AI Bubble, Inflation, and the Limits of Monetary Policy | Jason Furman

0:00Demetri Kofinas:What's up, everybody? My name is Demetri Kofinas, and you're listening to Hidden Forces, a podcast that inspires investors, entrepreneurs, and everyday citizens to challenge consensus narratives and learn how to think critically about the systems of power shaping our world. My guest in this episode of Hidden Forces is Jason Furman, professor of the practice of economic policy at Harvard University and a former chair of the Council of Economic Advisors who comments regularly on the U.S. economy, inflation, and monetary policy. Jason and I spend the first hour of this conversation exploring his economic framework and his time working inside the Clinton and Obama White Houses and how these experiences shaped his perspective on the proper role of government in the economy.

0:46Demetri Kofinas:We then dive into his thinking about artificial intelligence, whether we're living through an AI bubble, how he thinks about AI regulation, and where he expects to see the biggest productivity gains from the adoption of AI in the US service sector. We also discuss the limitations of our inflation models, whether we have a good working understanding of the causes of inflation, whether the Fed has implicitly raised its inflation target, and how large structural deficits and political constraints will shape the Fed's ability to manage that target in the years ahead. In the second hour, we switch to a discussion about the politics of affordability and a broadening sense, especially among younger Americans, that the social contract of their parents and grandparents is broken and that the costs associated with achieving the American dream have become insurmountable for almost all but the very wealthy.

1:35Demetri Kofinas:Jason and I debate the political appeal and efficacy of price controls and the extent to which tariffs, industrial policy, and currency depreciation can or should be used to reshape global supply chains and rebuild U.S. domestic manufacturing in areas deemed critical for national security. We also discuss the U.S. trade deficit and capital account surplus, what a weaker dollar may tell us about the government's policy objectives, how corruption, rule of law, and institutional decay may eventually feed back into foreign appetite for US assets, and what is at stake in the choice of the next Fed chair, both for the independence of the central bank and for the long-term credibility of American monetary policy.

2:15Demetri Kofinas:If you want access to this part of the conversation and you're not already subscribed to Hidden Forces, you can join our premium feed and listen to the second hour of today's episode by going to hiddenforces.io slash subscribe. All of our content tiers give you access to our premium feed, which you can listen to on your mobile device using your favorite podcast app, just like you're listening to this episode right now. If you want to join in on the conversation and become a member of the Hidden Forces Genius community, which includes Q &A calls with guests, discounted access to third-party research and analysis, and in-person events like our intimate dinners and weekend retreats, you can also do that on our subscriber page.

2:56Demetri Kofinas:And with that, please enjoy this thoughtful discussion about the US economy, government policy, and the long-term structural challenges and opportunities facing the country with my guest, Jason Furman.

3:12Demetri Kofinas:Jason Furman, welcome to Hidden Forces. Great to be with you. It's great having you on, Jason. I'm actually surprised that you've never been on the podcast before. You're a rather prolific commentator on the state of the economy. you've been in the economic policy business for at least three decades. Is that fair to say? Since the mid-1990s. So yeah, about three decades now. So tell me a little bit about you. Let's start even before that. I'm just curious how you got into the economics profession. What led you into this field? And then I'd love to learn about your experience in policymaking. Yeah.

3:47So I grew up and I just loved math and physics, but I also loved the political world and the set of issues that came up in political debates. And so economics was a really natural way to combine those two interests. You could think in a logical, disciplined, and rational way, but about topics that I think in some ways are more fascinating and more complicated and difficult than quarks and leptons, because people just don't behave quite as well as forks and leptons do. So I got into economics. At first, it was in a pretty academic way, and then almost accidentally got recruited for a job in Washington at the Council of Economic Advisors.

4:31Originally, my plan was to take a year off from graduate school, but I ended up loving it. So even though I finished my PhD, I have always since then been on a much more policy-oriented track and way of thinking.

4:44Demetri Kofinas:It's interesting that you mentioned physics as being your other passion. Do you feel like your perception of economics as a more deterministic discipline has waned in the year since you've been practicing it? Yeah, it certainly has. I mean, when you're in grad school, when you're doing your research, you're writing very, I don't know that I'd use the word deterministic models because you can have randomness and propagation and shocks and all of that, but you have completely self-contained, closed, fully understandable system described by a set of equations. And when I think of the world around me, partly I try to bring it back to some of those models, but importantly, I just said models plural.

