HEC Talks: Do Ideas Become Harder to Find with Philippe Aghion at HEC Paris Description of the event: Event by HEC Paris Mon, Sep 14, 2026, 8:00 PM (your local time) “Do ideas become harder to find?” with Nobel Laureate Philippe Aghion On September 14 at 8:00 PM (CEST), join us live for an HEC Talk with Philippe Aghion, one of the world's leading economists and a renowned expert on innovation, economic growth, and the dynamics of technological change and 2025 Noble Laureate of Economics At a time when AI promises to disrupt almost every sector, yet offers no clear map of the future, Philippe Aghion is a voice worth listening to. His work connects economic and technological change with the social conditions that make innovation sustainable. By showing how economies grow through continual renewal—and how creative destruction creates both opportunity and disruption—he has made a compelling case for innovation policies that combine competition, investment in knowledge, and meaningful support for those left vulnerable by change. Transcripts: (05:44) Please take a seat. The conference is going to start. There are still a few places in front, if you want, and we are going to set up some other chairs in the old owner. Okay. Hello, everybody. Welcome here. (08:10) My name is Antonin Bergeau. I'm a professor here, and I've been given the difficult task of introducing Philippe Aguillon in less than five minutes. And if you go to Wikipedia and look for Philippe, you will see why it's a very difficult task, because in five minutes you basically do not cover the introduction, and probably just the list of his achievements would take the whole day to spell out. (08:35) So instead of telling you about the biography of Philippe, let me just ask one question, which is why some countries are so much richer than others. Why is Norway 20 times richer than Honduras, for example? And that's a very complicated question that another economist named Robert Solow started to address in the 50s. (08:59) And what Robert Solow said and developed in a model that if you took a macroeconomist class, you all know, is that you can explain why countries get rich if you understand that... At some point in our development, people decided to save some part of the output so that firms can use this part of the output to build up some capital. (09:21) And because capital accumulates, and because capital uses production with people, then in this model, for the first time, you can explain why some countries get to some level of richness while others remain poor. And that was a very influential model, but it's a bit frustrating in the sense that you can't explain why, at a given level, of GDP per capita, a country can become richer or poorer. (09:46) How can poor country get richer and catch up with the richest country? And so 30 years later, Philippe Aguillon and his co-author Peter Howitz, that were at the time, I think, at the MIT Massachusetts Institute of Technology. decided to improve upon this model by explaining why would firms, why would people decide suddenly to invest into what we call technology, to use their human capital, to use their resources to improve the quality of the goods and the services that are sold in the economy. (10:22) They had a very clever idea and a very clever model in which they used the idea of Joseph Schumpeter who was another economist that in 1942 developed the idea of creative destruction. Philippe Aguillon and his co-authors did in their model is that they basically explained that in the economy you have firms and those firms sell goods or they sell services and they have profit from selling those goods and those services. (10:47) But you also have a lot of potential entrepreneurs, young entrants, that want only one thing is to replace those firms to enjoy the profit that they are enjoying. And so what those entrants are going to do is that they will invest some resources to improve the goods that already exist in the economy. (11:06) And if they manage to improve this good or to create new services that is better, they will replace and destroy the incumbent firm that is already in place. And they will enjoy the profit that they were looking for. But what they don't realize is that by doing so, they increase what we call the frontier. (11:24) So they increase the best level of quality available for a given service or for a given product. And because they did that, it will have positive effect to the world economy. And so for the first time, you have a model that explains both why you can have economic growth, where does it come from, But also, why would some firm enter a market, exit a market? So you get a much richer model that can finally be used to explore the effect of some specific policies, to explore the effect of some specific shocks. (11:57) And so suddenly you can develop a true innovation policy. And that's why this model has been so influential over the past 35 years. And that's why, you know, this is the type of model that wins you a Nobel Prize that was awarded last year to Philippe and his co-author and also a third economist. Perhaps more importantly, this is the type of contribution that earns you a doctorate honoris causa at HEC. (12:24) So, of course, you know, the end of the story is not quite here. There's still a lot of puzzles to address. One of it would be, of course, can this model help us to understand the effect of artificial intelligence and what we should do to benefit the most from it? What would be the effect of demographical change? And also, where do ideas come from? And do we face a risk of running out of ideas? And what would be the consequence? of such a threat. (12:56) And so without further ado and without talking too much, Philippe, please, if you can come and have the floor. Thank you very much. I think I don't need the... So I should say a few words. First, it's a great honor for me to be... introduced by Antonin. I had the honor to be one of his advisors. And I think when you get a... (13:44) Once you get the prize, you don't want to get another prize. What you want is that your students get the prize. And if I had to bet on one student of mine who is in the race, I would. undoubtedly better than Antonin. That's what I wanted to... So, anyway, I will do everything. (14:12) You know, in Toulouse, Jean-Jacques Laffont, you know... to attract Jean Tirole to Toulouse. At the time, it was the Gray Mac that became the Institute d'Economie Industrielle, IDI, and then became subsequently Toulouse School of Economics. And Jean-Jacques Laffont, it was exactly what he had in mind. He said, you know, I want to attract Jean Tirole because I would like Jean to get the prize. (14:40) And that was Laffont's idea. And I could say exactly the same with Antonin. I will do everything I can to create the best possible climate. So that's what I wanted to say. The second thing I wanted to say is that HEC... I was, I visited, I taught at HEC in 89, in 89-90 during one year. I came back from the U.S. (15:15) and I took a job at Delta. Delta was the precursor of Paris School of Economics. It was Boulevard Jourdan. But I also took a visiting position at HEC. And I was teaching. a course on growth which was joint HEC Delta and I was teaching first or second year HEC students in microeconomics during that year. (15:42) So I had a one-year experience as a professor at HEC, and then subsequently I've been very closely related to HEC. I have a friend of mine who sometimes he asks me to be a guest speaker, and I've been several times, and now we work very closely because now HEC is a great show. is that it's a wonderful research team, and we work closely together, and we are, in fact, working a lot. (16:16) But anyway, I won't say the setup and the thing, but we are very much, well, I feel extremely closely linked, which I say in many respects. So let me start presenting. So, in fact, Antonin already introduced the topic so well that it will look extremely redundant at the beginning of my presentation. (16:40) I call it, are ideas harder to find overall? So here is a picture of Peter, that's a photo picture of Peter Howitt. And I met, so I always tell the story. I have 40 minutes, so if I start making jokes, I think it will be way more than 40 minutes, and you'll be fed up. (17:02) But the thing is that the great, you know, I always tell that story. When I was a Harvard student, my dream was to enter the Harvard Society of Fellows, because it's a kind of first-class post-doc. You are paid like an assistant professor to... to just do research and have dinner with Amartya Sen or wherever. And everybody wants that. So I was shortlisted for the oral exam. (17:28) And usually, I failed my oral exam. I failed many, many exams in my life. And in particular, Robert Solow, by the way, was in the jury. He was a senior fellow. He was a MIT professor. But he was also, at the time, a senior fellow at the Society of Fellows. And he told me at dinner time that I completely missed the question that was asked of me. (17:46) And so forth. I didn't get the Society of Fellows. And as a fallback, I went to a very bad place called MIT. And that's a story that MIT loves, of course. And my neighbor was this guy. And my door was open. His door was open. We would talk. We would chat. He came. He was visiting for one year from Western