Go back

Alex Imas & Tyler Cowen | AI Is Moving Fast. Why Isn’t the Economy?

from The a16z Show

1h 4m 11s

Alex Imas & Tyler Cowen | AI Is Moving Fast. Why Isn’t the Economy?

This A16Z podcast episode features Google DeepMind's Alex Imas and economist Tyler Cowen discussing why rapidly advancing AI has not yet produced extraordinary economic growth. Both agree that model capabilities are accelerating dramatically, with agentic systems now performing real work rather than acting as fancy chatbots. However, they argue that benchmark hockey sticks do not automatically translate into economic takeoff. Cowen notes that faster progress often just means encountering real bottlenecks sooner, and that institutions, regulation, and physical infrastructure move slowly. He points to normal-looking market data and implausible revenue projections from AI labs as reasons for caution, while Imas emphasizes that the relevant competitive unit is the human organization, which benefits from comparative advantage and coordination that current agents lack. Imas's research suggests that even minimal AGI would not transform the economy; the real breakthrough would come when agents coordinate well enough to outperform human organizations, potentially breaking the Coasean ceiling on firm size. Such a breakthrough could cause abrupt labor-market disruption, though both guests expect a more gradual transition overall. They also discuss alignment through a mechanism-design lens, arguing that checks and balances matter more than perfect alignment, and that incidents like the Hugging Face episode stem largely from human error. On measurement, they see GDP as underrated but increasingly incomplete, with welfare-adjusted metrics, longevity effects, and structural models likely to gain importance. Both remain bullish on AI's long-term benefits but expect the path to be slower and more normal than many in the Bay Area predict.

