This is the 200th episode of the Science of Politics podcast, featuring economist Tyler Cowen.
Politics seems to be holding us back in a world of technological and social progress. Research has found health cures, invented magic new tools, and connected us all, often with public policy assistance. Yet, the American political system remains deeply divided and dysfunctional, with the relationship between science and government at a low point. Can we use social science not just to improve policy choices, but also to improve the functioning of the political system?
Cowen—an influential researcher, blogger, podcaster, and author—has led the Progress Studies movement, which seeks to understand why progress happens and how to accelerate it. The movement has gained institutional support and stimulated new policy ideas to improve living standards and human flourishing. But it has not yet cracked the code on translating these ideas into political success. How can science can be deployed to improve the American political process, and how much does the Progress Studies movement depend on successful politics?
Video
Transcript
Matt Grossmann: Welcome to a live edition of the Science of Politics from the Niskanen Center. I’m your host and political science professor, Matt Grossmann. Today, we’re celebrating 200 episodes of the podcast and I’m pleased to be joined by economist Tyler Cowen, one of the foremost public intellectuals. And we’ll be talking about the role of politics in progress studies and the relative importance of American political functioning to the future of human progress. Those who are joining us live can ask questions in the Q&A box, and I will get to as many as we can. So Tyler, you sparked this movement for progress studies. Tell us how it’s going. What are the most promising results in terms of ideas and institutional development so far and any disappointments?
Tyler Cowen: Well, we’re doing this podcast six years later, so that’s a huge plus. In a way, the rest is collateral information. The United Kingdom, the Netherlands, soon to be the EU either have or will have centers in something like progress or progress studies. There’s talk of it being done in other places. There’s a lot of people writing substacks with that kind of emphasis. So it’s gone much better than I would’ve thought. But that said, I don’t know that the actual world is going better. It definitely is on the tech side, but I think it’s going much worse on the politics side. So that’s the bargain we have right now.
Matt Grossmann: So how central is politics to progress? I know that’s a vague question, but it certainly seems that we’re in a time when we have lots of advancement, but politics is holding us back. So how important is politics, especially in rich countries and the United States to the success of progress?
Tyler Cowen: Well, very important obviously, but here’s a default hypothesis. I’m not saying it’s true, I’m just saying it should be our starting point. We should ask, is this true? That whenever you start to have a lot of progress, it disrupts politics you used to have. Your new politics is pretty dysfunctional and then you’ve got to fight your way through that. So it’s possible we have some kinds of progress to blame for where we are now. Now, we need to fix our politics to get to the next stages for progress itself. But I don’t think we should just paint the story of, well, it’s politics against progress because political institutions evolve slowly. Tech very often does not evolve slowly and there’s going to be this discord.
Matt Grossmann: And what about politics in terms of the sort of average country in the world versus politics of key countries like the United States? Is it possible to have political progress in most of the world where we see big declines in poverty and increases in human achievement while the key central countries in the world are not moving in that direction?
Tyler Cowen: That is possible. But the United States aside, I see retrogression in a lot of other countries. Turkey would be one example. Ecuador and Peru would be others. Chile, it’s not so sure. Mexico, it’s not so sure, but not obviously positive. You could say Argentina, there’s at least much greater hope. Not yet a proven success. But if you add up the positives and negatives, the balance has been the most negative since the decline of communism in my opinion. And that’s the whole globe. So that’s greatly concerning. And it may be due to common forces also, which again gets us back to the tech point.
Matt Grossmann: And what is the theory of politics behind progress studies? I certainly notice a lot of emphasis on policy and how to adopt useful policy, but it seems like the implicit theory is improved technocracy. That is we’re going to find the right solutions in science and research and then we’re just going to convince political leaders to adopt that. How realistic is that theory of politics and do we need a different one?
Tyler Cowen: Well, I don’t think there’s any theory of politics behind progress studies. I’m not even sure anyone progress studies or not has a decent theory of politics. To the extent I have a theory, it would be start with your own situation, ask what you can do best and then try doing a bit of that. It’s more a theory about oneself than a very large macro theory. So maybe it’s all about marching in the street or something, but I’m not especially good at that. So communicating ideas is something where I have a lot of experience. That’s what I do, talking to elites, but I’m far from convinced that’s like the right thing to do. I don’t know that there is such an approach, but you have what you have and me marching in the streets, I’m pretty sure really is a waste time.
Matt Grossmann: But is there a technocratic theory behind that view that means that perhaps is a block for its success? In other words, do you have to convince a broader group of people somehow or figure out how to collectively represent us better to get human progress rather than just find out the policies that would work best?