Ontario. (18:11) And he was coming to work on search and unemployment with Blanchard and Diamond. And one day, I went to his office and said, look, Peter, I think we should, I heard about Schumpeter, but there is no Schumpeterian model. Why don't we try? from scratch to write a model that would encompass Schumpeter's idea of creative destruction. (18:38) So creative destruction is the term that Schumpeter used to coin the process whereby new innovation displays old technology. So Schumpeter's always that progress comes about because there are all the times people who innovate, they build on previous innovation, but they make the previous innovators obsolete. (19:01) process. That's what drives. But he had no model. I mean, in my students' days, the leading model was the solo model. And already Antonin explained very well what the solo model. I think you all know the solo model, no? You all know the solo model? In a nutshell, you can say, well, what is the solo model? Well, it's a model of growth through capital accumulation. (19:22) You produce a flow of wealth with a stock of capital. You save part of the wealth you produced to to increase the stock of capital so that you can produce more tomorrow. And that's the source of growth in the Solow model. But Solow himself was aware of the fact that there are decreasing returns to capital accumulation. (19:44) That means that if you have no machine, to move from zero machine to one machine, you increase production a lot. But if you move from 10 machine to 11 machine, you don't increase production by much. So you cannot grow forever just relying on capital accumulation. You need something else, which is technical progress. (20:01) But Solow would not tell you where technical progress comes from. You would talk about the Solow residual. So we say, well, you can grow because the stock of capital grows. You can grow because population grows. And then there is the rest, which is... a black box called the solo residual and that you don't know where it comes from and our obsession with it was to open the black box of the solo residual so what we did with how it we said well you know let's have a new paradigm where the first idea is that long-run growth is driven by a cumulative process of innovation where each (20:32) innovator based upon previous innovations the second idea is that uh oh there is a problem oh yeah that's the you're right we don't need this It's redundant. But we need a glass of water. Thank you very much. The second idea is that innovation do not come from heaven. The results from entrepreneurial activities were motivated by the prospect of innovation runs. (20:59) If you innovate, you invent a better product or a cheaper way to produce. And that gives you runs for a while, until you are superseded. And that's what motivates you to innovate. That's something I tried to explain to some of my French colleagues, whom I won't name. But you know who I'm talking about, okay? Now they missed the second part. (21:22) You see, they learned growth with solo. But it's a misuse of solo. It's truly true. I mean, some of them, they're very wonderful people. But this they didn't get. That they didn't get. Okay? So, and the third idea is creative destruction. New innovation displace old technologies. New innovation make old technologies obsolete. (21:43) So, of course, at the heart of the paradigm, there is a contradiction. On the one hand, you need innovation grants to motivate innovation active investment. But on the other hand, these grants can be used, exposed by previous innovators to prevent subsequent innovation. Because they don't want yesterday's innovators, they don't want to be subject to creative destruction themselves. (22:04) And so that's... the contradiction. You need the innovation runs, but you want to make sure that yesterday's innovators will not use their runs to prevent subsequent innovations because they don't want to be subject to creative destruction themselves. So how do you manage this? Of course, competition policy is very important. (22:21) And regulating your market economy is largely about how to manage this contradiction. In fact, it comes all the time. Most of the work we do, you always come across this contradiction. Another thing is that So you need competition already to make sure that all the time you have new talents coming in. (22:38) You don't want yesterday's United to prevent new talents. Already competition policy is there to make new talents come in. But I could look also, you know, to which extent more product market competition would induce more innovation. And I always give the parabola. You are a classroom. You are all equally good. (22:55) But suppose some of you are the top of the class and the rest is the bottom of the class, the bottom half of the class. Suppose I bring a very good new element into the class. The top of the class will work harder to remain the top of the class. The bottom half of the class, they are already discouraged. (23:10) They will be even more discouraged if I bring someone very good. It's the same. Same with firms. The blue line represents firms that are close to the technological frontier in their sector. They react to more product market competition by innovating more. They innovate to escape competition. And for them, competition is a boosting force, is a boosting driver of innovation. (23:34) But the orange firms are firms that are far below the frontier, the technological frontier in their sector. They are discouraged by that. more competition, they do less innovation. But if you have a more developed economy, a more developed economy has more blue firms compared with orange firms. So the more developed the economy is on average, the more growth enhancing it is to have competition. (23:57) See, competition becomes increasingly important as a driver of economic growth, the more developed an economy is. OK? That's the basis. That's the basic law. OK, so now in a nutshell, and I don't want to bother you with the model, but it's just to say, you know, the way it was said, you have final good. (24:19) And you produce a final good with an intermediate input small y. And A is the productivity of the intermediate input producing the final good. So A is the quality of the intermediate input in producing the final consumption good. And innovation in this model turn A into gamma A, where gamma is greater than 1. Each time you innovate, you improve the quality. (24:41) You invent a new intermediate input, which is of a higher quality than the previous intermediate input. You drive out the producer of the previous intermediate input, and you become the new innovator. You make runs. You get monopoly runs until yourself, you are being superseded by someone, a subsequent innovator, that will go gamma times better than what you're doing. (25:01) And that's the first. that's what drives growth in this model and of course innovation in this model we assume that you can produce to produce small y you need labor and suppose you use one for one one unit of labor gives you one unit of intermediate input so small y is also the amount of labor used to produce intermediate input but there is another use of labor which is research and the idea there is that you have a Poisson rate of research and if you have z amounts of labor invested in in research and development, but new innovations come at rate lambda z. (25:36) You see what I mean? And so there are two. So you have a first equation that tells you that labor supply is equal to small y plus small z. That's the labor market clearing equation. And then you have an equation that tells you that in equilibrium, you shouldn't be indifferent either to work in manufacturing, to earn a wage, to produce the intermediate input. (26:02) or you should be different between this and doing research. And that's what we call a research arbitrage equation. So the labor market clearing equation, together with the research arbitrage equation that tells you that you are in equilibrium and different between being a worker producing small y or doing research and getting value of innovation, when you do that, put the two together, you determine the equilibrium amount of R&D as a function of all the parameters of the model. (26:30) So it increases with the productivity of research lambda, with the size of innovation gamma. It decreases with interest rate. When you have a high interest rate, you usually are more short-sighted. You are less forward-looking. You do less research. (26:48) And very interestingly, it depends upon the supply of labor, L. And that's why we call that the scale effect. And once you have that... The growth rate, the average growth rate, is lambda z log of gamma. So when we got this formula, when we got the research arbitrage equation and the lambda z log of gamma, Peter Howitt tells me that I told him we are on for the Nobel Prize. (27:08) And that was 38 years ago, almost 39 years ago. And we did the model in three months. It took 38 years to get the prize. So we have to be patient. We did a model between December 87 and end of February 88. We had the basic equations. We had that by then. But then 38 years. OK. But that's the solar residual. (27:32) You see, what's nice is that now I know what the solar residual is. It's the frequency of innovation, lambda z, multiplied by