Tyler Cowen on how AI will reorder economies, schools, and spirituality

Western skepticism and anxiety about AI (shown by Adelman and Ipsos market research) contrasts sharply with East Asian enthusiasm (70% excited and optimistic); this reflects that the West may have more to lose with service-sector jobs built around white-collar work.

“So there's data from two market research companies Adelman and Ipsos which shows that particularly in the Anglosphere people are skeptical, nervous and not excited about the prospects of AI. And if you go to East Asia in particular, those numbers are entirely inverted. You know, 70% of people are excited and and optimistic about the prospect.”

GPT-03's depth on classical music—when asked what to listen for in Sibelius's first symphony—is remarkably sophisticated and empathetic, better than nearly any human source one could consult, approaching or exceeding the knowledge of a tenured professor at a top-10 school but not quite the world's leading expert.

“I use 03 a lot to ask it about classical music. I'll say, well, I'm going to listen to Sibelius's first symphony. What should I listen for in the symphony? It's remarkably deep. You could almost say empathetic, thorough. It just understands the stuff better, I think, than any human source you would find.”

Projections of AI driving 10-25% economic growth are wrong because they ignore the critical role of human imperfections in systems and institutions; as one variable improves, remaining imperfections become more important, and adoption is constrained by regulatory obstacles, energy costs, and institutional resistance.

“They're ignoring the all critical role of human imperfections in systems and institutions. So the more powerful AI becomes, and I'm not a pessimist on that, the more it bumps against people who don't want to adopt it or institutional systems where that does not get incorporated into the workflows. It also hits higher energy prices, constraints from data centers, possible regulatory constraints. The general economic point is that as one thing in your economy gets better, the remaining imperfections become all the more important.”

Quality leadership reflected in voter preferences is one of the most important values; there's no formula to guarantee it, but it makes a big difference (example: if Britain had not had Winston Churchill, things could be very different).

“Just quality leadership that's appreciated by voters is one of the most important things. And there's no formula that gets you that, but it makes a big difference. Let's say Britain had not had Winston Churchill, right? Things could be very different. leaders ultimately reflect the people who choose them and we need to do better.”

Headwinds against economic growth include aging populations, low fertility rates, and governments doing many stupid things, which will lower growth rates independent of AI adoption.

“Well, we're aging, right? Uh we have low fertility rates. Our governments are doing many stupid things, especially in my country. All that tends to lower the growth rate.”

Western populations (especially Anglosphere) are skeptical and nervous about AI according to Adelman and Ipsos market research; East Asian populations are inverted (70% excited); this reflects existential anxiety about meaning and purpose in Western societies, compounded by the idea of being surpassed by machines.

“So there's data from two market research companies Adelman and Ipsos which shows that particularly in the Anglosphere people are skeptical, nervous and not excited about the prospects of AI. And if you go to East Asia in particular, those numbers are entirely inverted. You know, 70% of people are excited and and optimistic about the prospect.”

South Korea and China can build nuclear power plants much more quickly than the US and Britain while maintaining safety, demonstrating that slow nuclear deployment in the West is self-imposed and not inherent to the technology or to responsible governance.

“South Korea, China can do it much more quickly and it still seems safe.”

Young people should find something they genuinely care about—philosophy, DJing, calculus, cooking—and anchor their career toward it; school systems focus on credentialing and job placement (corporate law, consulting), but those boxes may not exist, so passion and intrinsic interest become the stable coordinates.

“I actually add a third thing, which is find something you really like and really just care about. It just doesn't it almost doesn't matter what it is. It doesn't matter whether it's philosophy or DJing. It doesn't matter whether it is calculus or cooking, but just something that you you really really care about and that you can anchor towards. And I think it's a it's a relatively perhaps a new way of thinking for people who are in their late teens and early 20s because the schools are have really been thinking about how do we position them to get that corporate law job or that job in the consultancy or you know you know you're fitting everything to get into that into that box and now you're being told that box may not be there.”

Core values to defend over next 20 years: good institutions, capitalism, democratic checks and balances, political competition, free speech, reasonable degree of transparency, quality leadership appreciated by voters.

“Well, good institutions, capitalism, democratic checks and balances, political competition, free speech, a reasonable degree of transparency. I know it does not apply to all governmental secrets. Just quality leadership that's appreciated by voters is one of the most important things. And there's no formula that gets you that, but it makes a big difference.”