5:28Don't just reduce it to one single model and combine that with a certain amount of just trying to listen to how businesses think, how workers think, what's common sense, and combine all of that in a way that is admittedly less disciplined than just hewing exactly to one model. But I think probably a better way to think about the world than just living within the confines of one model as well.

5:57Demetri Kofinas:So, you began to work, you said, on the Council of Economic Advisors in when? 1995? 1996. 1996, which was the Clinton administration, the second term. I feel like the advantage of working in politics is that you probably have a more realistic sense of what's doable as opposed to just what makes sense in theory. I'm curious how that experience working in politics. And tell us a little bit about what administrations you worked in, what period of time, et cetera. How How did that shape your perception of the economy of economic policy, what works, what doesn't work? Yeah. So I was in the Clinton administration first at the Council of Economic Advisors and started out doing macroeconomic analysis.

6:41I would help draft memos for the president on a lot of the different data releases that would come out. We'd get the data a day in advance. So really learn how to read data, think about how the macroeconomy functions. But I quickly moved to basically the front office of first Joe Stiglitz and then Janet Yellen, working with them on basically every topic, helping them to write speeches, engage, etc. I went from there to the World Bank and then came back to the White House to the National Economic Council, which is a little bit more political and strategic. Then in the Obama administration, I did the opposite.

7:19I started as deputy director of the National Economic Council, working on issues like combating the financial crisis, our fiscal plans, debt limit, fiscal cliff, all of those sorts of things, as well as technology policy, infrastructure policy. And then in that administration, I transitioned from there to the Council of Economic Advisors, where I was chair and worked on just about everything that we were dealing with. And in Washington, I think it's important to be able to wear different hats. When I was at the National Economic Council, it was a more political hat, more strategic, more what is possible.

7:59The Council of Economic Advisors is a little bit more like an in-house think tank within the White House. And there, I think it's important to be a little bit less political and a little bit more, here's the best option, here's the second best, here's the third best. and not just try to jump straight to the third best just because you think that's more politically feasible.

8:24Demetri Kofinas:Great. So we'll have an opportunity to apply it to a lot of the questions that I'm going to ask you today. So I have a lot of questions to ask you, Jason, as you can imagine, but maybe the best way to start is to just kind of ask you, what is it that concerns you today? Where do you spend most of your time focused on? How much of it is focused on trying to understand where the economy is, how much of it is what needs to be done on a policy side. Feel free to share your thoughts with me and then we can begin to dig into them. Yeah. I mean, I often pay attention to whatever the issue of the day is, often a fascinating issue, an important one, and one that comes up a lot.

9:02And right now, the biggest issue of the day is the AI bubble. Are we in one? What's the possible consequences if we are and it burst? Should policy be doing anything about it now? Should policy be doing anything about it later? I also confess I certainly spend a lot of time on what's going on with President Trump and his policies and the like, but some of that is frankly less satisfying and interesting. That doesn't mean it's not worth doing. It doesn't mean that I don't spend time and effort on it. But when things get very political, all of a sudden it's sort of two warring tribes. And in some case, there's truth and justice on one side, I think on certain issues like the Fed, which we could talk about.

9:49There's truth and justice on one side of the Fed independence issue. But the AI bubble has been in some ways more fascinating to me lately. All right.

9:59Demetri Kofinas:So let's talk about that. I also want to talk about affordability because if we're talking about what's in the news and what's current, I think concerns about affordability and especially their political ramifications for the president are top of mind. I know that this has been raised as a concern among some of the president's staff. The election of Zoran Mandani in New York City, I think points to the salience of this issue, certainly for Democrats. And perhaps there is a sense of urgency on the part of Republicans to develop some kind of platform to be able to respond in the midterms and in the next presidential cycle.

10:32But let's talk about the AI bubble.

10:34Demetri Kofinas:So just calling it an AI bubble focuses, I think, very much on the financial side of this, whether or not there's been over commitment of capital, if prices succeed, kind of sustainable levels. And then there is a larger question of AI policy, which I feel like I'm actually a bit confused about what this administration, and not just this administration, because I don't really know what the position is of Democrats or if there even is a coherent view on how AI should be regulated as a category, as a general purpose technology, what we want to accomplish with our regulations, et cetera. Let's maybe focus on that first.

11:12Demetri Kofinas:How do you view AI as a technology and what do you think the role of the government should be in regulating it? Yeah. So first of all, any Any predictions about the future of AI have to have enormous confidence intervals. So I am excited and hopeful for it. It's definitely enriched my life, both professionally and using it personally to help inform me on all sorts of decisions, large and small. And I think that it's sort of amazing that I'm using the very, very worst AI I'm ever going to use for the rest of my life. It's only getting better from here. So I think it's exciting. I think economically, I'm much more worried about having too little of it than having too much of it.