Tyler Cowen: Well, I think even very recently there are quite a few successes. It’s just the ways in which we’ve gone backwards are very discouraging. So the IMBY movement has also sorts of local successes. There are dozens of these around the country. We have not regulated AI to death. That’s still a story in progress, but it could have happened already and we haven’t. That’s a big success. What you would call broadly the space program seems to be going well, and that has a lot to do with policy. So there’s really a lot of policy areas. Nuclear energy is at least being contemplated again. It will be done in some countries. Huge success, which I had not expected. So that’s all great. I think there’ll be a lot more successes. But again, my worry is the failures might outrace them. So I feel we are actually convincing a lot of elites to do these things and they’re happening. The question is, can other issues do you in nonetheless like say the tariffs or policy on immigration?
Matt Grossmann: But one of our problems is that there are no sort of remaining trusted nonpartisan knowledge building institutions that are trusted by both sides. So how much is that an obstacle to the vision and is it possible for policy to be guided by evidence when the evidence builders are disproportionately quite disproportionately on one side of the political debate and trusted only by that side?
Tyler Cowen: There’s a lot of different questions and presumptions in there. First, I’m not sure there are no trusted authorities left. The US military for better or worse I think it’s pretty high trust rating still. I think the Supreme Court with some exceptions, but it’s mostly viewed as quite legitimate. The New York Times is not what it used to be, but people change their minds on things and it may not operate through a centralized authority, but say most of America deciding it wants to reject DEI, whatever your view of that is, there’s a new consensus. And it happened pretty quickly. You wouldn’t say, “Well, Walter Cronkite got on the Evening News and told us no DEI.” But it still happened. And what is mostly a good way. So I would partly challenge the premise. Again, it’s not all going in a good way. So say opinions toward vaccines, they’re far, far worse than 10 years ago. I don’t know how to change that.
Matt Grossmann: But you’re building, as you said, these institutions that are engaged in knowledge building, trying to convince government. And we’re on a podcast from a think tank that has the same basic theory. So you don’t think that there’s an issue when the people who build knowledge are seen as being on one side of a two-sided political system?
Tyler Cowen: I suspect that is not the core issue. I tend to think the core issue is some kind of negative emotional contagion. Some of that is due to events like the financial crisis, COVID, even 9-11. And then the negative emotion spreads through society in a manner that would be well described by sociologists. And that makes us more negative and bitter and cynical and we turn against many things. And yes, that does include the elites, but if you simply frame it as some kind of rebellion against the elites, I feel that the whole thing is probably being misunderstood. And having people have more positive attitudes, again, lower rates of mental illness and so on, that strikes me as more fundamental. I’m not saying I know how to do that, it’s quite clear I don’t.
Matt Grossmann: We had a lot of recent experience with some of these issues in the COVID experience where the role of experts in policymaking was quite strong but quite questioned. And lots of the after-action reports have been quite negative about how the experts themselves performed, how the relationship between government and experts worked. What should we be learning from that experience?
Tyler Cowen: Many people are learning the wrong lessons. So the experts, whoever you think they are, they clearly made some big mistakes. Suppressing lab leak would be one of them. A lot of what Fauci said I thought was really quite weak. When he told people in March 2020, go ahead on your cruise. That’s not the mistake he’s usually criticized for. But that was just stupid. But that said, if you look in the pandemic, what actually saved lives was plenty of evidence that it was vaccines and we had Operation Warp Speed, we got the vaccine, we got it out in record speed and our performance like far exceeded even reasonable optimistic expectations. So to say the experts just blew everything, well, the one thing that was by far the most important, they close to completely got right.
Matt Grossmann: But by the time they got it right, we were polarized on the role of experts and it made a difference in actual vaccine take-up. So-
Tyler Cowen: Sure. A lot of people died who didn’t need to die. Yeah.
Matt Grossmann: Okay. So-
Tyler Cowen: The real lesson of COVID is the experts did better than we think.
Matt Grossmann: The experts did better than we think.
Tyler Cowen: Yeah.
Matt Grossmann: But it seems like that’s not the lesson that people are taking. So if we had a similar circumstance in the future, we might have an even worse transmission mechanism from experts to governance in the public.
Tyler Cowen: That’s because of this negative emotional contagion. So another mistake we made, some states at least, was keeping the schools closed for too long. I’m not sure it was the experts behind that. I think a lot of it was small C conservative school boards and teachers unions. They’re not the experts. The experts were split admittedly. But the main problem was something other than the experts and the experts take the blame. So we’re looking for reasons to be negative about many things. That’s my worry.
Matt Grossmann: So one of the current bright spots that you mentioned in the YIMBY movement is this idea of an abundance agenda. How closely is that tied to progress studies and what do you see its prospects as?
Tyler Cowen: Well, I don’t speak for anyone other than myself. I read Ezra Klein and Derek Thompson’s book. I think I agree either with all of it or close to all of it. So that’s great. And I hope the Democratic Party takes that direction. I’m worried that they won’t, but I’m a big supporter.
Matt Grossmann: And how would you compare it to say the 1970s left/right convergence on deregulation that enabled some of these major reforms? Are there intellectual trends on the left and right that you see as converging usefully?