Superstars in digital markets now have greater concentration of advantage than 50 years ago—using AI well requires sophisticated system prompts that are 10,000 words of intricate instructions, creating a gap between those who master prompt engineering and average users.

“You see the instructions for the large language models that you know the claudes and the the chat GPTs get leaked and and they've got a prompt and they have the system prompt for uh for these these things and they're 10,000 words long. incredibly sophisticated and I look at my prompts which are two lines long and I think there's no way I could get to that degree of sophistication... we see this in other digital markets you know so I can make music now as anyone can but we have a greater degree of concentration amongst the superstars than perhaps we did 50 years ago”

Failure to reform education has two costs: (1) output loss as students are taught obsolete skills, and (2) psychological/identity costs as students graduate unprepared and feel they don't fit into the world—with some validity.

“I Well, it's both, but I think we're producing a generation of students who will go out on the labor market and be quite unprepared for what they're expected to do...you know, four, however many years in college, that was the time to do it. And we taught them things that in a way were counterproductive even. So some of it is yes output will be lower but I think many of the highest costs will be psychological people feeling they do not fit into this world and they'll be somewhat correct.”

Projections of AI driving 10-25% of economic growth are getting fundamentally wrong the role of human imperfections in systems and institutions; as one thing in an economy gets better, the remaining imperfections become more important, and this constraint will bind harder than AI capability will improve.

“They're ignoring the all critical role of human imperfections in systems and institutions. So the more powerful AI becomes, and I'm not a pessimist on that, the more it bumps against people who don't want to adopt it or institutional systems where that does not get incorporated into the workflows. It also hits higher energy prices, constraints from data centers, possible regulatory constraints. The general economic point is that as one thing in your economy gets better, the remaining imperfections become all the more important.”

Cowen does not expect huge job displacement from AI; instead, new jobs will be created incorporating AI into established institutions, but this process will be slow, and in the interim employers will be skittish about hiring because they fear new hires may become obsolete due to AI, causing a reluctance to hire that will outpace job creation.

“I do not expect huge job displacement. There will be an enormous number of new jobs incorporating AI into the routines of established institutions and also building out new institutions. Now, I do think what will happen is those new efforts will be a little slow to get underway. And in the meantime, a lot of employers will be somewhat skittish about hiring more people because they'll say, 'Well, ideally, I'd like to keep workers for 5 years. Maybe this hire will be obsolete because of AI. So, for now, I'll wait.'”

The question of whether AI will concentrate wealth in the hands of capital is uncertain; AI could easily become commodified through open source competition and Chinese competition, and it's unclear whether capital or labor will benefit more, though land may emerge as the biggest winner since AI cannot substitute for land.

“It's not clear that AI will concentrate more wealth in the hands of capital. The AI could easily become commodified. We have a lot of open source. There's potential competition from China. Those companies right now are not making money... Sometimes I think the real winners are people who have bought high-quality land because AI may substitute for capital and/or labor but it really cannot take the place of land. But I would just have a high degree of uncertainty as to where the incidents will fall.”

Globalization's contribution to economic growth over the last 70 years has probably dwarfed that of technological innovation; de-globalization represents an enormous institutional and economic disruption that will occupy corporate and government attention more than AI adoption will, creating a competing drain on resources and focus.

“globalization's contribution to economic growth has been so enormous over the last 70 years, it probably dwarfs that of technological innovation. It's a, you know, it's an institutional innovation that brought with it lots of technology innovations like supply chains and containerization and so on. And so delobalization is a really big deal, right? It's a really really big and distracting deal for companies and and governments to have to deal with.”

Emerging cultural signals (ultramarathoning, extreme vacations, vinyl record collecting, minimal tech experiences) suggest people are trying to preserve tactile, visceral, emotionally-rich experiences while also using infinite digital resources (Spotify); this 'barbell' approach is adaptive for an AI-saturated world.

“I put the, you know, the growth of ultramarathoning and extreme vacations, the rise of the vinyl record as a as a as a set of signals that people say that in amongst taking advantage of the infinite celestial jukebox that is Spotify. I also want to have some processes that are a little bit more tactile, a little bit more visceral, maybe a bit more emotional and and you know, by the numbers much more much more limited.”