Transcription

11510 Words, 63323 Characters

English
Speaker 1Institutions are not quite running as smoothly as either we might like or we might fear. We have these weird things happening with NVIDIA. If you look at prediction markets for what the anthropic IPO will go for, it's not flying above $2 trillion.
Speaker 2Do you expect the growth to continue or do you think it'll level out?
Speaker 1Benefits will exceed costs, but it will take some time, and the path will be quite a bit more normal than many people would have you believe.
Speaker 3What you actually need for real economic transformation is not just competence, it's the ability of these agents to coordinate with one another. There's this idea of agentic swarms. But if you unpack it, these agents are still very bad at coordinating. They're mostly just doing parallel processing.
Speaker 1Often the faster that some things move, that just means you hit the real bottlenecks more quickly. If you ask me the question, when will agents be 10% of GDP? I don't think it's soon.
Speaker 2What do the two different worlds look like? One in which GDP growth is a bit more normal and another in which it actually becomes double-digit?
Speaker 4AI models are getting better at an extraordinary rate. So why doesn't the economy look extraordinary yet? In this episode, I sit down with Google DeepMind's Alex Emos and economist Tyler Cowen to ask how rapidly advancing AI translates into economic growth. We unpack why better models don't automatically meet double-digit GDP growth. From institutional inertia and regulation to physical infrastructure, and finding productive uses for AI. Alex also argues that the bigger breakthrough may come when AI agents learn to coordinate well enough to outperform human organizations. And that raises the larger question. What would actually have to happen for AI progress to produce an economic takeoff? Alex, Tyler, welcome to A's in Z podcast.
Speaker 2Happy to be here. Thanks for having us. There's a lot to talk about. First, let's talk about some timely topics. The media report came out. There's been a lot of conversation. Paul Christian. I know his talk and tweets. There's a lot of sort of unrest and anxiety around sort of the state of AI right now and RSI. And I'm curious, Alex, what is your reaction to what we are seeing? And where do you stand on some of the proclamations being made right now?
Speaker 3So, I mean, I read it closely. I think it's undeniable that we're having incredible progress and the progress is accelerating. I think anybody who's looking at, you know, the model releases, they're becoming, you know, more frequent. They're becoming more powerful. The capabilities are going from, like, last November to this November, it's just an unrecognizable world, especially with the difference between, like, agentic AI and the chatbot era. You're seeing, like, what's coming out of companies when you have data coming out and, like, the difference between, you know, the chatbot era versus the agentic era. You're seeing real productivity now. Like, with the chatbot, you're kind of, like, using it almost like a fancy version of a Wikipedia or something like that. You're asking questions. It gives you answers with an agentic system. You're seeing actual work. And that level of work, the ability of these systems to actually do the work is just progressing so quickly. And you're seeing a lot of hockey sticks, basically. And that's what the meter report showed is that, look, here's another hockey stick. So, to me, to tell you the truth, I didn't update too much from the meter report because a lot of the data in the meter report has been around in other spaces. So, the S1 statement, you know, we knew about some of the unreleased models. I didn't update. I didn't update too much, but that's because I was kind of already there.
Speaker 2Is your view that the hockey sticks that you're identifying, that they will just sort of continuously go up? At what point might they peter out or sort of how bullish are you from here on some of these hockey sticks and maintaining?
Speaker 3So, I mean, like, the thing that you said, some of these hockey sticks is, I think, the main term. It's like, where are the hockey sticks going to be? How are they connected? So, in the meter report, to me, like, the key phrase was on well-verified, you know, and well-specified tasks. The big question is, like, what about less? The less specified tasks, a lot of work is not well-specified. I'm not one of those people who thinks this is like, all right, we're safe on the non-well-specified side. I think the models are becoming increasingly better on not very well-specified tasks as well. But I think, you know, it might be harder to do hockey sticks on those sort of tasks. There's also a lot of, like, the questions of, like, tacit knowledge, kind of Hayekian concerns about, like, the actual impact on the economy. I think, you know, Tyler and I are both economists. So, all of this information, I have to kind of sift through about what these hockey sticks mean for economics and what does it mean for economically valuable tasks. And the big questions for me are, like, every model iteration, what are the economically valuable tasks that this model is unlocking that the previous model wasn't unlocking? And this is kind of captured very imperfectly by, get the hockey stick. So, you can get a hockey stick that looks crazy on a benchmark, and then you're not really seeing a lot of acceleration in the real world. We're not there yet. I think we're seeing a lot of acceleration in the real world. But these are just, you know, some of the questions that I have.
Speaker 2Tyler, let's go to you. I feel like you've been a more sober voice in the conversation. Alex has also been sober, but I just mean, have you updated at all based on what's been happening here? What is your view?
Speaker 1Well, often the faster that some things move, that just means you hit the real bottlenecks more quickly. So, I don't update that frequently. But this is AI land, and in AI land, the exact dates really matter. So, we're recording on November 11th, 2026, and that matters. I would say since August, I've changed my views a fair amount, most of all because of the math problems. So, progress in areas where there are not many human bottlenecks. I see much more rapid progress, and the ability to create a new model that is clearly better, say, within something like three weeks. I hadn't thought that would happen as quickly. As it had. So, I would date that to a few months ago. But if you mean over the last week, the meter report, the last two weeks, I'm actually slightly, very slightly more bearish. I have no prediction, but it will slow down the sector a bit. If you look at prediction markets for what the Anthropic IPO will go for, they bounce up and down. But it's not flying above $2 trillion. You look at Anthropic's own projected revenue numbers. I forget the exact numbers. I wouldn't say I was underwhelmed, but they're not numbers that you would expect under the super-fast takeoff scenarios, and that's for their IPO. So, I'd say I'm like 1% more bearish over the course of the last week than I was a week ago for those reasons. Just the institutions are not quite running as smoothly as either we might like or we might fear, depending on your point of view.
Speaker 2Are either the sort of growth rates of OpenAI and Anthropic… Is your sort of base case that over the last, let's say, you know, two or three years, that those growth rates continue? And put aside those specific companies, but I more mean just the leading AI companies, like the sort of economic impact that we've seen, you know, whether those two companies maintain the lead. Of course, they're competing with the hyperscalers, too. Do you expect the growth to continue, or do you think they're leveling up?
Speaker 1Well, I was in the Bay Area a few weeks ago, and many people were pushing very high revenue numbers on me. And I asked them all the following question. I said, what are the… What are the other sectors that are going to grow sufficiently rapidly to justify the extreme revenue numbers you're giving me for Anthropic and OpenAI? And no one had an answer for that question. They didn't even try to give me answers. So I think that means the extreme revenue numbers are false. That doesn't make me bearish on the sector as a whole. I'm still bullish. I think it will be fine. Those are great companies. But at the end of the day, you look at market data, whether it's real interest. It's real interest rates, stock prices, anything. Market data still look pretty normal. And the whole world investing community is talking about AI. It's not some kind of weird mystery like it was back in 2014, where you could say I read a less wrong blog post, and now I'm really wealthy because I did. So I trust the judgment of the market more than the judgment of the Bay Area people who don't think that carefully about social interactions. So that's where I stand. It'll change the whole world. Benefits will exceed costs, but it will take some time. And the path will be quite a bit more normal than many people would have you believe.
Speaker 3Yeah, and just to add to that, I think Tyler is exactly spot on in the fact that, like, you need to start seeing the productivity numbers in the rest of the economy to justify the revenue increases. So at some point, the rest of the economy needs to be willing to pay for that product. And that increasing willingness to pay has to be from increasing productivity in the rest of the economy. And we haven't. We haven't really seen too much evidence of that. Not to say that it won't happen, but it does need to happen at some point. The other part of that is competition. So I think what's starting to happen, particularly in the last few months, is extreme price competition from the frontier labs. As far as, like, before, they were just trying to compete on, like, who has the better model. And now that the models are being used by increasing number of enterprises, and they're becoming quite expensive because agentic AI, by the way, uses way more tokens. So now companies are, you know, OpenAI, Anthropic, you know, Muse is in the game, DeepMind is in the game, trying to actually say, like, no, we're offering a cheaper model that's as good or maybe even, like, a little bit better or maybe a little bit worse. And so you're seeing token prices come down a lot. And so, you know, Ramp has a really nice data set to track these things. And you saw exponential increases in token usage and flat revenue numbers, right? Where is that coming from? It's because of price competition. So at the frontier, you're seeing way more price competition than you used to, like, six months ago.
