuits are being launched against companies just because the president or somebody else in the administration doesn't happen to like them. And so I think that that's something that we do want to keep in place as much as you can. Otherwise, you're moving towards China where antitrust is very much a political tool.

Yeah. I guess the last topic we can land on here is this sort of advent and rapid explosion of AI, LLMs. What do you— how do you feel the major world economies can find that balance with this in particular? This is— I find the conversation mostly around these things to be pretty Parochial. I think with a wider lens such as the one that you have is important. I mean, the things from history, the things from recent and past history, the things from kind of the argument you've been making, we obviously need big innovation with AI. We obviously need people to be there, this competitiveness. We want to figure this out in a way where we can use it well. But I'm not completely doomsday on this, but there are, when you look at the potential serious concerns, and my worry is if we have something like a generative AI, if we have something where we're not having this sort of, we haven't figured out the alignment problem, right? Of how much does AI align with human interest or not? If it starts to diverge from human interest, Yeah. just, are we in big trouble? Again, I'm not saying that we're going to get to a Westworld situation tomorrow, but I don't want to get there ever.

Yeah.

How do you understand the role of major entities or global economies to say, yeah, how do we push and spur on this innovation and keep exploring and see where this is? This is where The development and innovation seems to be at the moment in a variety of ways, but maybe the speed of it or how rapid it is, is a little worrisome because we're not— it doesn't— it feels like we're— feels like we're riding in the car with no brakes is kind of what it feels like. That's how it feels for me sometimes. It's like, whoa, like this is— we're just, you know, there's 3 people or 3 companies just spewing out all this stuff and we're kind of like trying to catch up and get like on board with it. And, you know, I think it's like anything else. You start giving things to mass amounts of people, you're gonna find some pretty nefarious ways in which to use it or things like that. How do we navigate this kind of landscape at the moment as it's very, very, I guess you could say, very ripe? It's very dynamic at the moment.

Yeah, so I'm not sure if there's one reason or one answer, I should say, to your question, but, you know, beginning with the, alignment problem. So I'm not sure that that problem can be solved for the simple reason that humans have quite different values if you look around the world values survey. And so hopefully we can encode some values that most humans don't want AI to kill all of us.

Yeah.

So hopefully we can encode that somehow. I haven't been overly concerned in the past about these existential risks, but what you're beginning to see is that in warfare, for example, It's too slow to keep humans in the loop. And you are beginning to delegate more decision-making to drones, for example, powered by AI. And once you begin to delegate more and more to algorithms, well, you are losing more and more control. So it is a concerning trajectory in some of these domains. I would say, though, that with large language models, with the AI that we're having now, I don't think I don't think that the risks are mostly that fundamentally different from what we saw with the internet, right? So yes, we have concerns that we don't want LLMs to be trained on dangerous information that can enable somebody to build a biological weapon. And we have the concern on the internet that we don't want people to post the blueprints of biological weapons on the internet. And so I think many of the concerns today are quite similar, but they could, as technological capabilities improve, become much more significant.

I guess the thing there is that I guess what I'm seeing, which is different— so you and I are both old enough that we remember before the internet existed, or at least for the popular public masses and home computer and So you and I have seen that kind of— I remember hearing these worries too, and oh, all these things, and it didn't turn out that way. I'm not even that worried about social media that much. I know everyone makes a big deal about it now and all these things, and phones are just destroying kids' brains and all that. I'm not fully on that train either, although there are some challenges. I guess the thing with the LLMs that I worry about is, is that we're, we're, we're not, it's sort of like the drone issue that you mentioned is, is that we're offloading human cognition and human intelligence to a machine where we're not thinking, literally thinking anymore. Because if you can get the answer, you know, I put in ChatGPT all these things, we're not using that skill anymore.

Right.

There's all of these cognitive processes that are happening with artificial intelligence of sorts. I mean, as an example, I mean, I'll have students that, you know, I have to explicitly state, you cannot use this to write your paper, or I have to find new assignments because they just put it in there and it spits it out. And then they maybe, you know, turn some things around and that's it. Like, there's a big problem in terms of thinking, in terms of, and for other things too. I know people that use it all the time. time. Again, it's a great tool and it can be super helpful of like, I need to find this wording. I need to do all these things. But I think sometimes, again, people are going to over-rely on this where I worry downstream of, are we fostering, fostering a generation of people where you have good cognitive abilities? Right. Are we going to be able to look at cognitive scores of people in the next 20 years that have been using ChatGPT and does their average go down one standard deviation or something? I don't know, right? But that's, I guess, my worry when I see it in kind of more everyday kinds of use or things like that. But I don't know, what do you think about that concern or that kind of— am I too alarmist on this?