12:00Certainly, there's risks of a lot of AI and things like the labor market, but I'd rather take those risks and figure out how best to manage them than the alternative, which is condemning ourselves to slow productivity growth. So for me, my starting premise is that this is exciting and we want more of it. Then when I think about public policy, first of all is to do cost-benefit analysis that there may be some costs of AI, but there's also benefits. And you can't just wait until you can prove everything is perfectly safe to move forward, because waiting for that is a way of condemning people to die by cancer and not get tutored in school and all the other things that it can do.

12:49I don't like general AI regulation so much as domain specific. For auto safety, are cars with AI safe or not? Just the way we evaluate the safety of any other car. For stock trading, look at that. For medical devices, look at that. So don't have an AI super regulator, have AI as part of all the other regulatory systems. And then finally, I think it's really important to compare AI not to the almighty, but to the alternatives. And the alternatives are humans. And humans are just incredibly, incredibly flawed in the way they make decisions. They show bias, they make errors, they get into car accidents, and all that stuff.

13:37enough. So don't ask, is the system perfect? Ask, how does it compare to the others? And by the way, an important part of that comparison is even if AI is a bit worse than humans, it's something right now, it learns really quickly. And so in some ways you want to do the dynamic analysis of not just where is AI today, but where is it going and how does the learning process fit into that?

14:01Demetri Kofinas:So, between I would say 10 and five years ago, between 2015 and 2020, the debate among luminaries, commentators, people that were interested in AI and conversations about narrow versus general intelligence, those conversations I think leaned heavily towards the AI safety perspective, which is to say there were a lot of concerns about how these systems would operate in the wild and how we would do our best to try and manage the objective function so that we didn't, quote, lose control of these systems. And I'm not even talking about Skynet. I'm not talking about these systems being embedded in defense technologies.

14:44Demetri Kofinas:I'm just saying in general. It seems that at some point in the last couple of years, really shortly after ChatGPT's launch, maybe a year afterwards or so, it seems like there was a kind of just, we flipped the switch in the public conversation. And we went from this being sort of the paradigm in which we thought about the trade-offs of regulation to just saying, well, we just need to throw as much government support behind these large companies that are leading the development of these foundation models because China, because we can't afford to regulate, otherwise it will turn into Europe. I'm curious from your perspective as a policymaker or someone that's been in the policymaking space and has a lot of relationships with people in and out of government, what do you think is going on here in terms of our policy approach to AI and do you have a view on whether it's efficacious or not.

15:43Yeah. So there were two things a few years ago. One was the safetyism that you said, and then another was a really intense concern around AI being biased, absorbing human knowledge, reproducing bias, doing it against unracial gender and other grounds. If you look at the Biden administration's policy on AI, in many respects, they were quite pro-AI At the very end, they did sign legislation on making it easier to build in that space to get through the permitting. But there also was an awful lot of let's sit down and worry, not just about these safety issues, but all these bias issues and the like.

16:29People I talked to in Silicon Valley think some of that did slow them down a bit. Not a huge amount. They were moving super fast under Biden, but slowed them down a bit. Under the Trump administration, you basically had all of that ripped up and full speed ahead type of attitude. I'm somewhat sympathetic to that. And I think it is one of the few things in the Trump administration on economic policy I'd probably put on the positive side of the ledger, and it's a quite important one, is increasing the, just trying to get rid of whatever bottlenecks, whatever is standing in the way of this technology.

17:06Now, 99 % of what's happening is the private sector moving on its own, and that would have happened regardless of who is president and probably even regardless of what policy. But you look at the United States compared to Europe, and it's hard not to believe that policy matters here too, and that's a big part of why we have tons of leading foundational models and Europe basically maybe has Mistral.

17:34Demetri Kofinas:So one of the limiting factors during the internet boom in the nineties was the bandwidth of analog phone lines, which the industry addressed in part through the transition from analog to digital. Today the consensus seems to have shifted from initially seeing the limiting factor as chip capacity, both in terms of quality and quantity of compute, to now being about the grid and the capacity to service ever increasing demands for electric power by data centers that will be used to train and run these frontier models. And even that thesis is under fire as a result of deep seeks innovations and inference compute efficiency.