Speaker 1And the very recent collections, I think those are being overplayed. I think a lot of data centers. still will proceed, but I think it will slow things down a bit. There's greater likelihood of regulation or somewhat tougher regulation or regulation a bit sooner, or Trump, you know, giving on the issue of guardrails. To me, it's not a major force, but again, it's going to make me a modest bit more bearish on the sector in terms of speed of takeoff, revenue numbers, anything else. By the way,
Speaker 2while we're on the topic, did the midterms cause you to update anywhere else, Tyler, or is it kind of, you know, thermostatic backlash is a thing and, you know, we should just expect these pendulum swings often?
Speaker 1Well, most of it I expected and indeed predicted, but Texas truly surprised me. I didn't think Texas would flip the way it did, and maybe I just haven't been to the state enough lately, but that was a shock.
Speaker 2Yeah, yeah. So, AI and the economy, you both talked about how, you know, you push back strongly against the idea of, say, you know, double-digit GDP growth at some point due to you're far more sober that, hey, it will be tremendously impactful for consumers. It will change a lot of things, but GDP growth, as you said, you know, a few minutes ago, Tyler, it might be more normal than we might expect. But before getting why that is, can you explain what would need to be true for, or what do the two different worlds look like? One in which GDP growth is what we just said, a bit more normal, you know, and another in which it actually becomes more normal. And Alex, you were alluding to it a little bit, like, paint that world to me of what it looks like if it's actually having, you know, massive impact on the economy.
Speaker 3There's a beautiful paper out there that outlines exactly the conditions you need for this, like, singularity-type growth by Ben Jones, Chad Jones, I believe Aguilón is on that paper as well. And essentially, what you need is a bunch of pretty knife-edge conditions to be met, particularly the main one is essentially no real bottlenecks. And the main bottleneck that can't exist is the idea production function. So you can't get decreasing returns to ideas. So basically, you go from, like, the old world where you have this technological breakthrough, and then you kind of have decreasing returns to that technological breakthrough. From that technological breakthrough to start creating more and more technological breakthroughs. So you need less resources. Less resources while getting more and more technology and more and more production out of that technology. And so you basically have increasing returns rather than decreasing returns. And this is the, you can actually estimate this parameter in the model. And if it's greater than one, you're in this, like, double-digit growth world. If it's less than one, you're in the standard kind of, like, diminishing returns, single-digit growth world. And if you look at the data so far, including, kind of, the meter report, the data that's out there, and you take that data and you estimate this parameter, it is decidedly less than one. So this could change. And some people say it will change. And when you get the full RSI loop, this is exactly what it's going to change. But right now, you need increasingly more resources in order to eke out increasingly more intelligence out of these models. I would
Speaker 1add to that, you can look at cases where growth has been very high, so not long ago, Guyana grew 40% in a year. That's crazy. But it's a tiny place selling one thing oil. There's a brief period in the 80s where the U.S. is growing over 8%, but that's coming out of a deep recession. After pandemics, after major wars, you can have temporarily high rates of growth. So if the U.S. were to screw up in some way, or we had another pandemic, some kind of terrible war, some other weird event, and that combined with a burst of AI-driven growth that might have given us 3% to 4%, like, could we for a year get to 10%? I'll say yes. I'd bet heavily against it. And the effect would be temporary. And a lot of it would come from getting out of the recession, war, or pandemic. But that's another way to think about it, fully consistent with what Alex was saying.
Speaker 2And if robotics emerge, would you have a similar view? Because that is impacting more of the real world, the real economy?
Speaker 1Well, I think that would likely come relatively slowly. Look at how long it takes to get a Waymo in Washington, D.C. We're still not there. And that's a pretty simple kind of robot. There's like forwards, backwards, and turns. And if you look, well, what are the valuations of robotic companies at? Way higher than it had been. Hard to estimate. But you don't just see them popping into trillion dollar companies. I mean, you could count Tesla as a kind of robotics company. But the market data says it will happen. It's making good progress. But not going to give us 20% growth like three or four years from now.
Speaker 2No. Right. Right. And so the smartest people who are, you know, the biggest believers in sort of this runaway growth, you know, the Dorkeshs, the Leopolds, you know, are friends who believe that. What do you think they think differently? Is it that they have a different belief on the speed of technology? Is it that they believe that it will sort of overcome some of these bottlenecks? Or can we steal man the other side?
Speaker 3I think, and Tyler, please chime in. I think the argument, I'm going to get a bit technical here. It's about having something called homothetic or non-homothetic preferences. And basically, what that means is that if you look at kind of things that get automated in the past, what ends up happening is the thing that gets automated contributes less to growth. So let's say you have a million robots. Let's bring robotics. The thing that they're making, because the price becomes cheaper, the thing that they're making ironically becomes a smaller part of the economy. And so they contribute less to growth. And so if you have non-homothetic preferences, which is this, you know, there was a really nice paper estimating these preferences of data. What ends up happening is that, let's say you have broad automation, people just, people, the value in the economy of what contributes to growth moves to the bottlenecks. That's just an inconvenient part of how growth is measured. And so the thing that's growing least will be contributing most to GDP. And that's why you get these sort of effects.
Speaker 1I agree with all that, but I would add a more sociological element. In my case, and I suspect Alex's, what is emotionally vivid in our minds is the long slog of human history, previous growth episodes, centuries of GDP data, productivity data. What I think is emotionally most salient, probably for Dorkash, Leopold, and many others, is conversations they have in the Bay Area, which have real information. But when you ask the big question, you know, with which emotional salience will you make better predictions? I would opt for what Alex and I have stuffed our heads with, or for what they imbibe all the time. But of course, you know, they would take the other side of that.
Speaker 2Yeah. So basically what you're saying is when you look at the last few hundred years, the breakthrough technologies from electricity to sort of, you know, the internet to, you know, everything in between, they haven't led to runaway growth. Or if they did, it was because of sort of, you know, it was a small base or, you know, after a war or something like this.
Speaker 1I do think this will likely be quicker than those. But still, you have a sense of the bottlenecks, whether it's the industrial revolution, history of the scientific revolution, implementation of electricity, the IRS still uses fax machines, and so on and so on. Try using edtech software, which I do all the time. It's a nightmare. They're not even up to, you know, how things should have been 10 years ago. Yeah. And that's not. Yeah. That it's not. Regulation is not the problem. It's a problem. But just inertia is difficult to overcome. And so much of our economy, it's government, it's nonprofits, it can be highly regulated parts of the healthcare sector, higher ed, general schooling. When you actually have experience with how those work, you could say like, I don't know, 40% of the U.S. economy is a kind of nightmare. In Europe, it could be higher, like in France. And those sectors might need to be that way. For other reasons. But still, it's a big dose of realism that not everything is like startups.
Speaker 2Yeah. Let's say you were both tasked by the president to come up with a set of metrics that determine whether AI is, you know, leading to our country's flourishing, you know, broadly defined. Obviously, that's a very big question. But is GDP one of the main metrics you would think about? Like, should we be aspiring for it to have a. A big impact on GDP? Or what are some of the metrics you would consider?
Speaker 3Yeah, I mean, so this is a really important question. So I think the other thing that I would add before even starting is like, I don't think Tyler and I really disagree with like the San Francisco consensus on like the capabilities. Like, I think, you know, I don't. A lot of people hear Tyler and I or people like us talk and they're like, oh, you're bearish on AGI or something like that. That's absolutely not it. It's more about how that affects the economy and the structure of the real economy. And so thinking through like what measures we can be using, I think GDP historically has been a really good tracker of, you know, how basically human welfare and flourishing, it tracks pretty well with the gross domestic product. But what Eric Bernal and Avi Kolas and others have been kind of studying is the fact that technology has actually started to introduce this sort of wedge where a lot of the products that technology creates are basically free. And so they think about the sort of like digital entertainment and digital products. And then they have something called GDPB, which is, or GDPD, I forgot the acronym. I'm sorry about that. But it's essentially a, you take GDP and then you amend it with a kind of a survey-based measure to actually see how much other products that are not necessarily priced or products that have become cheap or even free because of automation, how much the increased amount of those products is adding to human welfare. Because you could see a world where you're still kind of like in this four to, you know, three to 5% GDP range, which is already on the high side of the growth. But you have a world that's actually truly flourishing. Like, how do you pick that up? And I think the way to pick it up is through these alternative measures.