Tyler Cowen: At the state and local level, we’ve seen that with the YIMBY. I wouldn’t say the left and the right. I would say the progressive left and the libertarians and the other side just not fighting that hard yet in some of the cases. I’m not sure either the left or the right is actually pro-YIMBY. And my worry is those victories decrease to an asymptote and stop coming at some point. But on nuclear power, I wouldn’t call it a consensus, but you’ll find people from a lot of different political backgrounds supporting that one. So again, there’s plenty of signs of hope and there was never a real consensus on deregulation in the ’70s. Like Reagan actually didn’t deregulate when he had the chance. You could say he did indirectly on antitrust, but he didn’t get rid of many regulations. Mostly that was Carter or starting even with Ford. And the Democrats never wanted to deregulate many, many things at that time. So these things are always partial and piecemeal and you grab what you can.
Matt Grossmann: So it just doesn’t sound like much of the work that you’ve at least simulated so far in progress. Studies has used kind of political success stories, but some of the examples that you use for progress must have depended on getting the politics right. What have you learned so far?
Tyler Cowen: You need to be flexible and pragmatic and look at things through the lens of context and history and not convince yourself that you know more than you really do. I think those are great lessons for many things, but especially for politics, we have pretty good economic models, supply and demand and different modifications. We don’t really have great political models. So you look for talented people you can work with on particular things and hope you can get somewhere rather than nowhere. And again, I think it’s gone pretty well in a whole bunch of cases. So I don’t feel like, “Oh, we need to throw out what we’re doing.” We need to do more of it, do it better. But I’ve been very encouraged.
Matt Grossmann: So as a reminder to those joining us live, we are taking questions and you just have to type it into the Q&A box and we should be able to get to it. So one seeming success of public choice, at least in elite consciousness is making it clear that public officials and institutions also have their own interests that they pursue, but that has been associated with attacks on the state or the government, both from the left and right that are gaining new attention. That is that the state building might have been imperiled by people’s realization that we need to check those people who are in charge in government. Do we need to rebuild any kind of myths of governance in order to rebuild state capacity?
Tyler Cowen: We probably do, but right now, the problem is the reality more than anything. So we have at the executive branch, very cynical methods of governance, low levels of expertise, not in all places, but in many places, a lot of deliberate misinformation. And I’m all for nice myths, but right now we’ve just got to face that reality and hope it gets better.
Matt Grossmann: So there’s a commonly cited Andrew Breitbart view that politics is downstream from culture. And I know a lot of your work has been in the cultural realm. Should we view politics as really about changing culture first or is this stimulating a productive idea that politics should directly engage in changing culture?
Tyler Cowen: Maybe I think that’s also a too macro a level. So I wouldn’t contradict the claim that politics is downstream of culture, but also culture is downstream of politics. So our culture has changed a lot in the last months because our politics has changed. So it goes both ways. And I think the overall model ruling this is too complicated for us to usefully figure out in a way where we would find levers. And again, just to look for particular cases where you think you can do something and act on those pretty quickly and decisively and with good arguments, I think that’s the best thing we can try. So it’s all this big sprawling evolving mess. I mean even much smaller countries than the US but certainly America. And you’re never going to wrap your arms around the whole elephant. You’re never going to get most of what you want. But if you work hard at it every day, you’re not focused on getting the credit or getting power, you can get a lot more maybe than what other people think you can.
Matt Grossmann: When I first heard the ideas about progress studies, I thought about it as potentially an analogous to positive psychology, a movement within psychology that said we should focus on not just what makes people unhealthy, but what makes them more healthy. And certainly in political science, we are often reasonably accused of explaining why things don’t work rather than explaining how they can work. But if you think about that psychology example, it also brings to mind that a lot of those parts of psychology that we’re most involved in making those kinds of interventions had a lot of questionable research practices. They’ve been kind of central to the replication crisis and a sort of search for quick fixes. So is there anything that economics, political science, and sociology should learn from that or from its successes and its errors?
Tyler Cowen: Well, I think the main thing we need to learn in a lot of endeavors is everything we do has secondary consequences. They’re hard to trace out. There are things where you may think it’s a victory and it ends up looking like a loss. So if you told me the ardent supporters of President Obama that when Obama beat Mitt Romney, he will come to regret this, their jaws would’ve dropped and they wouldn’t have believed you. Maybe to this day, they won’t face up to that question. I’m not saying Romney was better than Obama, worse than Obama, but still, if Romney had won, I don’t think we would be in our current situation. It’d just be some very different world. And a lot of the original Obama supporters I think would be a lot happier. So that’s a cautionary tale. But again, at the end of the day, you’ve got to pick up your bags and do what you think is best, but just always be open to data and realize some of your victories might themselves give rise to bigger picture losses.
Matt Grossmann: But some of the positive psychology movement and associated with more the policy side of helped develop these nudge units and governments that people came to believe were promoting get-rich-quick schemes, too-easy policy changes with communications that would make big differences. And some of those are now closing. So is there, I guess, a danger that looking for the little places where we can make a difference corrupts the scholarship?