the log of the size of innovation. And that's where the z is the equilibrium z that you have here. And so now you open, we now, what used to be this black box of the solar residual, now you know what it is. And that's the basic model. (27:57) Okay? Now, one undesirable feature of that is that it tells you that if you had population growth, growth should explode. Because if the growth rate depends positively on the amount of labor supply, if you are in an economy where the demography rises, then you should, and keeps growing, you should have explosive growth. (28:21) And that's not a very nice feature. There are many other features, but the non-zero value feature is this feature. Okay, so that's where we are now. And now I will take it from there and say, well, there is... There is from there a debate, and we'll see how we get into the debate. And the debate I want to talk about is the debate, are new ideas harder to find or not harder to find? So on the pessimistic side, you have Robert Gordon. (28:50) And for Robert Gordon, you know, the view that Robert Gordon had of industrial revolutions is like a fruit tree. When you have a fruit tree, the easy, the most juicy fruits are those you get right away. And then you have to climb to get more fruit. Usually they are acid. They are less tasty. You see what I mean? And you may fall, by the way. (29:11) And he had that view of the... He had that view of the sequence of industrial revolutions. He said, well, you know, growth started with the steam engine revolution in the early 19th century. That was a big deal. Then the second industrial revolution, it was the electricity and combustion engine revolution that took place in the U.S. (29:37) in the 1920s, 1930s. That was still a big deal, maybe a bit less than the steam engine. And then you had the IT revolution. And Gordon's claim was that the IT was much less of a big deal than the steam engine and electricity. Now I don't know what he would say about AI. That might change his mind. And that was his view. (30:03) And so he had the feeling that, you know, we may run out of ideas, that somehow growth, we were very lucky, we had the steam engine revolution, and then we run into decreasing returns, and therefore eventually growth should be a parenthesis. You see what I mean? That's Gordon's view, or it used to be Gordon's view. (30:21) My feeling is that now with the AI, he doesn't think that way. The opposite view was from Joel Mokir. who we had the great privilege to share the prize with, and Joel, and it's very interesting because Joel Mochir and Robert Gordon, they are neighbors. (30:42) They are in the same department at Northwestern, and their office is next to each other. And they are very good tops. They don't fight each other. They invite each other. They like each other. But they don't think alike on this issue. For Mochir, new paradigms keep coming. emerging, and you never run out of ideas. And what I want to do is to take you on this debate. (31:03) Say, who is right and who is wrong between Gordon and Mokyr? Are you there? OK. So I will revisit both theoretically, although I will have very few equations. I just slashed the equation before, you know, like half an hour ago. You don't know what you are escaping from. And then empirically, okay? So let me start from the first generation of innovation-based growth models, like the Aguillon-Witt model. (31:36) But you saw the Aguillon-Witt model, G, the equilibrium growth rate, depended, was increasing in the size of population. That's what I call a strong scale effect, okay? So the problem with the first generation is that if the population size grows, you would have explosive growth because your growth rate would go up. (32:00) But that then came Chad Jones. Chad Jones is a very good colleague and a fantastic growth economist. Each paper, every year he produces one or two papers which are jewels. And I'm sure you've read a work by Chad Jones. He's a very deep thinker and always doing fantastic growth economists and economists more generally. (32:24) So what Jones did, he said, when you... This doesn't work quite well, because if you look at the total factor productivity growth of the U.S. since the 1930s up to the 2000s, you don't see any upward trend, even though the population, and in particular the effective number of researchers, the L, if you want, rose. (32:47) kept growing. So if we were right, if the agronomic model was OK completely, you would see that the growth rate should have also increased, like the population. And that's not what you're observing. So he said, well, there is something wrong there. So Jones, and then in the mid-'90s, we had that problem to solve. (33:13) So what he would say is to say, well, you know, instead of saying that A dot over A is an increasing function of the amount of labor invested in research, which is a function of the total amount of labor, I will put an A to the phi minus 1, where phi is less than 1. (33:36) That means I will say that the higher the achieved technological level, the more difficult it is to keep growing. the more labor I need to achieve a certain growth rate of productivity, the higher my current level of productivity. And he said, if you do that, then it's OK, because you get, at the end of the day, you get a balanced growth rate equal to n. n is the growth rate of l, of the supply of labor, divided by 1 minus 5. (34:04) And that would be in line with what I showed you. I showed you population growth. n is the rate of growth of population. And I showed you a constant growth rate. And here you get a constant growth rate. and you reconcile, that's what I call the semi-endogenous growth model. The problem with this model is that growth no longer depends on... (34:26) policy. You see, before I had the growth depending on policy. I could say, well, you know, I want to increase the productivity of research by investing in universities. I want to, competition would play a role, whatever, you know, monetary policy might play a role through R, whatever. (34:47) And all those things would impact on the growth rate, competition. And here, the problem is that in this semi-endogenous growth model, You know, you lose policy. Policy doesn't matter anymore, okay? So that's a bit frustrating. So to respond that there was a, in fact, there was a second generation of innovation based on a just growth model. Howit had himself a paper in the JPE, Journal of Political Economy, 1999. Alwin Young had a paper in 1998. (35:17) Segestrom had a paper in 1998. And what they said is like, look, you know, in fact, the way to do is to say average productivity is the average of productivity growth. over the varieties. And as population grows, so does the number of varieties. So what will happen is that your population grows, but that means that you just have more varieties. (35:42) That will not make the growth rate go up by itself. You see what I mean? And that will give you an aggregate productivity growth, constant and endogenous, equal to its average over all varieties. And all that will happen with population growth is that you will have an increasing number of varieties. (36:00) You see what I mean? Instead of having one brand of perfume, you have 10 brands of perfume. You see what I mean? But that doesn't make you grow faster. That was the idea there. But it would mean that, you see, on each variety, you would have... no constant aggregate productivity growth. You see what I mean? On each variety. (36:21) OK? Ideas will not be harder to find on each variety. OK, that was the try. But then came Bloom et al. So Bloom is Nick Bloom. He's a fantastic researcher. He's at Stanford. And I had the chance, you know, when I showed you the escape competition effect, we had this empirical paper with him, Blundell, Griffith, and Howitt. And he was our student at UCL at the time. (36:49) I guess you met him at the time, probably, no? You were close to that, were you? Almost, no? With Nick. And when they say that, well, look, you know, you are claiming that, OK, you have more and more varieties, and that within each variety, you should have a constant growth rate that you would not run into decreasing returns on individual varieties. (37:13) And he said, well, look, let me do something. So it's Bloom, Jones, Van Rynen, and Webb, okay? So what they said is, no, let's look at particular activities. For example, the production of microchips and transistors. There again, you know, there is the Moore law. The Moore law is that, you know, you double the number of transistors. (37:38) I forgot every how many years. I doubled the number of transistors per chip. And what they showed is that to get Moore's law, you need much more researchers than you did in the early 1970s. You need 18 times more researchers today to double the number of transistors per chip than you did in the 70s. (38:00) So even if you look... At the micro level, not at the macro level, even within varieties, ideas seem harder to find. You see, even within varieties, research productivity seems to go down. Whereas the whole idea was to say the thing would go down is that with more population growth, you have more varieties. But within each variety, ideas don't become harder to find. (38:23) And here they say, well, even at the micro level, ideas become harder to buy. And they look at the