Speaker 1And longevity effects could well prove important. So, Eric, say you live to 120. I don't think I'm going to make it. That will, in complicated ways, boost GDP and even more complicated ways, possibly boost per capita GDP. But it is something distinct from GDP that you get all those extra years. And economics can value that. And that's another way in which it may become more important to revise GDP. But, you know, as Alex said, GDP will become somewhat less accurate over time as AI progresses and we'll need to revise and improve it. But in the short run, it's just not that bad a measure. I like to say GDP, the most underrated statistic that there is.
Speaker 2Yeah. In the 2010s, Peter Thiel, you know, famously decried, sort of the lack of progress in atoms relative to bits. But here we are, you know, 10 years, 15 years later, we've got, you know, self-driving cars, you know, early flying cars. We've got, you know, major, you know, you know, the biggest companies are now hardware companies, you know, NVIDIA, SpaceX, Tesla. Tyler, would you go so far as to say that, you know, that isn't a concern as much anymore? Or how would you reflect on that?
Speaker 1The atoms will break through, I would say, almost for sure. But we still don't have them. Like the contribution of driverless cars to actual human welfare is tiny. Like the biggest contribution is that it's fun. And you can tell your friends I was in a Waymo. It's not really better than an Uber. For me, it's a little worse because they go by the speed limit. It's fun in a science fiction kind of way. Most of America still doesn't have it. Europe, forget about it. So we're still going to have it. It's still not there. That's the amazing thing. And it gets back to robotics also will be slow.
Speaker 2Yeah. I think I asked you 10 years ago, when would the great stagnation be over? And I think you said something along the lines of when we have self-driving cars. I can't remember if it was deployed or, you know, how sort of diffused it was. But you still believe that?
Speaker 1More or less. But that's a matter of degree. And they clearly work. And the new Tesla is pretty phenomenal. I've been in Alex Tabarrok's lately. Yeah. A few times. But in terms of market penetration, I don't know the numbers. But again, it's tiny.
Speaker 2Let's go back to a different topic. Alex recently wrote a related paper about it. But I'll start with you, Tyler. The agent swarm topic, sort of the hugging face incident. Did that update your views in terms of what's possible or make you get concerned about something or any reactions to what happened there?
Speaker 1It didn't update my views much at all. And I'm curious to see if Alex agrees with this perspective. But as an economist, you think first and foremost in terms of principal-agent theory. And unless the conditions are very restrictive, the principal just can't control the agent that well, no matter what contract they try to write. So when we replicate that result with AIs, well, it's interesting, but I'm not very shocked. The answer was always going to be checks and balances and what we might call separation of powers in the political realm. Not you have perfect alignment. So again, if you're going to do that, you're going to have to do it. So if you come at this from the economist's point of view, all the hand-wringing about alignment, like, yes, people should work on that a lot. But at the end of the day, you're like, hey, I learned that when I was 17. Like, OK, it's a real problem. But Alex, what's your thought on that?
Speaker 3I unfortunately did not learn that when I was 17. I think we once again agree. I think, you know, there's like these sort of different perspectives on this issue of. How do you best. Ultimately, it's a question about, like, how do you best get AI to be in the service of humanity? Right. That's that's the question. That's the question of alignment. And one kind of way of thinking about it is to get the models themselves aligned with like something like a constitution. The models at an individual level are aligned in this way because they have the same preferences as the human beings that are that are deploying that. But another perspective, which is what Tyler mentioned, is like you align things. You don't. You don't almost like don't even care what the preferences of these agents are. They could be extremely individually misaligned, but you can create a mechanism around them where the incentives are such that the agent behaves in an aligned way. And this is the mechanism design perspective in economics where, like, as a principle, I don't need the agent to want to help me necessarily. The agent could really, really want to just help themselves and do whatever they want. And the whole study of mechanism design is how do you create the system of checks and balances? In order for the agent to still be working in the principal's aid interest. And I think that's that is. I from the research that I've been seeing coming out of the labs, both like internally, externally, like a lot of people are thinking this direction that we really need to take a mechanism design perspective. And as Tyler said, like, I wasn't, you know, shocked by what happened.
Speaker 1I'm not aligned with the governor of Virginia, but it's a pretty well run state. And I'm happy here. And we have two party competition, sort of. And it works.
Speaker 2Do you think the way that Anthropic or OpenAI or some of these companies or just the sort of AI safety community thinks about alignment? Do you think that is sort of not the right way to think about it or confused in some sense?
Speaker 1Well, there's how they think about it and how they talk about it and how they talk about it is way too shaped by Twitter, which can be quite hysterical. How they actually think about it, I believe, is a lot more reasonable. How it ever looks from the outside. So I don't have some big speech about, oh, we've got to totally reform them. I think they're pretty practical. They're companies. They have IPOs coming up. They don't want to get sued too much. They don't want to be overregulated. We already have checks and balances. Not all of them good, to be clear. But the reality is the actual problems we've seen, you know, the hacks, they're mostly coming or maybe entirely coming from evil human individuals where the AIs are doing what the humans told it to. And that is a huge problem. But that is a kind of, well, we have the alignment. The evil human says do this and the AI does this. And that's where the costs are coming from.
Speaker 2Yes. Some people make the comparison that alignment is in some ways it's for the employees. It's sort of like safety in general is like in 2020, 2021 and sort of around that time, there was, you know, sort of more awareness around social issues. And so CEOs had to sort of. Say these say these things to please their employee base. And, you know, some of them were true believers and some of them weren't. And similarly, perhaps around the safety, there's a lot of concern. And so some people are making that that comparison that that the CEOs aren't as true believers as they may be indicating.
Speaker 1This checks and balances in these companies as well. It's partly why the companies work well, that you have to actually end up pushing through things that will work. And that's somewhat different than what everyone is saying. And there's a winnowing and selection. Set of mechanisms within the company. And people want promotions or bonuses and they want to do things that can work. And again, it will end up going better than how it will sound on Twitter, I would say.
Speaker 3Yeah. And I think like hugging face is like a perfect example of like if you unpack like what actually happened, it's just a bunch of human error on top of human error. Right. And so like it's great to have an incident. I mean, great in quotes, but it didn't like cause a massive disaster. Right. And it exposed a lot of like, you know, broken incentives for the agents. So it was, you know, you see how, you know, the companies have reacted to the hugging face by learning from it and creating better environments. And so this is nothing is smooth. It'll be kind of a little bit down, a little bit up. But, you know, the slope is positive.
Speaker 2The I mean, some people were, you know, Paul Cristiano, I think, said in his speech, he's like the most likely cause of death, you know, a year from now or two years from now will be an AI. Related death and that that to you sounds sort of, you know, hysterical. Strong disagree. Yeah. Yeah. Alex, let's go deeper on your paper. Why don't you share some of the findings that you. had and some of the questions that still ring.
Speaker 3Yeah, so like the paper, I believe you're talking about the Organizing Intelligence paper with Andrew Koh and my colleague Shane Legg. And basically, the idea of that paper is that it starts with this kind of puzzle that, you know, AI is becoming very capable on particular dimensions, but we don't still see any like visible economic transformation. And the first kind of thing that we start with is that there's been this focus on individual agentic benchmarks without kind of thinking about what those agents are actually competing with in the economy, where the thing that's the competitive, the relevant competitive unit is not the agent, it's the organization. And the organization in historically has been referred to as a form of super intelligence, because of this concept called Ricardian comparative advantage. So basically, if you have a bunch of jobs, the person who does the job is not the average person, it's the person who's best suited for it. So like the envelope of the intelligence in an organization is way better than if you kind of average across the population. And so these the AI agents are basically competing against forms of human organizations or teams of humans. And so you shouldn't compare a individual level benchmark to how well a human does something, it should be across the entire kind of what we call mosaic intelligence. So the thing that I think is really important is that you don't have to be a human to be able to do that. You don't have to be a human to be able to do that. So the thing that we so then we kind of start layering on like what sort of how do capability