Tyler Cowen: I don’t know. It’s maybe too nonspecific a query to really answer. I think a lot of what government has done with nudges has worked great. So there were things done with vaccines. Well, they made them easier to get for a lot of people. It got people more vaccines more quickly. So it’s not all some big failure. In any movement, many things are oversold. Maybe that’s in the nature of movements that the oversellers will get a lot of the attention and then eventually there’ll be a backlash. Maybe some of that you just have to accept as natural. You could say some of that was true with asset privatization in the ’90s for instance. A lot of that was oversold or maybe there was a small gain. People expected like their water utility or railroads would just be perfect and sort of cheap like toothpicks, and just some sort of balance is needed.
So I think if you just take time to read about these histories or maybe you lived through them, you have a better chance of making the right decisions, I think is all I can say. There’s no promise of a path where you avoid all of these pitfalls, but you just much rather have some people around who know a good deal of the history. That’s your best chance, I think.
Matt Grossmann: So political science has been interested in these recent trends that you mentioned at the start, and it’s stimulated a lot of literature on democratic backsliding. What do you think of that research so far, and particularly in the US case, to what extent can we usefully interpret the Trump trends in the US, especially the people who are most alarmed by them in the scholarly community through comparative patterns or views of democratic backsliding?
Tyler Cowen: I don’t view it as democratic backsliding at all. I view it as very bad and I view it as the United States returning to a lot of its earlier illiberal traditions, something like Palmer Raids. But a lot of what the Trump people are doing is trying to run roughshod over the judiciary, which at least in a literal sense is the non-democratic branch of government. In some ways, it’s more democratic by some definitions than what we had. I just think that’s bad. So I’m not worried about democracy per se, but I also think simply keeping democracy per se is quite overrated. It’s always having a well-functioning balance of powers where many of those powers have strong non-democratic elements is what’s important. And the democratic backsliding framing to me induces people systematically to misunderstand that point.
Matt Grossmann: Well, how well did we interpret the first Trump administration and what should we learn from that for interpreting the second?
Tyler Cowen: Well, so far, they’re very different, but how much personnel matters seems to be an even bigger lesson than we had thought, because the first time, Trump wasn’t ready. He appointed a lot of normal Republican appointees. They thwarted many of his worst tendencies. This time around, there’s four years to prepare also and plan and resentments are building. Only one other case in American history where a president hasn’t interrupted second term like that, I think, and I guess that’s dangerous, especially when it’s Trump, but it might be not great more generally to have this interruption before a second term. There’s somehow too much planning that goes on by one branch of the government but not the others.
Matt Grossmann: Well, one of those planning prospects was taken up in think tanks and usefully combined with some ideas that Elon Musk had to really make pretty fast a retrenchment in government where others had failed. But it’s definitely not over, and we may end up just sort of paying a lot of people not to work or paying a lot of contracts that don’t end up getting done. And obviously, we’re cutting and stopping and pausing in areas that might hurt us down the line. So how should we assess how that’s gone so far?
Tyler Cowen: Well, some of those are very positive developments from my point of view, and some of those are negative. So I wouldn’t say cut funding for the NIH. I would not cut staff at the FDA. Those I think are big mistakes. But can you cut half the people at the Department of Education or Social Security or Health and Human Services? I just think there’s a lot of waste you can clear out that even with bad rushed methods, I think you can do a lot of good and some of that has been done. Because DOGE has some bad parts or more broadly, the Trump administration has some huge bad parts, just DOGE overall gets a bad name. But if you look at it carefully, do you actually defend the current level of employment at the Department of Education? I think it’s quite clear DOGE is in the right and the value of sending people a message that half of this can go away and three years later our kids will be just as well educated, I just think that’s very, very high and all the rush to criticize Trump, most of which I agree with, that lesson is getting lost.
Matt Grossmann: So yeah. What success stories do you think there will be from this initiative and where will be the ones that bite most in the long term?
Tyler Cowen: Well, I mentioned a few cases where I thought they were making mistakes. Most of the other cases I think are good. Now, I think it has not much more left to run. I think Elon will retreat from very close ties to the Trump administration, and as DOGE loses the attention of Trump and as Congress becomes more important as we actually have to pass a budget, I think we’ve seen more than two-thirds of DOGE so far. A lot of the cuts at USAID think we’re great. Cutting back on vaccines was a big mistake. The interruptions to PEPFAR were a mistake, but a lot of it is clearly necessary in my opinion and desirable. There’s so much of USAID that just went to pay third party contractors in part because the law insisted on it. So it was money paid to American companies who did become parasites, the nonprofit sector parasites and not really help say Africa very much. And just a simple rule, you find a few things in Africa that are very good. You send them money, you don’t involve too many intermediaries. That’s where we ought to end up. We did make some mistakes, but again, a lot of what we did I think was very good.