production of microchips, but they also look at corn and say similar thing in agriculture. The yield growth of corn doesn't show any upward trend, yet you need increasingly more researchers. So there again, it's another sector. (38:53) And another sector is biotech. You say, well, you know, that's the effective number of researchers has gone up, but the research productivity, which is in fact the economic growth in that sector divided by the number of researchers, has gone down. So even within individual sectors of the economy, ideas seem to be harder to find, unlike what the second generation was assuming. (39:23) So what we did is that with Antonin, our colleague Timo Boppart of the University of Zurich, and he's also at Stockholm, and Jean-Félix Brouillet, who is at HEC Montreal, we said, no, that's not the end of the story. Because in fact, the way things happen is that it's true that within lines, you get decreasing returns, but you reset the innovation clock all the time. (39:49) And so the idea is that it's true that within a line, Once you've created a new line of product, a new variety, you can increase quality on this variety, and that's what I call development. And we can accept the idea that there will be decreasing returns on developing a particular existing line, a particular existing variety. (40:14) But research continually ensures new products, new product lines, with fresh development opportunities, effectively resetting the innovation clock. So the idea there is to say... Gordon or Jones, they are right within a line. But we keep discovering new lines. And that's why we reset innovation. And you see that, we'll get to economics, but you know that very well. (40:35) There are domains that were very much in fashion. General equilibrium theory, when I was a student, everybody was working on general equilibrium theory. And then it fell out of fashion. It did like this, and then like that, okay? Then you had game theory. Then you had incentive theory. Then you had contract theory. (40:54) We use a lot of contracts, but as a research domain, it has gone down. Of course, creative destruction will never go down. That's of course, because we... It's an inexhaustible subject. But you see, that's the argument. So on each line, eventually, development opportunities are exhausted, which will lead to technological stagnation and exit. (41:18) So the world is as follows. You have, I create a new variety. I use labor to produce, but I also use more labor to increase the quality on my variety. I can improve on my variety. I call that development. And that's what I do. But at some moment, I can no longer develop. I face what I call an obsolescence shock. (41:40) That's a strong version of ideas harder to find. At a Poisson rate, at a random rate, whatever, at some moment, you are unable to improve. So initially, you create a new line. You employ labor to produce, but you also employ labor to increase the growth rate of quality of your product. on your line, okay? And for a while, you're booming. (42:04) But at some point, you cannot, you run out of steam. You can no longer. And at the end, and since the rest of the economy is growing and you are not, eventually you exit. And the fact that you exit leaves room for new entrants that come in, create new lines, develop their new lines. And that's where the growth process goes. (42:22) You see what I mean? And so what we say is that the exit of those who face the obsolescence shock opens the door to entrants who introduce new product lines through research, unlocking a new wave of development opportunity. And the resulting cycle of research, living with research, they create new lines. (42:43) Once I created the new line, then I become a developer. And then development allows me to improve the quality of the land until I face the obsolescence. I still remain on the market for a while, but eventually I am really no longer competitive at all and I exit the market. And that leaves room for new entrants to come in. (43:01) You see the way it works? Okay? So, for example, take the camera industry. Each film, early film cameras, were developed from simple box models into sophisticated single-lens reflex systems. And that's your improvement. So cameras, you see the old cameras, and then you have the cameras of my childhood, which are better than my parents' cameras. (43:24) They were still cameras. You would develop the film, you would go to a store, and after one week, you would get ugly pictures. Okay, back, okay. But then what happens is, as improvement opportunities in film-based imaging exhausted, research delivered a fundamentally new product, the digital camera, which opened up new avenues for subsequent development. (43:46) For example, sensor technology, image processing, software-driven features, you see. And that's how the thing worked. And just to tell you something, so we developed the model with Antonin and our co-authors. It's really this model where, you know, when you create a new line, part of the labor is to produce, part of the labor is to choose the rate at which your quality grows until you face the obsolescent shock, okay? And then eventually you exit. (44:16) But you have also research activity that says you can employ labor to create new lines. And once you create new lines, the value of the new line will be the expected value of what you will be able to generate as profits on the new line when you create it. And you have a kind of free entry in research. (44:32) So the thing is that this model, you still have the Jones term here, but you have an additional term, which is quality improvement, the rate at which you improve quality. And that rate is endogenous because it will depend on policy variables. And I will not give you the exact formula, but the important thing is that this DQ is a function of, for example, how much you subsidize development of other things in the economy. (45:04) And therefore, you have the Jones term, but you have an additional term, which is endogenous, which responds to policy. That's the thing, okay? And I don't want to give you the formula for this. So it works like this, you see. I enter at this level. I keep improving my productivity. And then I stop. And for a while I remain until I exit. (45:36) Someone else will come later. The thing is that when you come later, the economy has made progress, so you benefit from the progress. That's the spillovers. So you enter at a higher level than I entered because the fact I innovated has moved up the average productivity. So when you will enter tomorrow, you will enter at a higher level because you take advantage of my improvements. (46:02) Okay? So you enter this level, you improve until you can no longer improve, and you exit. And you see how it works. If you take the upper envelope, you have growth. So on each particular variety, you exhaust, but if you take the upper envelope, you keep growing. You understand the way it works? Okay? And there, for example, if you subsidize development, but you move from the gray line to the black line, and of course, you will increase the growth rate. (46:39) You understand that? Okay? Because you will always develop more. So the next guys will always start higher they would have without the subsidy on development. And so on and so forth. And that's how you restore. So you see that the... So that's very much the theory part. You see, to say we can, in fact, Gordon is right and Jones within a line, but Mokyr is right because you have a reset. (47:09) You have an innovation reset all the time. There are new lines, and of course the new lines, they will start higher because they will take advantage of all what has been done before. They will not start at the same stage as you are. Okay? Follow? Okay, now I will move into empirics. I say, well, okay, that's very nice, but, you know, can you give us an example? And I will give you an example, and that's based on work with Antonin, Luc Paloxevich, Raphael Bargon, they are at the Innovation Lab, and Gaëtan Rassenfos is at EPFL Lausanne. (47:49) And what we want to do is to say, well, we would like to look at economics. Remember, I mentioned general equilibrium theory, contract theory, etc. We would like to be able to talk about topics, and we would like to look at the productivity of research, the evolution of the topic, and then we would like to look at the upper envelope. (48:14) So we would like to know, is it the case that ideas become eventually harder to find within topics? But is it the case that the on-handing emergence of new topics keeps growth happening? And we start here with economics. We are currently doing chemistry. And we want to look at other scientific subjects. (48:39) So are you there still alive? Are you still alive? Are you? Because I hear coughing. When I hear coughing in September, it's not a very bad sign. You know, when you hear coughing in February, you can think, well, you know, it has to do with the weather. But today the weather was beautiful, so coughing now with the nice weather, I say, oh, oh, oh, I'm losing something. (49:03) Okay, so I need your attention. Because so far what I've done is nothing. Now that I need your attention. So let me start from there. OK. So what we do is the following. We look at the economic literature from 1950 to 2020. And then what we