No one was ready for the Industrial Revolution; institutional responses (poor laws, trade unions) addressed side effects but did not matter compared to the scale of change; by analogy, modern actors cannot prepare for or control AI transformation—only choose whom to be and what to build.

“I think we learned from history that no one is ready for a fundamental change. Britain had an industrial revolution. Different things were done with the poor laws, trade unions. Arguably those were a good idea, but they really didn't matter that much compared to the overall scope of the change. I think that's the important point for me.”

Real economic progress from AI will take a 20-year time horizon because we need to invent whole new ways of doing things and may need an entire generation of institutions and firms to die off and have new ones born again built around AI at its center, much longer than tech entrepreneurs anticipate.

“Now, you can look at particular self-contained systems. One of the first we had was chess. AI and chess is virtually godlike. AI and programming is getting there. It will take a little while, but it becomes a kind of self-contained system. maybe parts of graphic design, but all that put together is well below 10% of the economy. And once you have to interact with more complex processes, we need to invent whole new ways of doing things. We may need an entire generation of institutions and firms to die off and have new ones born again that are built around AI at its center. And don't get me wrong, all of this, I think, is going to happen, but it's much more like a 20-year time horizon.”

Working with AI machines requires two capabilities: (1) ability to use AI tools effectively, and (2) sense of direction/purpose derived from passion, which guides what to ask AI to do; direction must come from human values and interests, not from the tools themselves.

“Well two things. Can you work with these machines? you have the tools to do that and do you have a sense of the direction in which you'd want to point them because they're extremely powerful. So that has to come from your passions.”

Despite AI being smarter than humans in economics, demand for Cowen's services as an economist is up, not down, because the real-world constraints on deployment are severe and persist.

“What's interesting is how little it matters. So demand for my services in so far as I wish to supply them. It's up not down.”

GDP is an underrated metric for measuring economic progress; where people want to live correlates strongly with GDP (with few exceptions like oil monarchies), and the top 10 countries people would want to be born into are all high-GDP countries, suggesting GDP captures something real about human welfare despite criticisms.

“I think GDP is an underrated metric. If you put aside the Gulf, oil, and gas monarchies, where people want to live is where GDP is high. And the intangibles are pretty closely correlated. Like take the 10 countries you'd most want to be born into. They're all high GDP countries, right?”

Historical analogy: England was a backwater for most of history until coal became central, making England arguably the most important nation; by analogy, instead of coal, AI could make Saudi and UAE into world superpowers alongside US and China.

“I would say one is England. As you well know, England was a backwater in world history for most of the millennia. Number of reasons why that stops. But coal is a big part of those reasons. And England then becomes arguably the most important nation. So instead of coal, maybe now it will be AI. And it could be the world will have a third or maybe fourth AI power. US, China, Saudi, and UAE. That would be a very strange setup, but it's quite possible.”

Nearly a billion people have voluntarily agreed to use ChatGPT regularly without any mandate; AI is coming from one place very quickly, but the opt-out for citizens is becoming increasingly difficult even if easy now.

“Nearly a billion of us have voluntarily agreed to use chat GPT regularly, right, with no mandate to do so. But on the other hand, it is starting to change behavior. What is the opt out that a citizen has if they wanted an opt out? What does that actually look like over a 10 or 15 year period? I think it will be difficult over time.”

Some people will form a cognitive group choosing to experience being human without AI, wanting to feel grass underfoot, physical exertion, mental struggle; subgroups around the world already do this, but most will choose partial AI integration.

“I imagine a kind of a cognitive group, a group who say we want to experience being human as human. We want to feel the grass under our feet. We want to feel the sweat. We want to feel our brains ache when we think through a problem. And we don't want to engage in a consistent way with this world. And we see there are subgroups around the world that do that. But I also suspect that many of us will will choose to do that as part of our highly AI infused lives, right? That we will choose to say we're going to go on that hike and we're going to take a minimal AI support with us.”

Western anxiety about AI is partly due to 'challenge of affluence'—lack of grand narrative after Cold War's end; affluence allows for meandering whereas competition forces urgency and adaptation.

“The argument might be that we are suffering from a challenge of of affluence anyway. It's been meandering for for decades and it it has been, you know, also embedded certainly from the perspective of the West with trying to find a large story after the end of the cold war and the collapse of the the Soviet Union.”