Matt Grossmann: So we have one question from a viewer about whether we should teach more economic history in high school. That might be too easy. So I’ll say, what should we teach more of and where are we getting it wrong?
Tyler Cowen: Well, how to use advanced AI programs. They can teach you better economic history than even PhD economists, much less high school teachers. So we need to radically reshape how we do all of education. This will be slow and painful, but we have the tech to do it right now. It’s not some hypothetical. Very few even PhD economists are qualified to teach economic history. Never mind these high school teachers.
Matt Grossmann: So you’re comfortable with whatever the aggregate lessons are that come out from the AI tools? You just think it’s just a question of dissemination?
Tyler Cowen: Well, I think they’re more objective than any media source I know of. Even something like Grok to me is really quite objective. It will tell you bad things about Elon Musk if you ask it. Students learn how to use them pretty quickly. Some might need to be somehow in more user-friendly form because the LLMs do not themselves motivate. You might need humans to motivate children to use the LLM, but it’s this incredible breakthrough. And right now, we’re kind of just letting it sit.
Matt Grossmann: It seems like there’s lessons from economic history that you want to be central to progress studies. So if the LLMs are already there, then is maybe knowledge creation is not the issue when it comes to understanding progress?
Tyler Cowen: Well, what people ask them is very important, but if you ask the best LLMs questions about history and progress and economic growth, their answers are very, very good. You can always have your own preference for a slightly different answer, but I have to say overall I couldn’t wish for better answers. I just think it’s a godsend and a miracle. And again, we’re doing very little with it, but at some margins, that educational problem is being solved, has been solved.
Matt Grossmann: So as we get AI advancing, there’s seems to be some presumption that that will somehow spill over into the political system. I was just at an event of scenario planning where there are all sorts of AI futures involved, but a lot of the assumption was that the political process would change. I am comfortable viewing outdated institutions as much more stable than people might think. So how inevitable is political change or institutional change with social and economic upheaval that might be caused by AI?
Tyler Cowen: I think radical change is inevitable. I don’t feel I have very good predictions for what exactly will happen. First order effect is probably to increase the relative influence of the United States and China, that other countries will hook into one of those two AI systems and that will determine a lot of other things they do. That to me feels like the soundest prediction I have, but many, many other changes will come. It’s hard to even imagine them all. We might be able to balance the budget. That’s a shock. I wasn’t expecting that. If AI boosts our economic growth rate by half a percentage point, which is not a guarantee, but I find plausible, there’s some kind of smooth glide not into full balance of the budget, but we avoid massive tax hike, spending cuts, and that’s a big change in our politics. Right?
Matt Grossmann: So you mentioned that so far it’s a success story that there’s sort of been hands-off approach to AI. How dependent is AI progress on policy choices or politics? There’s certainly a lot of folks in that world who seem to be acting like it is the most important political problem of the moment, but there might also be a case that it succeeds in spite of or because of lack of political attention.
Tyler Cowen: Well, there are all these state-level AI bills. I think there’s 4,000 of them. I’ve read summaries of some of them. And at some point, they’re going to pass in a meaningful way and I think we will need federal preemption and over time treating AI as a national security issue. Of course, there is no such single thing as AI nor a coherent legal definition of AI. AI you could say includes drones, not just large language models. We’re going to need better drone defense. We’re going to need to worry about truly building quality drones inside the United States without reliance on Chinese parts. That will require our government to do some things it’s probably not very good at doing. I think that’s a big problem for anyone. I don’t think there’s a simple fix with a tariff for subsidy here and there and then it happens. I think we genuinely don’t know how to do it. Those are some of our biggest challenges and they’re coming pretty quickly.
Matt Grossmann: How should we think about the role of economics in policymaking, especially at a time when we’re enacting the kinds of policies that there’s almost universal consensus among economists are not going to work in the form of these international tariffs?
Tyler Cowen: Tell me again the earlier part of the question.
Matt Grossmann: Well, economics can have a strong influence on policy and presumably that’s important for the success of progress studies we’re living.
Tyler Cowen: Yeah, they’re listening to Oren Cass, right? Not to economists. So I don’t know how to change that. I think Trump is very ornery in his views on trade. The state’s back to the 1980s. My hope is that business interest groups acting through the House of Representatives will force Trump to cut a deal with bigger questions involved and that will limit the tariff hikes. I think that may plausibly happen, and that’s I think my best hope rather than Trump wakes up some morning and starts listening to economists.
Matt Grossmann: Even Cass does not seem to be on board with the way that these are implemented. So that raises the question of whether even if you do get influence for some of these tools, the way they’re implemented politically may not be in line with getting the results you want.
Tyler Cowen:
And this is a very unusual administration in terms of how it’s run. I think often there is no plan, no center of gravity. There’s may be eight or nine different views. They’re all tugging on Trump’s cape for his attention. You end up with these very weird forms of cycling with super, super short time horizons. I think that’s a better way to understand the tariffs than to think, well, there’s some big theory behind it all. Maybe there are eight or nine theories, but no one of them is ruling the roost.