want to do is that we use machine learning to identify topics. (49:37) And then we will look at the evolution of the topic. If there is a topic, popularity is rising and then falling. And then we will try to look at the overall, you know... research output in economics. So we will see that we have an algorithm that will identify a local optimum of 90 topics over the period 1950 to 2020. (50:04) And we will see that across the majority of topics, you have a bell-shaped curve, so ideas become harder to find. But still... we get evidence of reset all the time. There are new topics coming, and no evidence that the ideas are exhausting faster in line with Mockier. So within topics, Gordon and Jones are right, but you have this continual, this unending reset of topics. (50:32) So let me tell you how do we... Of course, to identify topics, you could use Journal of Economic Literature classification. But the codes are assigned after the emergence of a topic, and they are assigned by authors, leading to inconsistent classification and strategic manipulation. So what we will do is to use machine learning to determine topics endogenously. (50:52) And that's what the paper does. So how does it work? It works like this. In a nutshell, there will be two slides, a bit rough, and then we think we'll be done. Suppose I fix the number of topics. And we will use abstracts. So we use papers, and we use the abstracts of the papers. So suppose you have k topics. (51:19) That's the number of topics. And now take a paper j, and the paper j will estimate a vector of topic shares. So you know how important each topic is in the abstract. That's how you know. You look at the abstract of the paper, and you can see the weight. of the various topics in the abstract of the paper. (51:43) Okay? And now when you take a topic, you have the probability of a word coming in, given that topic, that's something else you find. You see what I mean? How strongly a word is associated with a topic. So what you could say is that you could say, well, you know, I know the weight of the various topics in the abstract for paper J. (52:08) I know the probability of a word W given the topic. So I could predict the probability of a word in the abstract. And it should be this. But you will compare the predicted probability of a word in the abstract to the true abstract. You see what I mean? And if you start from a, you may start from a guess on the zetas, you will get that something which does not correspond right away to the true probability of words in the abstract per topic, okay? So, sorry, the true probability of words given the paper. (52:48) So what you will do is that given the initial value for zeta and the initial probability of words per topic, you will, you see, You will adjust for the zetas. You see, well, the initial zeta was not good, but I will get to a new zeta. And then I go on and I go on and I go on until you get to the best, to the zeta star. (53:13) You see, you will get, for any paper, you will get a vector zeta star, which such that if you start with your other zeta star and the probability and the optimal... probability P, W, K, such that with those, you get back on your feet. You see what I mean? You say, well, I will indeed predict the exact probability of the words in the abstract of the paper. (53:34) Do you understand that? And that's how you do. You have an iterative method. And that's where the machine learning plays, to find the theta stars. You see, that will make sense that, you know, if I know the importance of topics in the paper, and I know the probability of word of the p of w given topic, I fall back to exactly the share of words in the abstract, you see, of the paper. Okay, so that's the way. (54:03) Do you follow what I say? Yeah, you understand how you iterate? OK, when you have the PWK and the thetas, you can predict the PWJ. But of course, initially, it will not correspond to the true probability of a word in the abstract. So you keep iterating on the thetas until you fall back on your feet. (54:24) OK, so now that's conditional on a K. But now you could say, well, what I can do is that I can let the number of topics vary. And suppose there is a fixed cost per topic in order to compute the optimal number of topics. So there is a trade-off. You would like more topics so that you minimize the distance between articles within a topic. (54:49) But on the other hand, you don't want to have too many topics because the fixed cost is too high. Do you see what I mean? So, in fact, what you will do is that that trade-off will be very important to determine the optimal number of topics. Now, you have also something. You want to have no surprise in the model. (55:07) That means you would like, in fact, for a measure used in machine learning called perplexity. It is a measure of how surprised your model with a set of stars is when it encounters new publications it had never seen before. So if you have lower perplexity, the better the model generalizes to unseen data. (55:28) So if you want a model not only that you can predict, but if I bring new... New articles, new things, it does not invalidate the model. You have a stable model. which is robust to bringing new papers. So when you have a small number of topics, the benefit of adding one more topic is high. You have a very rough model, and each new topic covers a meaningful part of the data that was not captured before. (55:57) So when the K is small, adding one new topic will give you a marginal benefit greater than the fixed cost. When you have K large, new topics can only be built by splitting existing topics. and you pick weaker distinctions, and there the marginal benefit of adding a topic will be less than the fixed cost. (56:15) So you see, now you represent like this. That's your perplexity per number of topics, you see? And you see that initially, when you add topic, your perplexity goes down. If I have only one topic, I will be surprised by having new papers. But if I have more topics, it's more stable. So you will see the public city going down. (56:39) But then the public city starts going up, net of the fixed cost. So you get the kind of optimal number. You don't want the number of topics to be too small, because then you don't capture anything. And you don't want it to be too large because the fixed cost has gone up. And you want not to be surprised by new data. (56:59) And you get, what's nice is that we get an optimal number of 90 topics between 1950 and 2020. So that's the number, we get the number of topics. And then once we get the K, with the method I described before, you get the theta star. And that's what we do. Now you suffered, and now we will harvest. So what I will do now is that we look at the journals. (57:25) We get the papers from publications. So we use a publication provided by the French CNRS. We restrict our attention to journals classified in general economics or specialized economics. We expand the list to include top 300 journals in the 2025 Science Mago list of best-ranked journals. And we select journals available in open, Alex. (57:48) that are in English and provide information on abstracts. So we get 365 journals that give you this number of papers. And now that we have that, and we have 90 topics. Okay. So now we can look at the findings. So first, I will tell you that the popularity follows often a bell shape. So what is the popularity of the topic? It's the share of economic papers for which the theta of this topic K for this paper is among the 10 topics. (58:25) You remember, when you have the zetas, when I have a paper, I have the importance of the various topics. For some topics, the zeta is very high. Those are important topics for the paper. But you have some topics for which the zeta is low. So what you do here is that you want to know what is the share of papers for which the zeta is among the top 10 topics. (58:45) You see what I mean? I would like to know for how many papers, for which share of papers, this topic is among the top 10 most important topics. And that's what I call popularity. It's the popularity of the topic. How popular is general equilibrium theory? How popular is game theory? And what I will show is that out of the 90 topics, on 62 topics, the popularity will do like this. (59:14) It's an inverted U, essentially. 