People will use emerging tactile preferences (ultramarathons, vinyl) to ground themselves in a world spinning at 'nano speed' around them; this is psychological ballast not logical optimization.

“you know, by the numbers much more much more limited. You know, I think we're giving signals of the the sorts of things that we'll want to hold on to, not because they're logical way better, but because they help ground ourselves in a world that will be spinning at nano speed around us.”

A leap in real interest rates above expectations would signal to markets that AI productivity is higher than currently priced, causing Cowen to reconsider his forecast; current real interest rate increases are modest and mostly attributable to US government borrowing, not AI capex.

“Well, if I saw, for instance, real interest rates make a big leap upwards for some reason other than higher budget deficits, I would think, oh, the market's telling us something and then I would change my mind. Real interest rates are up a bit. It's probably mostly because of higher US government borrowing. A modest amount of it is probably from higher capex expenditures coming from the AI sector and that does reflect higher growth prospects for the future.”

AI adoption will require an entire generation of institutions and firms to die off and new ones to be born around AI at the center, but this will take approximately 20 years, not the 5-10 year timeline Silicon Valley imagines.

“We may need an entire generation of institutions and firms to die off and have new ones born again that are built around AI at its center. And don't get me wrong, all of this, I think, is going to happen, but it's much more like a 20-year time horizon.”

People at Cowen's stage of life (established, no retirement pressures) face minimal downside from AI disruption and can experiment with it freely; young people (20-25 years old) face genuine uncertainty about career viability and must choose between flexibility, learning AI, and accepting uncomfortable ambiguity.

“Well, it's easy at my age. I have no plans to retire, but if I had to retire, I could just do so. So I don't feel I face any risk. I can experiment with it, try new things without a truly significant downside. But if I were say 20 years old, 25 years old, I would be highly uncertain about my future career plans. I wouldn't be sure what to do.”

Education systems are designed to incentivize students to get good grades; this skill (optimizing for external validation) will become obsolete due to AI; schools show no signs of changing their incentive structures despite this predictable obsolescence.

“The whole system is set up to incentivize getting good grades and that's exactly the skill that will be obsolete. Yet our institutions of education I don't see that they're changing at any level whatsoever.”

Homework and computer-graded exams are made obsolete by AI; effective education requires faculty to act as mentors rather than information deliverers, which is time-intensive, non-formulaic, and requires genuine pedagogical skill—not all teachers can execute this model regardless of training.

“And it means admitting that homework is obsolete. That the easy ways of grading people like computer- graded exams, they're obsolete. Faculty teachers need to be more like mentors. That can be rewarding, but it's very time inensive. It's not something you can do by formula. It's not something an okay teacher will even necessarily be very good at.”

The conversation involves 'violent agreement' between Cowen and the guest on a 20-year timeline for AI economic transformation; but decisions and engagement can happen now, making the question what 'good looks like' over the next 20 years.

“this is a process of course that will play out likely over decades rather than years. I mean, these processes tend to and we talked about at the beginning of this discussion that we have this violent agreement of a 20-year shift in the economy, but it also feels like there are lots of decisions that could be made now or at least engaged in or participation that could happen.”

Young people using AI as a 'perma companion' to work through relationship status and feelings represents a more profound change than AI speeding up business processes—this is AI as intimate counselor reshaping how people think about themselves.

“The younger cohort, which I guess was people in their sort of early 20s, use it as a perma companion for, you know, not to not to figure out Turkish economic history in the 13th century, but to figure out the status of their relationships or make sense of their of their feelings. That to me feels like a more profound change than speeding up a particular bit of a company's processes.”

The TV series 'Humans' (from ~a decade ago) featured a character (Matty) who questioned why she should study when a robot would always be a better doctor, more patient, better at diagnosis, better at delivering care—this encapsulates the identity crisis people will face as AI surpasses humans in high-status domains.

“There was a a TV series called Humans about a decade ago. In humans, we had lifelike, very realistic humanoid who would help us around the home. And they were so good at everything. And one of the uh characters was a girl in a teenage girl called Matty, very bright. And the mother said, 'Why aren't you studying?' And Matty goes, 'Well, what's the point? The robot will always be a better doctor than me and more patient and better at diagnosing and better at delivering care and better at the empathetic side.' And it really sort of in a way encapsulated a decade ago some of what you've described now.”