Matt Grossmann: So for viewers, we do have time for additional questions. I want to ask a little bit about the role of the public intellectual. We have an age of media fragmentation, maybe intellectual fragmentation and public intellectuals sometimes are just like social media influencers. On the other hand, there are people who still see this as a kind of a golden age for public scholarship, a lot of opportunity to get it out there. How do you see the state of the public intellectual?
Tyler Cowen: Well, it’s been good, but it’s all going to change in less than a year. I mean, even current AI, the best models, they can write a lot of pieces better than you or I could. The world just hasn’t woken up to this yet. I’m not even sure you know that. Most people I speak to don’t. I was at an event with Danny Roderick not long ago. He didn’t seem to know it. So you asked, how much will it change government? It will change our sector enormously. And that in turn will influence government. It’s such a big change that will come so quickly. I don’t know how to predict it, but people like the two of us we’re really, like I’m ready for it. But other people I meet, they’re not ready for it. They don’t even believe it’s true. And then there’s people who will recite the fact that it’s true, but they have not emotionally internalized that fact because they find it too paralyzing or demoralizing. So that’s what I would say about public intellectuals. We’re in for huge changes very soon.
Matt Grossmann: Well, I used the new opening AI models yesterday and this morning, and I know you’ve been impressed with them. So we will see where that goes. So tell us about your own journey, though. You’re switching to a new place for your own writing, and you’ve obviously worked at the New York Times and other places in the top sphere of the media influence system. Why are you making the move that you are now and how do you see the role of those kinds of parts of the ecosystem?
Tyler Cowen: Well, I’ve started writing columns for the Free Press, which is edited by Bari Weiss. I like the fact that it’s a startup. I’m very enthusiastic about charisma driven projects. I think they give their writers a lot of freedom to express themselves, and it is internet native in a way that many other ventures are not. And I thought I should try it. So far I’ve been very happy.
Matt Grossmann: And how do you see its role compared to the more traditional role of the op-ed columnists in the legacy institutions? It sounds like you think everyone’s going to be beat out by amateurs using AI, but maybe not.
Tyler Cowen: Well, they might be beat out by pros using AI. I think outlets where personalities have a more central role have a better chance of surviving those changes. And that’s definitely the free press. So if you’re publishing quite generic material, like if I read a generic op-ed in The Guardian, maybe by someone I haven’t heard of, AI is already much better. And I just don’t think that can survive. And maybe the Guardian was not surviving anyway. So there’s other outlets they get by on prestige. That will continue for quite some while, but maybe not really be very influential. That could already be the case. And I don’t know. It’s going to be such a change. I wish I had better predictions. I can say I’m going to do more podcasts. I’m already doing more. More public appearances. We all need to think more in terms of charisma and our personal presence, because on that, the AI is not even in the same ballpark.
But if you’re just writing somewhat generic material, I think you’re done. And generic, the benchmark is tougher to meet than you might think. So experiment and ask it to write 800 word pieces as Matt Grossmann would say on topic X, something you know a lot about, and just see how good you think it is and then plan accordingly. And o3 Pro will be out in a few weeks. That will be much better than o3. So if it’s out in a few weeks, it must exist already. They need to do things to get it ready to release. But my goodness, and what comes after that? We’re in a cycle where about every three to four months there are big noticeable improvements. And again, that’s something we’re just not used to. It does not typically happen with other technologies. So we’re asleep.
Matt Grossmann: So how important to progress studies is actually the creation of new knowledge through experiments or through a collection of new data. Because if it’s mostly just compiling lessons that are out there somewhere in the pre-existing literature, then maybe AI will be a big boon. Because-
Tyler Cowen: AI is original though. People don’t get that but it’s true. It’s already original. But over time, as the cost of tokens falls, you can just pose your AI an arbitrary number of questions and have some other AI grade the answers for both correctness and originality, and you’ll have a bunch of new hypotheses in a day and they’ll be worth investigating. So people who think it just regurgitates the internet, that’s a major fallacy. It’s one of the reasons why people are asleep on this. They think like, “Oh, we humans were original. The AI is just kind of quoting rewording Wikipedia.” That’s wrong.
Matt Grossmann: Well, I don’t think that, but most people in the research community still believe that original research in the form of new data collection and/or new experiments or research designs are necessary.
Tyler Cowen: Strong agree. But that’s what humans will do. They’ll specialize in that part of the process and they will not be these generalized scientists the way they had been. The AI will write the paper for you and do a lot of the thinking and do the stats work and do the programming. Soon. This isn’t 10 years away. This isn’t science fiction. We just need to have a bunch of things that already exist, get converted into products that in fact are used in labs and universities.
Matt Grossmann: So it sounds like the potential with both researchers and AI or augmented by AI is pretty strong for understanding the sources of progress. What are the fears that you would have in how this comes about?