improvements will actually affect the economy in equilibrium. So this is a formal theoretical paper. And we look at different levels of intelligence. So we start with minimal AGI, where the AI models are as good as average human capability on every single dimension. So minimal AGI. And we show that even when you get to this point, you're not going to have a huge impact on the economy. Because again, because the organizations are not the average human ability, they're better than that. And so what you're going to start seeing is actually wages change. Because, you know, production becomes cheaper. And so that's going to affect human wages. And so what we show is that what you actually need for real economic transformation is not just competence, like ability on particular tasks to change. It's the ability of these agents to coordinate with one another. If you look at Hugging Face, and even like how the solutions to the Millennium Problems, there's this idea of agentic swarms, right? They're not coordinating. They just do, they're mostly just doing parallel processing. And if you have benchmarks, specifically on coordination, you could actually show that one agent does better than two agents splitting the work, even if they can communicate. Whereas if you give the same thing to a company, you're not going to be able to do the same thing. Whereas if you give the same thing to two people, they actually, they have division of labor, they communicate, and they do way better than one person with double the resources. So the agents are still pretty bad at coordinating, but if they get better at coordinating, you can basically break the Coasean ceiling on firm size, and you can get very, very different firms to emerge. So that's the basic premise, is that once coordination becomes better, the economic space of possibilities of what firms look like, what they're doing, what they're doing, what they're doing, what they're like, what happens to wages, what happens to employment, that's when the real kind of visible changes that are going to happen.
Speaker 2And what's the biggest bottleneck to figuring out coordination at the moment?
Speaker 3To be honest, I don't know. I think it could be as simple as it's not being trained into the models. It's not part of post-training. I actually don't have any insight on that.
Speaker 2Yeah. We've talked about some areas where you agree. I'm trying to find some areas where there's some difference of opinion. Maybe let's try employment. Alex, I think you're more worried about sort of the transition happening so fast that there might be something sudden happen on the employment side. Why don't you share your concerns or views?
Speaker 3I'm basically worried about the potential. And again, I actually agree quite a bit with Tyler that things are potentially going to move pretty slowly as far as the effect on the economy. Let's say there's a breakthrough in coordination. What we show in that paper is that if there's a breakthrough in coordination, there is the potential for wages and employment to suddenly there's like a phase transition as far as what will happen to the economy. In that case, there still will be new sectors and new potential opportunities. But the sort of rematching that we have in the economy that, you know, because of transitions used to take quite a while, rematching doesn't have an opportunity to happen. So you have these large, larger pools of unemployment. And that leads to kind of these recursive loops where people kind of get scarred out of the labor market. So that's my personal worry. But, you know, as far as that's a scenario that I'm worried about, likelihood is kind of a different story.
Speaker 1I have a related worry. It's close to the same thing, I think. But it's just hard to tell right now if you're hiring someone, will the person be good at AI? The skills they used to need, we had imperfect indicators. Oh, they went to Cornell or they got all A's or they had a good interview. And now the skill, will they be mentally flexible enough to pick up AI? You know, we don't know often. So that could increase the natural rate of unemployment. Good matches will be harder. Search will take longer. I don't think that problem will be with us forever. Eventually the AIs themselves will figure it out. And there'll be certifications for how good a person is in AI. But in the transition, everyone's a novice. You might be like well-established the way a lot of people in the tech world are, but otherwise we're all novices at this very important new skill.
Speaker 2And so I guess, what does that mean for, I guess, I'm curious how you think about sort of employment, wages, sort of these questions.
Speaker 1Well, some additional slowness in hiring people, but there'll be an incredible number of new jobs. So I'm pretty bullish actually. And labor markets have been fine. So, but of course there'll be disruptions and transition problems as there always are, as there were with the China shock, the industrial revolution. So I expect a new set of those, a bunch of them we haven't even figured out or seen yet, but they'll come.
Speaker 2Absolutely. To what extent will there be just sort of an explosion of agents sort of buying, transacting with other agents? In this kind of like almost parallel economy and, you know, there's so many more of them, they're 24-7, they can, like, how, to the extent that, you know, agents become the dominant sort of impact or sort of force on the internet and like, what is the right mental model of thinking about what that will look like or what that won't look like?
Speaker 1I don't know what you mean by the word explosion. Now, if they rose from basically zero to 6%, you could call that an explosion. 6% of GDP is enormous. Even that doesn't seem that soon. If it's like one and a half percent, I could see that, I don't know, within five years. Just if you're going to force me to give guesses, you could call one and a half percent an explosion. GDP is huge. That's one of the core insights of economics. GDP is huge.
Speaker 3Yeah. And I mean, like, even with the agents that we have now, they're already incredibly capable. The sort of consumer grade things that live on your phone. But, you know, that's the thing. Like, there's just completely unintuitive bottlenecks out there. Like, websites are designed for human beings and they're all different. And so, you know, you're still getting into issues with where your super smart agent is like booking you a flight, booking you a seat right next to the toilet every single time, every single time. And then you're like, I don't want to use an agent next time. Or the fact that there's all of this regulation. I mean, that's the big one, right? So like agents, companies are spending billions of dollars trying to do basic choice architecture on their website in order to upsell humans on products, right? They don't want to, agents don't have the same psychology. And so they're not going to be prone to these sort of, this sort of choice architecture. So these platforms have an incentive not to allow, you know, somebody else's agent to go in and shop for them. And you're already kind of seeing it. So like, if you pull up Muse or something like that, it can't shop on for you. I
Speaker 1don't use agents to manage my calendar, although they could, because I want to know mentally in my mind what's coming each day. And the only way I can remember is if I have to write it down an extra time on my calendar. And that's stupid. That's my flaw. But I don't think it's going away. And the agents could be 10 times better, whatever that means. And I'm still going to want to write it on my calendar. So I think, oh, December 18th, I'm seeing a Renaissance music concert. And I know that off the top of my head. So there are these funny indivisibilities in how we allocate our attention that are often some of the biggest bottlenecks.
Speaker 2Going back to the regulation part, Tal, you seem more sober that this sort of the midterms, you know, might not lead to sort of, you know, data centers stalling indefinitely. I'm curious just for your broader thoughts on AI sentiment, how much of the negativity is real in the sense that it will lead to, you know, we saw like Bloomington. You know, trying to outlaw usage of it as well to some degree. Like, what are your thoughts here on What's, how did you think about the political economy there?
Speaker 1Well, we have 50 states, there's counties, there's cities, there's even other countries. It seems that we'll slow down a bit, but still basically go through. You know, the new Texas governor is already backtracking. They're like, well, we just want the labor unions involved. They need to pay good wages. You know, the projects we have now will go ahead. The others we're going to look at again. Josh Shapiro, the rhetoric was very strong. The reality is quite different. So it raises costs, but I don't think ultimately it will matter that much. But will there be a bunch of places that absolutely say no? Probably, but again, 50 states, I'm glad we have them. Maybe we'll end up with 52.
Speaker 3Yeah, I mean, just to piggyback on that, I think like it just completely depends on the path, right? So it's like things, everything's kind of endogenous, including political economy. So, you know, the reason that. the governors are backtracking is because like unemployment rate looks great. The economy is growing at a reasonable amount. Everything kind of looks fine economically. So like there's no like instrumental reason for there to be this sort of political economy reaction. But the thing that I think people who are projecting these very extreme scenarios don't realize is that with the beginning of an extreme scenario, there will be the beginning of an extreme political economy reaction. And that will kind of endogenously constrain things. So it's just going to be a back and forth. And what the data will end up looking like to our and to, you know, people living in the future is that things moved a lot slower than people would think because of this sort of interplay.
Speaker 2It seems like, you know, the sort of amount of money one can put in politics, you know, everyone was worried about Citizens United a long time ago. And, you know, corporations are going to control everything. But, you know, Elon put in $500 million. Or whatever it was and couldn't get the outcome he wanted. Is this a sort of exposed limitations of money in politics? Or is it just that, you know, tech people, or not just tech people, corporations may be very good at allocating for their business, but maybe not, you know, that great at allocating in sort of the DCM political realm?