No, I worry about that too. And I mean, I've even seen students in class during discussions prompting an LLM and then reading out the answer in real time. And so this is certainly something that is happening, right? But then I think the solution to that is this, right? Pen and paper. We need to go back to examining students the way that I was examined 20 years ago.

That's right.

And we can go back to verbal examinations and so on. So I think that's the question that we're something that's actually very solvable through changing institutions and incentives. We're not after students maximizing exam score. At the end of the day, we're after learning. So it's a different objective.

Right.

And so I worry about that too, but I think that's within our power to fix. I think the question you raised earlier is a harder one. So which sort of— which markets structure do you want? Do you want open source AI, which everybody could tweak? That's probably going to be very beneficial for innovation, but it's obviously going to be extremely hard to control. Or do you want one firm or two with a monopoly on AI? And so I don't know what I prefer in terms of National security and safety, something in between most likely. But exactly where to draw the line, I don't know. What I feel relatively comfortable in saying is I think it's more likely that the more open system will drive innovation and progress. And so, and I think an underappreciated risk is a lack of productivity growth. It sounds very unfashionable to say, but But it is a real risk because an economy that doesn't grow eventually becomes a zero-sum game. The only way I can improve my standard of living is at the expense of yours. And so when you see me doing well, it means that you must be doing badly. And that creates very, very bad politics. And so many of the risks that we worry about AI, they are real. But many of them are also not huge. And I think the growth problem that we do have is actually bigger than many of those. I mean, how else are we going to pay for things like Social Security or educational system, healthcare, et cetera? So we need growth. And so we need to make sure that we also don't stifle progress in a detrimental way in the pursuit of safety.

I guess the sort of final question is picking up on that is where do you see us going? Where do you see— we have these big juggernauts that have some challenges. You talk about in the book, China, US at the moment, they have a lot of innovation or they've I don't know how you'd want to measure it about growth. I mean, the US feels a bit stymied, a bit. But yeah, I mean, biggest economies and biggest militaries, all these things, but it doesn't feel— I don't want to say sustainable, but it feels like there could be some issues. Where do you see innovation going?

So I'm very concerned. And so I live in Britain. People here are very concerned about British growth over the past 2 decades, getting almost impossible to do anything meaningful full reform-wise because the fiscal space is not there. It's challenging in Europe on the continent, but actually US and China are not doing particularly well either. And so since 2004, US productivity growth has been very disappointing. Since the Great Recession, productivity growth in China has been stagnant by many estimates and very slow by more optimistic estimates. And so far, we're not seeing AI in the productivity statistics either. And it seems to me that both in the US and China, policy is driving in a direction that makes it less likely rather than more likely that we will be able to harness the gains of AI. American politics is becoming much more transactional. And if you look at the tariffs, it's not just that they are reducing competition from abroad, which is important to check incumbents. But it's also, in addition to that, creating a very elaborate system of exemptions, which means that in the end of the day, you're going to get CEOs and leaders of corporates which are good at navigating politics rather than good at driving tech. And in China, you see that political objectives have taken precedent over objectives around growth. And that means that they're relying more on state-owned enterprises than they did. 10 years ago. And by any measure, foreign-invested and private firms have been more progressive in both productivity and patenting terms. And so I think from where I'm sitting, I don't see many things moving in a direction that gives me hope. The main hope would perhaps be that Europe now actually gets its act together, having begun to internalize the fact that it's on its own. And let's hope that that doesn't lead to another wave of techno-nationalism in Europe either, but rather reforms that foster completing the single market and foster fostering dynamism in Europe. And so, technonationalism can maybe provide a little bit of a cushion for some firms and make decline less rapid, but it's not a growth strategy. And so, Europe needs to find a growth strategy if it wants to grow.

Yeah, no, I fully agree. I share the same concerns. Obviously, you're closer to this, so I think those are important challenges we have to really, really consider. Uh, your book does that. It's, uh, it's very, very good. Uh, everyone should go pick it up. Uh, Carl, thanks so much for coming on the, the podcast. It was a, it was a fantastic conversation. So, so big thanks for your time, energy, and, uh, wisdom. It's, uh, it's much appreciated.

Thank you. It's been a real pleasure to be with you.