Tyler Cowen: Well, some people will become, I hesitate to use the word stupider or lazier, but they’ll let AIs do their thinking for them. So if you’re ambitious as a thinker, you’ll be become much smarter. You’ll know more. You’ll have access to an almost infinitely large army of PhD level research assistants. And some people will do great things with that. Other people will just let the AI do their thinking for them. And some of the charges, Karl Marx levied against capitalism might actually come true for these people. It’s alienating and it makes you more passive, which I don’t think in general has been true up until this point. But we’ve never had such strong intelligences of a general nature in our machines. So that would be one of my worries.
The other is just the effect on politics, as we said. And again, it’s not that I have some doom-like prediction. I just think any big change, there’s a decent chance that can go wrong. I think on average it’s more likely to help democracy than hurt it because it’s pretty objective. But so many different things will be changing.
Matt Grossmann: And how about use in policymaking? Obviously, there’s lots of potential for more efficient solutions to problems. On the other hand, we just had an example possibly with a tariff policy developed via poor use of AI. So where do you see its use in governance?
Tyler Cowen: Well, there’ll be some subset of countries that just use it well and do much better. I don’t really have predictions as to exactly which ones. Wouldn’t shock me if Singapore are on the list and other places you might expect. But the places that don’t use it will fall behind relative to the places that do. And you’ll have poor countries. Say you’re Peru or somewhere in Africa, you’ll have the option to upgrade everything you do to world-class intellectual levels at an affordable price. But actually doing it will require some institutional will. And I think we’ll see a lot of divergence in which countries do it, how much, and how well. But say like you live in Kenya now. Kenya has good internet connections. You have the world’s best teacher at some price. I know it’s expensive for a Kenyan, but it will get cheaper. There’ll be group plans, it’ll be in schools. Imagine that Kenya has better medical diagnosis than the rich guy who goes to his human doctor in Beverly Hills. But that’s already the case. It’s not science fiction, that’s today. It’s the Kenyans using it.
Matt Grossmann: So you mentioned that one of your predictions is that most of the world will have to be in the US or Chinese sphere so you presume that this kind of confrontation or continuing dominance of the US and China is the most likely near term scenario. We talked a lot about our potential for understanding the US. What about our potential for understanding China? How well are researchers doing that and how likely is it that we’ll be able to have some better sense of how it operates or to actually improve its operations?
Tyler Cowen: I don’t think AI will help much with that. It depends on a relatively small number of people who keep their plans pretty secret. There may be ways in which our national intelligence services can learn more things, but the binding constraint there is not intelligence. It’s like trying to predict what Putin will do next. If you gain 30 IQ points, I don’t feel your prediction would really be better. You might avoid some basic mistakes. So that will just always be tough and I’m worried how many countries will prefer the Chinese AI over hours. To say you’re going to hook up your school system to one of the two AIs. Do you want American or Chinese product? Countries will have to choose. Now, some will be hybrid or joined. We might have open source is good enough that that’s a third option. It’s very complicated. But I still think at the end of the day, US and Chinese soft power will go up an enormous amount. And also national security systems, cyber defense will depend all the more on either a US or Chinese system that you buy, you rent, you procure.
And they’ll also be private sector because our government can’t build AI, right? So it’ll be an indirect mass privatization of huge parts of the economy. Not like true privatization in the sense of pure market competition, but private in the sense is done through government procurement and enacted through private companies so they will become much, much more important. And that’s another big change in politics coming soon.
Matt Grossmann: So I meant beyond AI though, just the progress studies movement, presumably understanding how we got so many people out of poverty in recent Chinese history is an important part of that potential for understanding progress, but also having some sense of how we get progress in a system like China.
Tyler Cowen: Well, Chinese people now have access to the truth about China through AI. Again, they may not do anything with that. I don’t expect their card system to crumble or anything. But for the first time, they can ask and have answered all the same things you and I might talk about without direct censorship. If it’s Chinese AI, you can’t ask it about Taiwan or Tiananmen Square. But if you ask it a question, how would you estimate the relative importance to Chinese growth of export promotion, migration to cities, privatization of some state run companies and five other factors? It gives you as good an answer as any Western scholar would.
Matt Grossmann: And the political system is at least as good at integrating that knowledge?
Tyler Cowen: We’ll see. Again, we don’t know. It’s possible the CCP will be so afraid of AI that it will turn its back on it and just end up permanently stupider. That would make life easy for us. It’s unfortunate for humanity. I hope they don’t do that. But it would make life very easy for the US national security establishment because we’d just be way smarter than they are forever.
Matt Grossmann: Is there anything we didn’t get to that you wanted to include or any questions you have for me?
Tyler Cowen: Well, what will you do next in terms of your own work and how does that stem from what you see going on in the world?