62 topics out of 90. It is either like this or, voila. And then you have 13 topics that are either growing or declining. But if they are either growing or declining, you could say it's still this because if it's growing, you could say it's the sloping part of the bell shape. (59:34) And if it's declining, it's the downward sloping. So you see that there are only 15 topics that are not unimodal. that go like this, you see? All the other topics, either they are directly, you can see the bell shape, or you can see the upward sloping of the bell shape and the downward sloping of the bell shape. (59:51) So for example, here, I look at exchange rate, money, the supply output, that's kind of monetary. It's macro, real exchange rate, monetary, and you see that you have the bell shape on this topic. But the topics are generated endogenously. You have, for example, price, inflation, commodity. That's another topic. You get more or less a bell shape. (1:00:15) And here, what I'm doing is that I average, I look at various decades, the 1950, 1960, and what I do is that I average the trajectories in popularities over all topics that follow a bell curve and peak during that decade. And you see when you average, you still get the bell shape. Okay, so the average of the bell shape are still the bell shape. (1:00:39) So I took all the average of all topics whose popularity peaks over the various decades, and you still find the bell shape. And here is another topic. Here you have the increasing part. It's very interesting because when you have credit economy, macroeconomy, maybe that's a result of the financial crisis rising here. (1:01:04) And here is either theory, for example, function, utility, maxim, et cetera. And you see the bell shape very much. So that's the bell shape thing. What I can do next, because I don't want to bother you. I can show now the growth. So here, what you have is that you look at the number of papers for each topic. (1:01:29) I look at the number of papers, not the share of papers, the number of papers for which this topic is among the top 10 topics. And you see that the number is growing. You see? So you see that the popularity in terms of share is bell-shaped. But in terms of number of papers, even topics that stop being the most important, in terms of absolute number, they still keep producing, you see. But they are growing less than other topics, but they are still growing. (1:02:00) Now you could say, well, how about the quality? So what you could do is to say, I will now look at quality. I will look around the peak, how the quality evolves. So you look at the evolution of quality around the peak of the topic. And what you can see here, I look at citations. You see? Even after the peak, quality keeps growing. (1:02:34) Okay? So you can measure quality by three-year citation or five-year citation, citation after three years, citation after five years, and you see that it's growing in quality. Okay? So you have no evidence that publications are becoming more incremental over time. Okay? Now you can look at originality and generality. (1:02:55) So you can look at how general or original you are, and each what you can look is that you can see before and after the peak, and you see that before and after the peak, after the peak, originality does not go down, and generality does not go down. And I will look at another measure, backward and forward similarity. (1:03:20) What is backward similarity? Forward similarity, you want to be forward similarity. You want when you produce to have many other people subsequently doing like you because it means that you had impact. And what you see is that in years relative to the peak, the forward similarity goes up. Although the backward similarity, the backward similarity means that you are different from previous. (1:03:47) You would like your backward similarity not to grow because you would like to be not less original than before, but you want to be more impactful than before. And that's exactly what we observe. We see that you have growth in the forward similarity, not in the backward similarity. So over time, papers become not less impactful, even though the originality doesn't fall. (1:04:22) I don't know. That's about the growth. So I gave you evidence that the share of the popularity of a topic measured by the share of papers where it's among the top 10 public topics do a kind of bell shade most of the time. Then I showed you that. And in terms of counts, even though in terms of share, you have a decline, in terms of counts, it doesn't fall. (1:04:48) It keeps growing. Or in terms of quality, because you could share, you could say, well, you have more and more papers, but that are less and less original, less and less impactful, et cetera. We saw that that's not the case. Now I would like just to finish up talking about a new reset. So here, what I do here, I do the following. (1:05:06) I say that... I look at top two combinations. So I will retain a combination of two topics if for at least 10 papers, these two topics were the main two topics with the highest status. So if I take two topics, K and K prime, and I find that this K and K prime were among the main two topics for at least 10 papers, I will say that's a big combination. (1:05:36) And I will keep it only if that lasts for at least five years. So if you find two topics that are leading for at least ten papers, not only this year, but in the following five years, you will say this combination of two topics, I keep it. Okay? What this shows is that... Every year, you had new combinations of two topics that fulfilled this requirement, that were the top among at least 10 papers and remain so far at least, and not the same as before. (1:06:07) And you see that you keep having all the time new combinations of two topics that are top in a lasting way. So that means that you have innovation reset, you see. You keep having new combination of topics. So many of them were not seen before, so it could be completely new topics, or it could be two topics that were not combined before. (1:06:30) But you see that the combination of two topics, you keep having combination of topics that are... All the time. So that's the innovation reset. So you see, so you have within topic the bell shape most of the time. But in terms of growth, in quantity and quality, you even after the peak, in terms of numbers or quality, topics keep growing. (1:06:52) And then what you see is that you have this continuous reset. So now we are doing the same with chemistry and with other fields because we can use the same methodology. So Gordon and John. that ideas may become harder to find within particular fields or topics, yet due to the reset, ideas don't seem to be harder to find overall, as Mokyr would have argued. (1:07:20) And that's the big debate that we have with Jones. And now you see why is that important? Because I don't know if you studied semi-endogenous growth model. Now we go beyond semi-endogenous growth. The debate now, it took 30 years to conclude the debate. How long do I have left? Who is timing me? Zero? One minute. (1:07:41) Okay. So AI and growth. So I don't have much to AI and growth. What I can tell you is that AI stimulates growth, not only because AI automates tasks in the production of goods and services, because AI makes it easier to find new ideas even more. Even without AI, ideas have not become harder to find. (1:08:01) But of course, AI boosts the production of ideas. New ideas are often recombination of old ideas. And with AI, you can recombine much more. So for example, Antonin has looked at patents granted in the US from 2010 to 2025, excluding those that are in computing data processing. And then what you say is that you create a panel of CPC codes, which are technological codes, and you say, well, I want to know for each CPC code, I have a measure of AI exposure, which is a share of patents that touch any computing (1:08:47) data process. So you have an exposure to AI. And what Antonin shows, in a nutshell, Antonin will explain much more than I do, much better than I do, is that whenever a code, a technological class is exposed to AI, to AI, in fact, it boosts the patenting. That's the thing. So AI boosts the production of patents. (1:1)(1:09:15) Whenever you become expose d to AI, it boosts your production of patents. So my view overall is that, technologically speaking, I believe that already without AI, ideas were not harder to find overall. But with AI, even less so. It's even easier to find new ideas. Now you have constraints on the process. (1:09:42) One is human capital. We may lack human capital. Education. If we have a bad education system, it will be a limit. Limited freedom. You have countries like Hungary or other countries where you limit the ability to find new ideas because there are political constraints and competition. And just to tell you, but I don't want to spend much time, but when you had the IT revolution, initially it boosted growth enormously between 1995 and 2005 in the U.S. (1:10:12) , but then growth went down because you had the emergence of superstar firms that became Google, Microsoft, Amazon. The concentration increased a lot during the year of the high growth because this firm expanded a lot through merger and acquisition and through building many more establishments. The number of establishments for large firms rose much faster. (1:10:33) That's the red line than the number of establishments for small firms. So this firm became very pervasive. But as a result, entry of new firms was discouraged. So what happened with the IT is that initially the superstar firms emerged. They initially boosted growth in the U.S. But eventually, because they became so pervasive, they ended up discouraging entry of new firms. (1:10:53) And that's why you had the decline. And why is the problem with AI? Is that if you look at the upstream segment in the AI value chain, for example, the cloud is dominated by Google, Amazon, Microsoft, and you have only one big actor for the market for graphic processes, the upstream segments of the AI value chain are very much dominated by few firms. (1:11:11) And there is a danger that this kind of thing could happen. And that's where competition policy is important. So I just want to tell you that... My view is that ideas are not harder to find overall, but