AI will change religion; some people will start to regard AIs as oracles or even gods, and what humans view as uniquely human qualities will evolve in ways hard to predict, leading to pretty big shocks.

“I think it will change religion. Some people will start to regard the AIs as oracles. In some cases, even gods, what we view as the uniquely human qualities will evolve. Those matters are all hard to predict, but it seems obvious we're in for some pretty big shocks.”

At this moment, the right thing to focus on is engagement and participation in AI decisions; citizens can make decisions now or participate in processes that shape outcomes.

“we have this violent agreement of a 20-year shift in the economy, but it also feels like there are lots of decisions that could be made now or at least engaged in or participation that could happen. So, what does good look like in that frame?”

Sub-Saharan Africa has enormous youthful demographics and low current human capital; AI tutors and assistants providing access to humanity's collective knowledge via mobile devices could increase human capital significantly and enable leapfrogging in economic development.

“the global center might might shift a little bit. You talked about subsaharan Africa which of course has this sort of enormous youthful demographics and very low human capital levels right now could increase significantly once people have these tutors and assistants and really all of humanity's collective learning at their fingertips.”

AI will change religion: some people will regard AI systems as oracles and gods; this reflects how technology shapes human meaning-making and identity without democratic deliberation—most people have not chosen this, it is happening by default.

“I think it will change religion. Some people will start to regard the AIs as oracles. In some cases, even gods, what we view as the uniquely human qualities will evolve. Those matters are all hard to predict, but it seems obvious we're in for some pretty big shocks.”

Amish and Mennonite groups are examples of attempted lifestyle segregation that gradually fails (intermarriage, drift toward mainstream); this historical pattern suggests cognitive Amish will similarly fail to maintain separation over generations.

“like Christian scientists they refuse a lot of health care dwindling in number a lot of the Amish become Menanites a lot of Menanites you know marry Christians and are not that Menanite anymore”

Young people should find something they really care about (philosophy, DJing, calculus, cooking) to anchor themselves; it matters less what the domain is than that it's a genuine passion, because traditional credential paths may disappear.

“I actually add a third thing, which is find something you really like and really just care about. It just doesn't it almost doesn't matter what it is. It doesn't matter whether it's philosophy or DJing. It doesn't matter whether it is calculus or cooking, but just something that you you really really care about and that you can anchor towards.”

Education systems are optimized to incentivize getting good grades, which is exactly the skill that will be obsolete in an AI era, yet institutions of education are not visibly changing at any level.

“the whole system is set up to incentivize getting good grades and that's exactly the skill that will be obsolete. Yet our institutions of education I don't see that they're changing at any level whatsoever. There's a lot of hand ringing about how do we stop people from cheating and no look at what should we actually be teaching and testing.”

Cost of not unfreezing education systems: a generation of students going to labor market unprepared for what they're expected to do, feeling they don't fit into the world and being somewhat correct; psychological cost is high as it plays into broader existential narratives.

“Well, it's both, but I think we're producing a generation of students who will go out on the labor market and be quite unprepared for what they're expected to do...many of the highest costs will be psychological people feeling they do not fit into this world and they'll be somewhat correct. It plays into a number of stories and that we you know fictions that have have come through over the years.”

The European Union will face the biggest challenges with AI adoption: active hostility to AI, lack of own significant AI models, numerous laws making AI harder or technically illegal, insufficient power/data center capacity, lack of competitive governance (too much harmonization preventing regional experimentation).

“The biggest problems might come in the European Union where there's active hostility. At least your government will make noises about wanting to do things...European Union, I don't see progress there. They don't have their own significant models. There are numerous laws that make AI much harder to do or even like technically illegal. They need to get their act together. There's a number of uh institutional challenges in the European Union. One is that getting things done right getting power, which getting the data centers uh is particularly difficult. it doesn't have sufficient governance competition.”

Sub-Saharan Africa has enormous youthful demographics and low human capital that could increase significantly with AI tutors providing collective human learning; this region could see a major growth opportunity.

“the global center might might shift a little bit. You talked about subsaharan Africa which of course has this sort of enormous youthful demographics and very low human capital levels right now could increase significantly once people have these tutors and assistants and really all of humanity's collective learning at their fingertips.”