Speaker 1Dare I say that academia is slightly underrated by Bay Area people. People who study campaign finance have known this for a long time. And I think it's a good thing. I think it's a good thing. There's a lot of evidence that spending doesn't matter that much. You can think that's good or bad, but none of this surprised me in the slightest. And I don't think it's that the people are fools. You might think they're fools in some broader sense of just having political aspirations at all. But there's just, apart from some number of local races, there's no simple way to spend a lot of money and get the politics you want. It's checks and balances. Even though no one in the system is like aligned with you, you're not going to get the politics you want. You're going to get the politics you want. It's a beautiful
Speaker 3system we have. I'm like just every day. I mean, I'm not joking. I'm like, it's, you know, the American system is incredible and competitive markets are incredible. It's just everything just, you know, it's not perfect, but exactly as Tyler said, like there's all this magic that comes out of the system.
Speaker 2Yeah. I'm curious if you were both interviewing each other, what you would, what you would, what are you most curious about? Tyler, I'll start with you. Where are you most curious about Alex's perspective on, on, on something or, or, or, yeah.
Speaker 1Well, your origin story of how you became an economist.
Speaker 3Well, it's, that's, that's a very self-serving question, Tyler, because it has to do with you.
Speaker 1Okay. We'll say more.
Speaker 3Yeah. So, I mean, so I was, I'm an immigrant from Moldova. So, you know, I very, you know, grew up on food stamps and things like that. And I grew up in Moldova. So, you know, I grew up in America. And I, my parents didn't know anything about economics. I didn't know anything about economics. I, I went to college and I was pre-med because I knew what a doctor was. My parents knew what a doctor was. And in college, like you had to take like prerequisites. And I took an econ one class and I thought this was kind of fun, but I'm still pre-med. And I took it, I got good grades, made, made my GPA higher than my pre-med classes suggested. And then, you know, I was going to go to medical school. I applied to medical school. I got into medical school. And then, you know, I started reading Marginal Revolution and I started listening to, uh, you know, that I, I also heard like an NPR article, uh, uh, program with Richard Thaler. And between the both of you guys, it was just, oh, I can do this with economics. This is incredible. I only knew one part of that story. So thank you. I knew the Marginal Revolution part. Yeah, no, I mean, it was, but it was, because you were at that point, this was 2008 or nine. Um, there weren't a lot of these economists like, you know, actually showing you how much you can do with the topic. And so I actually, you know, I pulled out of medical school of the, of the process. I went to the only school that accepted me, which was, you know, thank you, UC San Diego. Um, and, uh, yeah, the rest is history. I, I, this is just like the, such an incredible field and knowledge base to kind of understand to the extent that we can, which is like not very high about, uh, about what's going on in this, you know, complex system that we call the economy.
Speaker 1Can I ask a follow-up? Will and should Moldova become a part of Romania?
Speaker 3I mean, my, yes, that's my view.
Speaker 1And they will. I figure they should, but they will.
Speaker 3Okay. Yeah, I, I, they were originally, I mean, it was the country was called Asarabia. My grandma, you know, she was from Basarabia. She wasn't from Moldova. Um, and it, I think it, this is probably a controversial opinion, but I, I, I'm actually not from Moldova. I'm from Transnistria. Ah, okay. Now it's all clear. Yeah. So like, you know, that we were refugees from Transnistria to America. So there was a civil war. We were like in basements hiding from bullets and all that type of stuff. And then we, we ended up coming to America as refugees. But I think it's like, you know, the, the, if you know, Moldovan culture, it's very similar to Romanian culture. Um, the language is almost the same.
Speaker 1When the VCs say always ask for the origin story, they're so correct.
Speaker 2Yeah. The, um, Alex, what, what question would you ask Tyler?
Speaker 3Um, I, I want to know your, because you were so early on this and I don't think I've ever heard this story of like, how, how were you kind of attracted to AI and thinking about it? So seriously. Um, and you were, you were very early about like the transformative potential of this technology. Like, uh, and I'm just curious, like how, how, how did you, how did you realize that earlier than most?
Speaker 1Well, all chess players have a story like this. I was in early years at like a chess prodigy. So I was age 13 playing in New Jersey and they wheeled in a computer called Tinkerbell that played against the humans and it sucked. But I saw what the progress was like. But the point about chess players is this. They all know if they're any good, the chess is not like pocket calculator stuff. It's intuition. And if you see computers conquer an area that you already knew was mostly intuition, you will find a very easy to believe they can conquer a bunch of other areas. So Kasparov was early to this. Ken Rogoff didn't need much AI pilling. He was early to this. And the chess people were the first ones there. And that's a very simple direct story. And that's how it happened. And Demis too. Demis too. Absolutely. Common stories. You see it, you know, kick butt by the 1990s and it beats Kasparov, Deep Blue. And Deep Blue is not a great model, but it's like, okay, it can do this. So you're just waiting for it to spread to everything else. But it took a long time, in fact.
Speaker 3Yeah. I think this is like a great example of like, you know, a lot of people would have looked at AI and said, hey, this, it can't do this. It can't do that. It can't do that. And it can't do that. But it's, it, it, that it's going to be able to do almost everything because, you know, like thinking about like RLVR or something like that, like the verifiable reward part of post training. A lot of people like use that as like copium, like, oh, that's not verifiable. That's not verifiable. But like, if you train the verifier, almost anything will become verifiable.
Speaker 1The fact that you can criticize it means it's verifiable almost, right?
Speaker 3Yeah, exactly. Exactly. So it's like, it's, it's, uh, and, and chess is like a really great example. Like you see, if you're a really good chess player, you know, exactly how much that, that intuition is required and how far that's going to go in
Speaker 1the future. Marecore is working on the poets. I'm not sure how far along they are. Uh, but it seems to me it should work. Mm-hmm. And create better poetry. Right. To write great poetry. I think, uh, in less than two years, it will be possible if people put their minds to it at all. I don't mean, I don't mean a big major effort with billions, but just some kind of effort at all. Bring in a dozen poets, pay them something, right? Doesn't have to be that much. Let them loose, have some checks and balances on the process. It'll work. Yeah.
Speaker 2So some people at the firm think that, uh, the idea of, of taste being uniquely human is, is a coke that, uh, the, the, the systems will, are already good at it.
Speaker 1They already have great taste. You ask them about classical music, like what's the best recording of Sibelius's fifth, whatever, they give great answers. than the penguin guide right now.
Speaker 2Yeah. Fascinating. Okay. So there's a few grab bag questions I want to ask before we wrap. One is, did you guys see Dorkesh's sort of essay he collaborated with somebody on sort of basically saying that Piketty, you know, wasn't right about the last century, but his model of, you know, RG, et cetera, might be right in the AI century. I'm curious if you had a take on that, Alex.
Speaker 3Yeah. So this is with Phil Trammell, I think. Capital in the 21st century. I think that's a, that's, I think it's the, the idea that economic activity may be divorced from labor and be kind of focused on, on capital. That, that's kind of the broader idea of the piece. I think, I think many people took, took kind of had issues with the actual methodology of the piece, which, you know, I won't, I won't speak to, but the broad idea of the technology potentially increasing this concentrate, the, the, the reliance on capital. I think that is a, that is a scenario that could happen. But there's alternative scenarios and this is things that I've written about. And I think Tyler also alluded to this, where depending on there's alternative scenario where like, you know, there's just more jobs created and these jobs are valuable because of the fact that they're scarce. Right. So the idea is that, you know, value moves towards scarcity. These, this is kind of like the whole bottleneck logic. And then if those jobs are in a rich society, they, there will be a lot of value going to those jobs. And it's just a question of how many of those jobs exist and whether it can support the entire economy. And historically, that's been the case. But that's been, that's the real big question mark. But these are things that we just need to watch for with data. So like, by paying attention to the data, we're not going to be able to get to the real big question mark. The paper with, you know, an economic policy for AGI just basically makes that argument. We need to be tracking where, where the economy is going with AI in a much more comprehensive way.
Speaker 1And I don't see that that piece really made a strong case for the latter proposition. It was mainly arguing capital can do a lot, but if the capital is commoditized, like I know people who say all the rents will go to land, all the rents will go to energy. I'm not sure that's true, but those seem more likely than all the rents go to capital. If you had to say.
Speaker 3Yeah. That's right. It goes to scarcity. And if scarcity is energy. That's where the rents will go.
Speaker 1Now, maybe AI itself makes energy less scarce. So that's why I'm reluctant to predict, but I at least see the case to think, well, the Gulf countries will make out like bandits and so on, or the landowners, but AI will change enough things. Say we can terraform land quite cheaply and desalinate water very cheaply. It doesn't seem so implausible. Well, maybe it's location of land or the best retirement spots or like Tahoe is so special. Because you're near the Bay Area, but you don't pay the taxes or, but it's very hard to predict, but think through scarcity, most of all.