18:14Demetri Kofinas:I'm curious, what role do you think the government should play when it comes to things like the electric grid, which are part of the public infrastructure that everyone depends on, including players at the economic frontier, but also everyday consumers? Yeah. I'd say the biggest role for the government to play is getting out of the way. And the main way it's in the way is on some of the issues around permitting that you need to build the energy supplies to build the chip supplies. I don't know, by the way, if the scaling is going to keep working and that you just keep doubling the number of chips and you double how good the models are and you double the economic impact of those models.

19:00There's a bunch of me that is actually quite skeptical of that. And I look at ChatGPT and a lot of the improvements it's making seem to be less about the pre-training and more about refining it and fine-tuning it afterwards. but I'm not an expert on this and I'm a believer in the government doesn't decide. I don't decide. Lots of different companies, lots of different investors make these decisions and then the market will sort out which of them work. So what the government needs to do is make sure that process can happen. It doesn't need to subsidize anything, doesn't need to give grants. More than enough of this is happening.

19:37It just needs to make sure you can build the stuff for at least one path that these companies are going on a path that so far has worked quite well and may or may not run into heavily diminishing returns.

19:51Demetri Kofinas:So I actually feel like this would be a good opportunity, Jason, to ask you a question about your philosophy on the role of the state in the economy. I have been very influenced by Bill Janeway's book, Doing Capitalism in the Innovation Economy, and he has this model that he describes as the three-player game, which includes the state, financial speculators and the market economy, which ultimately is responsible for commercializing and broadening the gains from the investments made by the state at the earliest stages of the innovation cycle. And this model was deployed rather successfully throughout the period of the Cold War.

20:25Demetri Kofinas:But it seems to me that we're at a stage today where the United States for the first time since maybe the mid to late 1970s, when the Soviet Union's growth rate began to lag consistently behind America's, and certainly by the 1980s where the gap in productive investment had grown and the Soviets were spending something like 10 to 20 % of their GMP on defense, that the US is facing a competitive alternative to its economic model in the form of China's state-driven capitalism with authoritative characteristics and a strong bias toward investing in achieving technological supremacy. And there is a kind of ideological war going on, certainly in certain avenues of social media where you see a lot of videos of Chinese investment and infrastructure and all the sort of ways in which they've dominated the electric car market.

21:20Demetri Kofinas:How do you think about what the appropriate role of the state in America and other Western democracies should be, especially in light of this geostrategic competition? And how has your view changed or not changed on this issue in the last, say, 10 years? So I think in the last 10 years, I've almost become a little bit more orthodox in my thinking. Maybe it's that I'm teaching a principles of economics class here at Harvard. Maybe it's looking at all the ways in which things can go badly wrong if you try to get too creative and too cute about things. But the way I think about it, and this is just sort of a heuristic description of my thinking, is if you follow all the conventional recipes, you get to the 85th percentile of economic performance.

22:12If you depart from them, you might turn into Singapore at the 100th percentile, or you might turn into Venezuela at the zeroth percentile and say, you better be really, really sure of what you're doing and really careful. So it's sort of small C conservative about not straying too far from it. Now, what does this mean applied to technology? Well, most of technology is something that the government isn't particularly good at picking winners, losers, isn't particularly good at picking approaches. So largely, I think it's for the private sector to do. But there's two big conventional market failures that we'd teach in a class like mine.

22:54I definitely don't teach like markets solve every problem, but I teach try to identify the externality, try to identify the way in which concentration is causing a market failure, et cetera, and then try to have a remedy that brings you closer to what you would have gotten if the market could have functioned better. So in the technology space, I think there's two and possibly a third. The first is the intersection between technology and national security. The market's never going to make enough drones. It's never going to make enough missile defense systems. You just can't leave it to do that on its own.

23:32The government needs to do that. The clearest place that is in technology is in things like fabricating microchips here in the United States. We're just not going to do that without subsidies. I think that is a good thing to do for national security reasons. And so I'm glad we're subsidizing it. Second is there's huge spillovers in knowledge that aren't captured by any given company. That's a positive externality. And so the government has an important role in both basic research and also has a role in possibly things like tax credits for research and development. There's then a third issue, which I'm less sure about, which is competition.

24:18And I'm a believer in antitrust. I'm a believer that the government need to take steps to ensure that there's more competition. Here's where I've changed my mind a little bit based on what I see as the data. To me, in the AI space, it looks like there's an enormous amount of competition. The lead company has changed multiple times. Some of them are brand new startups. Some of them are incumbents. Some of them are DeepSeek using relatively little in the way of chips. Others are open AI using enormous amounts in the way of it. And so I think there might be less role for antitrust in this space than I would have expected a couple of years ago.