inappropriate institutions can hamper the process. So my worry is not on the technology. My worry is on the institutions. (1:11:33) If you have lack of freedom, if you have bad schools, in France nowadays we are facing a big problem with education, if you have inappropriate competition policy, that's where you are unable to harness this fantastic force which is to... reset the innovation clock and keep inventing all the time new ways, new product lines, new research paths that will boost new ideas. (1:12:03) So thank you very much. Thank you. Thank you. Great. Okay. So good evening, Professor Aguillon. Thank you so much for your time and this illuminating talk. Although, so we escaped from your difficult equations. I understand you tried to escape from our questions. But so we would like to follow up on what you've shared and ask you a few questions for which we hope you'll forgive the simplicity given the precision of the lecture you've just given us. (1:12:56) So I'll let Lou ask you our question. As we've understood, and in response to the title of your talk, ideas are not becoming scarcer. But isn't the real risk more than the depletion of ideas, the absence of industrial policy, as you mentioned at the end of your talk, and competition policy in particular? In other words, what framework, in your view, is most conducive to creative destruction at the level of the state? But you need, I mean, if you believe in the framework, the idea is that you generate growth by having all the time new talents. (1:13:31) And the new talents, they create new firms. And they grow their firms. Of course, the problem is that when they grow, first they have to be able to grow. So you need an ecosystem where they can grow. So, for example, in Europe, there are many startups. They don't grow. They go to the U.S. because they have venture capital. (1:13:48) It's usually a bigger market. We have a very segmented market in Europe. So they don't grow much in Europe. And they go and grow somewhere else. So you would like to get talented people who created firms. They can grow their firm. (1:14:04) But the problem is that you want to make sure that once they've grown up, they won't use their power to prevent entry. And that's why competition policy is important. And, for example, if you look at AI, I told you in the upstream segment of the AI value chain, you are too few firms. You want more competition. Open source, I believe, very much. You know, the Digital Market Act that insists on data sharing. (1:14:26) You should have more data sharing to facilitate entry, et cetera. But at the same time, you may need industrial policy to build computing power because AI, it's both data and computing power. Data, we have plenty in Europe, but computing power, we don't have enough. can in fact reinforce competition policy. (1:14:44) It was very much the view for long that, you know, the two were counteracting. that if you believe in competition policy, you should absolutely avoid industrial policy. I believe you can have a competition-friendly industrial policy like the DARPA model in the U.S. or like France 2030, the competitive part of France 2030. (1:15:01) There will be a note, an unsaken part, a big part, showing that France 2030, the competitive part of France 2030, in fact, works very well. So that's my view. Although I'm a pro-competition person initially, but I believe some people rejected industrial policy because they believe in competition, and I think that's extreme. (1:15:29) You also spoke about education as a fundamental pillar of innovation, and you are a professor at the Collège de France and London School of Economics, so this is a subject that concerns... Education. Sorry? Education, yeah. So yet the French school system is built on a very early and fairly rigid selection in comparison, for example, to the American university model, which is more flexible. (1:15:51) Would you say that the French school system allows for the kind of social mobility needed for the disruptive talent to grow? Yeah, I will tell you, it used to. I mean, we used to have a French, you know, primary secondary schooling system that was It's not performing as well as it used to, and it' s not being the social ladder that it has been in the past. (1:16:12) And you try to understand why. I think a big part is that the iPads and the social networks have terrible effects. Children spend too much time looking at the... And they have parents also who spend too much time on the mobile phone. And I think that's a big pity. I think it's very important to go back. (1:16:32) That's why, you know, the Americans would tell you about the no-excuse chart of schools. It's schools where you emphasize writing, reading, grammar, and basic calculation, and without, no AI, no iPads, no nothing. You just won't concentrate. And I think the important thing is to have that done at school. (1:16:53) So I think it's very important to have classrooms that are not too numerous, well-trained teachers. I think that's very important. And that they should concentrate on basic. And the homework should be done at school because now increasingly less parents help their children. They are on their mobile phones. (1:17:13) So I think that's what has to be. Because when you have a good schooling system, you get... First, anybody, even children from poor backgrounds, can get very far. We have the thing that when Xavier Jaravelle calls the lost Einsteins or the lost Marie Curie, many smart kids born to families with none. (1:17:35) with poor families that cannot bring them the knowledge and aspiration to become entrepreneurs, researchers. And the schooling system in France used to be able to make up for that, no much less because, in fact, you see that children no longer concentrate on reading, writing, demonstrating a theorem, and I think that's a big problem. (1:17:58) It's a student issue for you. Yeah, that's right. Next week, we will be welcoming Jean-Marc Jancovici, the French engineer. He argues that economic growth has historically been tied to fossil fuel consumption and that the combination of climate constraints and resource depletion makes continued growth impossible or realistic and achievable. (1:18:19) In your view, is this thesis a forced degrowth? ve hopes or hydrogen or whatever that will slow down the you know the rise in temperature then you will have innovation to better adapt to climate change better air conditioning system we saw you know the importance of air conditioning last summer many people died because of lack of air conditioning we need to have cheaper and maybe more energy effective air conditioning we need we have to build dikes because the sea level is Those are adaptation innovations. I think there are many things that the degrowth advocates say that are accurate. They say we should become more sober. We should not keep our lights on. We should sort out our garbage. We s hould maybe not fly every day. I think they're right. They also say that we should move from quantity to quality. (1:18:45) I believe very much. They also say that measured GDP growth does not capture the true... growth in Livingston. All those things they say are right. But I believe that we will need still innovation. And what I call innovation is innovation in our behavior. But it's also innovation to find new sources of energy that are cleaner. (1:19:06) source of energy like nuclear fusion i still ha (1:19:37) There are also innovations to cool down the air. I'm sorry that you will find that crazy, but the physics department at Harvard works on geoengineering. When you have a volcano eruption, you have particles of sulfur in the air, and that makes a screen between us and the sun, and the air cools down. They are trying to master this technology to say, well, could we? use this technology to cool down the air. (1:20:00) There is another idea of giant shades in space that will also play the same role. I don't know. And also the idea that you should move from quantity to quality. So maybe one hope is that as we become richer collectively, we demand more quality than quantity. And therefore, the innovation will adapt to our demand. (1:20:20) And we'll also move towards more quality enhancement rather than quantity. And maybe we can also speed that up through policy. And so those are all the ways we have to. I believe that carbon tax alone will not do the job. It's important to have carbon tax to make sure that once you have the alternative clean technologies, then. (1:20:37) People move away from the dirty to the clean. And that's where the carbon tax will be used. But if you don't have the alternatives, carbon tax alone will not do the job. So you get the yellow vest movement. The yellow vest movement, you have these people who live in suburban areas in France. And suddenly, you increase the price of gas oil. (1:20:54) They were using gas oil cars. They had no alternatives. And they revolt, of course, because there was no trams. There was no S-bands. There was no nothing. In Switzerland, you could do it because you have a very developed system of S-bands and trams. (1:21:11) Speaking of policy... So my response is that to say, innovation, if you don't believe in innovation, you become Malthusian. If you believe in innovation, you see that the world is not finite. We could find also in other