Matt Grossmann: Well, my next project is about how to make policymaking in a world of two party polarization and parity in the US. Certainly I’m already incorporating AI models and all the writing and research that I do and have experienced some of the same things that you do. But I guess I actually maybe am more convinced that AI can help because I actually think a lot of the knowledge actually is already out there and just hasn’t been effectively integrated from different spheres of scholarship so that’s something that… So when I make that point about AI, it’s not me saying it’s just Wikipedia. It’s maybe me saying it’s doing the most important thing that we haven’t been very good at lately, which is to integrate findings from lots of different areas.
Tyler Cowen: Yeah, I agree with that. Rule by Wikipedia would be pretty awesome on some issues right now. And in terms of your own career, what changes are you making in what you do to prepare for AI?
Matt Grossmann: I wouldn’t say I’ve made a ton of changes beyond how I conduct research and literature reviews and writing. But I do wonder maybe you have a better answer what to tell young people who are coming into the field. Obviously, we can tell them that they should become very fluent in the latest AI models. But if you’re seeing all this change coming, especially to our profession, what do you tell the new entrants?
Tyler Cowen: I tell them to stay flexible and be suspicious of anyone who thinks they know how it’s all going to go, basically. And to be fluent with the models, of course.
Matt Grossmann: Do you view the youngest generation of scholars as better at this than us and more prepared?
Tyler Cowen: No, I think they’re worse in some ways. Depends what you mean by younger. If you mean 14 year olds, they might be in great shape, but the people who right now say are 31 and in the second or third year of a new assistant professorship, they’re the most vulnerable because they’ve invested heavily in something that could just easily not match the human comparative advantage. And they don’t have the broader set of skills that older people might have who kind of learned everything before the internet, right? Those people in a way may do much better than the people who are 31 and who are just specialists in a skill that will be obsolete for human beings.
Matt Grossmann: So we’ve had lots of obviously difficulties in the research process. I mentioned the replication crisis and all kinds of questionable research practices, but there’s also just lots of errors in working from data generation to conclusions. Is that something that AI will just be inherently better at or will it replicate some of the same problems that we’ve had in human research?
Tyler Cowen: Well, pretty soon it will ruin a lot of reputations. That will be significant. It won’t affect the world, but it will affect people we know. It’ll go back and read everyone’s papers, find all this plagiarism, find things that don’t replicate, find inconsistencies. And again, this is not far away. I don’t think most papers matter. So the real question will be, what are the few new things AIs or whatever we have we’ll come up with? And I think the big advances other than AI itself will be in the biomedical sciences. Social sciences will not be that exciting, but will actually do things like fix most cancers. And most people will die of old age or accidents, certainly like your children, and that will be amazing.
Matt Grossmann: And so is that just the inherent nature of those problems are actually easier to solve than the social problems or is that something about the way that social research is conducted that means that it will be less able as researchers to take advantage of the AI skills?
Tyler Cowen: I think the latter. Bio problems are genuinely very, very tough to make progress on. And there’s a lot of literally moving parts, but doing things like simulating the cell so you can dispense with a lot of experimentation will be very, very important. We’re on the verge of being able to do that, just trying out new biomedical ideas at a super rapid clip. We’re starting to do that now. Also, material science for reason similar to bio will advance at the super rapid rate and we’ll get a lot more new useful materials, a lot more new ideas for drugs. We hope the regulators don’t hold them up. The FDA might be overwhelmed, but I think it’s all looking phenomenally optimistic. And the notion that 40 years from now we’ve beaten back almost everything that kills people, that’s my default scenario.
Matt Grossmann: And why shouldn’t we be able to make as much progress in economics and sociology and political science with the same tools?
Tyler Cowen: I don’t think the simulations help us that much. So you could simulate different macroeconomic policies. You might decide say that stabilizing nominal GDP on a growth path is a good idea. I just don’t think that gets you very far toward actually making it happen. Whereas if you come up with a cure for cancer, people will take it, right? That’s a big difference.
Matt Grossmann: So the problem is in the implementation or the buy-in of the political system, you think the ideas will be out?
Tyler Cowen: Just the individual humans, not everyone gets a vaccine. You can’t mainly blame politics for that, right? A lot of people just made the wrong decision. Some of its politics. So in economics, a lot of people just have bad ideas and it’s hard to get better consensus ideas through with or without AI. And then a lot of things like macroeconomics, I don’t think we’ll ever really solve it. It’s not how good the AI is, just there aren’t really clear answers maybe.
Matt Grossmann: Well, I promised you an hour, so we’ll end here. But maybe one last bit of good news forecasting for us because I know that progress studies has in its sights that we will make progress and we hear a lot of bad news. So what’s the optimistic case for human progress?
Tyler Cowen: We live in the most exciting time I have ever seen in my adult life or any life at age 63. That means a lot is at stake. There’s the chance we blow it completely, but there’s never been more opportunity to do human beings good than right now. And it’s by a lot, not just a little. So I hope we make a go of it and succeed.
Matt Grossmann: Well, thanks so much for joining me. Appreciate it.
Tyler Cowen: Thank you, Matt. Good to chat.
Matt Grossmann: And thanks to all of you. We’ll see you next time on the Science of Politics.