Speaker 2Mm-hmm. Yeah. And in our world on that note, we think that the big scarcity is on supply and in the short term, that demand is much bigger than that it seems. But that rests on the sort of assumption or sort of the belief that when you talk to the super users, they sort of like can't get enough. And there's the super users. It's really a small percentage of people at a small percentage of companies. And AI is still not broadly diffused yet. And I think the counter to that claim would be, well, it's not having measurable impact on companies yet. So it is sort of speculative that, you know, the more that AI is used, it will eventually lead to impact on these companies in ways that justify the massive increase in spend. But, you know, we just started a new fund actually called Machine Age Fund, which is, you know, doing data centers and other elements of the sort of infrastructure stack to be able to mitigate some of these supply challenges. Leopold's paper, Situational Awareness, does that sort of stand to you as like one of the more important sort of, you know, public intellectual pieces in technology in the past few years? Or I'm curious what you think of the legacy of that.
Speaker 1In anything, not just technology. Yeah. But that said, I think he's too enamored of fast takeoff. As I told him at the time, he's assuming simulation can substitute for experiment. And thus you get this never ending loop where science just keeps on accelerating. And I don't think simulation very much subs in for experiment. It will in some areas. Alex? We agree.
Speaker 3I think, you know, I'm too Lucas critique pilled for, sorry, another inside baseball. But it's like basically why you need to understand kind of like the deep structural parameters. Of the problem in order to actually be able to predict and what simulations do and what AI is really good at is predicting. But it's it doesn't take the endogenous response of the system a lot of the time into account. And we'll we'll we'll see that kind of play out more and more.
Speaker 2I think the Robin Hanson's Age of M. How do you think that that stands up or will the legacy of those ideas be?
Speaker 1Well, I think it's wrong. It's a brilliant book and I love Robin. But when it came out. Which is now a while ago, I said, Robin, AI is going to get there first. And Robin says, no, AI might get there, but M has to come first. And I say, Robin, M probably will never happen. But if it does first, it needs AI. And I think my side of that debate is looking very good, but it's a great book to read for getting ideas and just shaking you out of, you know, the dogmatic slumbers.
Speaker 2Do you have any, we're right now talking about, speaking of timely issues, sort of Nvidia is sort of is in a big crosshairs at the moment. Uh, with sort of the Jensen situation, as you sort of view the public discourse, what do you think is misunderstood or, or how are you sort of interpreting, um, you know, what this potentially means or what lessons, if any, to, to take, or what do you think, Tyler?
Speaker 1You know, we've already said alignment is impossible. The stakes were so high, his behavior, the company, something seems out of alignment. In fact, it will be fine because of checks and balances. I don't know how they're going to operate. Uh, I think it's a great lesson. And all the different things we've been saying,
Speaker 2Alex, anything you do that.
Speaker 3Yeah, I think, um, you know, I'm, I, I, I, I think this is not really going to change the trajectory much. I think like checks and balances work. We have diversity in the supply chain to, to a large extent. Um, everything is oversubscribed demand creates supply supply creates demand. Um, so I think, I think things are going to, uh, you know, the, the, the, the system is robust.
Speaker 1The biggest change might be if something goes wrong at the macro level and NVIDIA had to step in as a buyer or lender of last resort with a weakened or gone CEO founder type. There are a lot less able to do that than they would have been had all this never happened. Like, yeah, that matters. But in the medium term, it's going to wash out and it could slow things down a bit. But again, the structural forces will take over sooner or later.
Speaker 2If you had to expect what is the next genre of sort of situational awareness like document that is going to be remarkably prescient and greatly impact our society, where would you expect that it, that to be, would it be in something in biotech would, would still be in something in AI? Like, how are, how are you thinking about sort of what are some underrated ideas right now that might be massive in the years to come?
Speaker 1Well, they're underrated to me too. I mean, everyone talks about biotech and I'm fully on board, but I think that will be especially significant. Because of regulation and the inability to use simulation to address the main issues. But again, Eric, I think you'll live to 120 and I think I will not. That's the simple way to put it.
Speaker 3I think we're going to get better models of, to the extent that AI will create better models of micro and macro phenomenon, like better structural models. So something like the model of the cell or like, like, like using digital twins. To do macroeconomics or something like that, to the extent that you are going to be able to put structure on things that we need to be able to use to predict, that's going to be the real big blow up, I think. And because that kind of gets to that idea in economics, like the reduced form versus structural debate, where you, when you understand the structure, you can really create brand new things. And I think AI, the potential for AI to do that, to help you understand and recreate that structure. I think that's really underrated.
Speaker 1It'll be great. Raj Chetty has a new paper out on digital twins.
Speaker 3Exactly. Digital twins. That's exactly what I'm referencing.
Speaker 1But he needed, you know, five co-authors and a lot of money and a research center, which is fine. It's wonderful that he's done it. But that also goes to show you something.
Speaker 2Let's close on the university question, actually, because we said earlier, you know, the academy is underrated in some ways, certainly by Silicon Valley. And that was now that we've, you know, given the props. I do. want to ask you guys, where are you most excited in terms of innovation in higher education and where is the most sort of urgent need? We, of course, are interested in the question having incubated Horowitz and Dreesen Academy. Maybe, Tyler, we'll start with you.
Speaker 1I hope you all do something and spend more money on it than what you've announced and put some of your best people on it. I don't really know what your plans are, but I'd love to be involved in some way. And as with you, Austin and Minerva and other innovations, we need much more experimentation, innovation, and measurement in higher ed. And if you're a part of that, I'm thrilled.
Speaker 2Amazing. I would love to have you. Alex, where are you most excited or what advice might you have to anybody trying to innovate in higher education?
Speaker 3I think more innovation, the better. I think higher ed's going to change drastically in the next, I don't know, I don't have a great timeline for it just because there's such slow-moving institutions. But I think it's going to, the market forces are going to lead to a shrinking of our standard idea of higher ed, of what that sector looks like, the traditional version. But I think we're going to have a ton of new innovation and we're going to have, I'm very bullish in higher ed in the long term. That's what I'll say.
Speaker 2Are you both bullish on the field of economics at the moment? Are you encouraged by what's happening right now? Where do you stand there?
Speaker 3I mean, I think the economic toolkit, I think the economic toolkit, I think the economic toolkit, and the economic way of thinking is extremely underrated. So I'm extremely biased too. But I think the way that I see things play out, like we had the earlier question about alignment and the economic toolkit is perfect for thinking about that through this mechanism design perspective. And you have this type of thing going on all the time about endogeneity in the political economy question and thinking through these sorts of consequences, thinking through the Lucas critique about you estimate a parameter but that's actually not a structural parameter. So it's not going to be valid to actually use that parameter in order to actually make the decisions. So all of these sorts of things are super important and they're going to be more important in the coming days, I think. So I'm bullish.
Speaker 1And it may just become social science. So the things economists don't know, the AI fills in for, the basic mental toolkit becomes more valuable. This is originally a Chicago idea from Gary Becker and others. And you just want really small, smart people who have some skills, a bit like those that we now call software engineers, but it will have its own name. And they have their fingers on the dials and they manage a bunch of AI agents and they're empresarios and they know economics and do all of social science. That's what I expect. So a true realization of the earlier Chicago vision.
Speaker 3I'm from the University of Chicago, so I buy that vision. I'm happy with that vision. Maybe we'll do it at the University of Chicago.
Speaker 2Amazing. Full circle. Wonderful moment to end the conversation. Tyler, Alex, thank you so much for coming to the podcast.
Speaker 1Thank you. My pleasure. Thanks a lot.
Speaker 4Thanks for listening to this episode of the A16Z Podcast. If you liked this episode, be sure to like, comment, subscribe, leave us a rating or review, and share it with your friends and family. For more episodes, go to YouTube, Apple Podcasts, and Spotify. Follow us on X at A16Z and subscribe to our sub stack at a16z.substack.com. Thanks again for listening and I'll see you next time. I'll see you in the next episode. As a reminder, the content here is for informational purposes only. It should not be taken as legal business, tax, or investment advice, or be used to evaluate any investment or security and is not directed at any investors or potential investors in any A16Z fund. Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a16z.com forward slash disclosures. We'll see you next time.