planets, we can find, you know, materials. And when you believe in the world, but the whole thing is to direct innovation in the right direction. (1:21:28) So which policy tools, because it's not, firms spontaneously will not innovate in the right direction. That's where you need carbon tax, green industrial policy to steer innovation in the direction you want. But I think that's my difference with Young-Avish that they don't believe much in green innovation. (1:21:44) in the broad sense that I described. So we understand that state and policies have a huge role in this. So as we enter in a French presidential campaign period, I would like to hear your perspective among the declared candidates. Is there any economic policy that you find credible? And also, what would be your very first move if you were to be appointed French minister of economy? In France, we have a debt problem, a public debt problem. (1:22:19) We cannot ignore it. So to be able to borrow at lower rates, because we need to borrow if you want to invest in education, research, and all that, we need to have our public spending grow less fast than our GDP. I think that will be very important. We have to show credibility. It's not huge what we need to do, but we need to do it. (1:22:39) So there are various things we can do in the short run. In the short run, we need to show that we can take some measures that other countries have done, but it has to be to make our public spending grow less fast than GDP. In the longer term, we have a problem of employment rate in France. The employment rate is too low. (1:22:57) compared to Germany, for example. If we had the same employment rate as Germany, our debt problem would be much less than it is. So it's partly the fact that we don't work long enough in terms of, you know, we retire too soon. But how we deal with that? not the way it was done in 2023 when we put this compulsory age. (1:23:14) I think we need to put more flexibility and to factor in painability much more. But we need to increase the effective age. So that, I think, will be important. And for the young, we need to improve still the training systems and the schooling system to make. And maybe we need a better flex security system like in Denmark to make sure that the employment rate of the young people goes up. (1:23:39) And then the rest will be productivity growth. So that's where we need an ecosystem. For example, in France, we have to, but we can increase the amount of long-term research funding. We don't have enough venture capital. to boost, to encourage risk-taking by small firms. We don't have enough institutional investors to encourage. (1:23:58) And the big problem we have is that we have lots of savings in France and Europe, but the savings are invested to finance pensions and innovation in the U.S. So that cannot go on. So you need to say, well, for example, you have life insurance in France. If you invest in the U.S., you will not be taxed the same way as if you invested in Europe. (1:24:14) I think something has to be done there. So there is a whole kind of thing at these various levels to deal with our debt problem, but at the same time to put us on the track of, like the Draghi report, to say, you know, we need to get on this track of creating the ecosystem to move from mid-tech incremental innovation much more to breakthrough frontier innovation. (1:24:37) And that's something we can do. Speaking of the Draghi report. One last question. You played an active role in drafting it. It was published in September 2024. It paints a stark picture of the European Union's economic and technological decline. On the economic front, is Europe going far enough, and can we consider a common economic policy, or at least a common innovation policy, to spur disruptive innovation across Europe? I'll tell you something. I'm very pessimistic about what we can do in 27 countries. (1:25:11) I think because there is a lot of political economy. We don't have a single market because each country has its own regulation to product market regulation to European product market regulation. And that's also very much for political economy. You want to win elections. You want to show that you are tough, that you do protectionism and whatever. (1:25:29) And we face this problem in six months. So I think... I believe in the coalition of the willing. I believe that research part, the ERC, there I think we could do something. The ERC should become more independent, the European Research Council, and should be able to fund research on the long-term basis. (1:25:47) Fund like the LABEX, Antonin has done this very important work on the Laboratoire d'Excellence, showing that wherever you have Laboratoire d'Excellence in France, it boosted breakthrough innovation in neighboring industry. We need Laboratoire d'Excellence for... in Europe in general, and that should be managed by the European Research Council. (1:26:04) So that would be something we could do maybe. There is the idea of the 28th regime to make it, you know, if you are new firms, but I think whatever can be done there. I think the rest should be done by a coalition of the willing. You could imagine France, Germany, UK, some other countries say we do John DARPA, equivalent of the Defense Advanced Project Agency, which is this way to reconcile industrial policy, competition policy in areas like defense, AI. (1:26:29) energy transition, we could have joint institutional investors. So I believe that there are policies that we can do, coalition of the willing, with those who want to do things, including the UK, I think, because they are very good in defense and other sectors. I believe much more in coalition of the willing than doing everything in 27 countries. (1:26:49) I don't think they're taking off, except maybe the ERC. Thank you so much for the evening. To conclude with a moment of silence, I'll now turn over the floor to our faculty dean, Andrea Messini. Thank you, Professor Argonne. Thank you, Philippe. It was amazing. It's difficult to go down to her now and to move to the official conferral of the honorary degree. (1:27:21) This is the highest honor that an academic institution can confer to an academic. And it goes to exceptional individuals who made an incredible, outstanding academic contribution, who had also influence on policymakers, on the academic community, on leaders. And I would say from tonight, it also goes to scholars who have passion. (1:27:40) And I think that was pretty clear from your speech. I will not repeat, of course, your academic contribution. That's pretty clear what you contributed in terms of our understanding of economic growth, of innovation, of creative distraction. But I think also what is important is your ability to go beyond academic contribution, your ability to understand and to help us understand the importance of institutions, and you mentioned a few, to understand how public policy can shape and foster and support innovation. (1:28:10) And I think that is something which is pretty much in line with the... DNA of HEC Paris, and that's why it is so meaningful, so important for us tonight to give this award. At HEC, we believe that academic excellence is not an end in itself. It is a means for achieving something greater, which is really contributing to the creation of a shared prosperity, of course, by supporting innovation, by stimulating economic growth, but also by helping decision makers design the right policies. (1:28:38) And of course, innovation and entrepreneurship are important instruments for doing that. I think you also help understand that if we don't study how that innovation, how that disruption has an impact on those who are going to profit from that or those who are going to be disrupted by it, that will not be enough. (1:28:57) So I think your academic contribution, your entire academic life speaks highly in that respect and contributes to achieving those goals. And that's why we're particularly proud of giving you this honorary degree. So in recognition of your outstanding contribution to the field of economics, in recognition of your contribution to studying economic growth, innovation, entrepreneurship, creative destruction, in recognition of your ability to influence policymakers. (1:29:23) academics, but also generations of students, particularly honored on behalf of HEC Paris and the entire academic community to confer upon you this doctorate honorary scouser of HEC Paris. So congratulations and welcome to our community. I'm speechless. Last year, people say I'm the most talkative, speechless person, but I feel very speechless. (1:30:21) At the same time, lots to say. And I'm so grateful to all of you, and Jan, and we are... I very much look forward to joint work together because, in fact, what we are doing in the lab is very much with HSA. HSA is a big part. And so I see that as a milestone in what I hope to be a very fruitful... collaboration. I think we can do a lot. (1:30:52) I should say it's a fantastic place and we have, you know, we already work together and we are doing lots together. So I think it's a fantastic encouragement. Thank you so much. Thank you all for coming. Can I invite the faculty to come on stage so we can take a picture with Philippe and have a great evening all of you. (1:31:29) Thank you very much for being here tonight.