Podcast Summary

Key Points:

  1. AI models are improving at an extraordinary and accelerating rate, but this progress has not yet translated into visible economy-wide transformation or double-digit GDP growth.
  2. Economist Tyler Cowen argues that faster AI progress mainly means hitting real bottlenecks sooner, and that market data and institutional inertia point to a more normal growth path.
  3. Alex Imas contends that the relevant competitive unit is the human organization, not the individual, so agents must learn to coordinate—not just perform tasks—before real economic transformation occurs.
  4. Current AI agents remain poor at coordination and mostly perform parallel processing, and a breakthrough in coordination could break the Coasean ceiling on firm size.
  5. Extreme AI revenue projections lack justification because no other sectors have been identified that would grow fast enough to support them, and frontier labs are now competing heavily on price.
  6. Runaway growth would require knife-edge conditions such as no diminishing returns to idea production, and current data estimate that parameter well below one.
  7. GDP remains a useful but increasingly imperfect measure, and alternatives such as welfare-adjusted metrics, longevity effects, and digital-twin structural models will grow in importance.
  8. Alignment is better framed through mechanism design and checks and balances rather than perfect agent alignment, and recent incidents like the Hugging Face episode reflect human error and broken incentives rather than catastrophe.

Summary:

This A16Z podcast episode features Google DeepMind's Alex Imas and economist Tyler Cowen discussing why rapidly advancing AI has not yet produced extraordinary economic growth. Both agree that model capabilities are accelerating dramatically, with agentic systems now performing real work rather than acting as fancy chatbots. However, they argue that benchmark hockey sticks do not automatically translate into economic takeoff. Cowen notes that faster progress often just means encountering real bottlenecks sooner, and that institutions, regulation, and physical infrastructure move slowly. He points to normal-looking market data and implausible revenue projections from AI labs as reasons for caution, while Imas emphasizes that the relevant competitive unit is the human organization, which benefits from comparative advantage and coordination that current agents lack.

Imas's research suggests that even minimal AGI would not transform the economy; the real breakthrough would come when agents coordinate well enough to outperform human organizations, potentially breaking the Coasean ceiling on firm size. Such a breakthrough could cause abrupt labor-market disruption, though both guests expect a more gradual transition overall. They also discuss alignment through a mechanism-design lens, arguing that checks and balances matter more than perfect alignment, and that incidents like the Hugging Face episode stem largely from human error. On measurement, they see GDP as underrated but increasingly incomplete, with welfare-adjusted metrics, longevity effects, and structural models likely to gain importance. Both remain bullish on AI's long-term benefits but expect the path to be slower and more normal than many in the Bay Area predict.

FAQs

Better models don't automatically produce double-digit GDP growth because of institutional inertia, regulation, physical infrastructure bottlenecks, and the difficulty of finding productive uses for AI. The rest of the economy must also see productivity gains before it can pay for these products.

You would need knife-edge conditions, most importantly no real bottlenecks in the idea production function, so that returns to ideas increase rather than decrease. Current data suggest returns are still diminishing, placing us in a normal single-digit growth world.

Agents need to get much better at coordinating with one another, not just performing tasks competently. If coordination improves enough, AI organizations could break the Coasean ceiling on firm size and outperform human organizations.

Not soon. Even reaching 1.5% of GDP within five years would be enormous given how large GDP is, and 6% is far off.

No. They largely agree with the San Francisco consensus on capabilities; their skepticism concerns how quickly those capabilities translate into visible economic transformation.

Rather than relying on perfect alignment of individual agents, use mechanism design: build systems of checks and balances and separation of powers so that even self-interested agents behave in the principal's interest.

Chat with AI

Loading...

Pro features

Go deeper with this episode

Unlock creator-grade tools that turn any transcript into show notes and subtitle files.