UC Berkeley graduates want to stay in the Bay Area and may not find the jobs they're looking for; there's an outflux every year but they're not willing to move like from Berkeley to Nashville, creating a local talent oversupply problem.

“There's plenty of people, so you graduate from Berkeley, UC Berkeley, which is an excellent school. A lot of those people want to stay in the Bay Area, and I'm not sure they'll all be able to get the jobs they're looking for, and that will for them be a problem. Again, I'm unsure how much it will show up in the aggregates, but that's a place there's an outflux every year. They want to hang around. They're not for the most part looking to move like from Berkeley to Nashville”

The trend of national income moving from labor to capital (about 10 percentage points over 50 years in the US and West) depends on how you count high earners like Mark Zuckerberg—if counted as labor, change is minimal; it's unclear whether AI will concentrate wealth in capital hands or if AI becomes commodified with open-source alternatives.

“The previous trend depends on whether you count Mark Zuckerberg as labor or capital. If you count as labor, if just a very well- paid laborer, there hasn't been much of a change. But I would say it's not clear that AI will concentrate more wealth in the hands of capital. The AI could easily become commodified. We have a lot of open source.”

With AI in particular, no one can be that far ahead of you in terms of current models—you start off very close to the frontier even though it's intimidating, unlike other sectors (e.g., fracking) where experts are decades ahead.

“Well, the great thing about AI is no one can be that far ahead of you. You know, it's not like all these other sectors of the economy or fracking. There are people out there who are decades ahead of you in fracking, but in terms of current AI models, you start off in a funny way being very close to the frontier even though it's intimidating.”

Avy Balwit, based in Silicon Valley, observes rapid iteration speed, febrile excitement, and compounded learning capability in Valley companies that rivals or exceeds the 1997-1998 dot-com era velocity, but with deeper knowledge of how to build and scale companies.

“she's in the heartland of Silicon Valley, right? and and that you know one of the things that I found fascinating actually is the speed of iteration within the valley. I haven't seen it as febrile and excited since the 97 1998 period but based on much much more knowledge of how to build companies, how to iterate, how to test things. uh and the the speed of learning there is really singing to San Francisco's ability to as to oblomerate all of this incredible talent.”

Good near-term outcomes require: (1) China does not win the AI race, (2) whatever emerges in Saudi Arabia and UAE is secure, orderly, and benefits everyone; these are unprecedented events with no experts, so outcomes depend on humility and luck.

“a good immediate future is that China does not win this race and also that whatever we erect in Saudi and UAE that it is secure and orderly and and works for everyone that's a hard thing to manage. This is these are unprecedented events like none of us know what we're doing. There are not really experts you can consult in a way.”

Market prices (currencies, equity valuations, real interest rates, volatility) across US and UK show normal conditions with no evidence of imminent radical transformation from AI; markets reflect skepticism toward claims of near-term AI-driven disruption.

“The very best estimate we have is that which is embedded in market prices and there are plenty of market prices that indirectly give signals about what will happen. I would say if you look at currencies, equity prices, real interest rates, levels of volatility, really anywhere you care to look, the world looks pretty normal, at least for our two economies.”

UAE is providing ChatGPT Plus subscriptions to all citizens, creating a politically-educated base of AI users ahead of most democracies and enabling rapid learning from AI exposure at population scale.

“UAE is buying chat GPT plus subscriptions for all of its citizens. So the political base there will be very well-versed in what's going on probably way ahead of the United States.”

In AI development, no actor can be far ahead because the technology moves so fast; even with intimidation, newcomers start near the frontier—unlike other sectors (fracking) where decades of lead time can exist.

“The great thing about AI is no one can be that far ahead of you. You know, it's not like all these other sectors of the economy or fracking. There are people out there who are decades ahead of you in fracking, but in terms of current AI models, you start off in a funny way being very close to the frontier even though it's intimidating.”

A leap in real interest rates for reasons other than higher budget deficits would signal to Cowen that he's wrong about AI's slow economic impact; if markets see high productivity from AI, demand for capital rises, raising real interest rates. Currently real rates are up modestly, mostly from US government borrowing.

“Well, if I saw, for instance, real interest rates make a big leap upwards for some reason other than higher budget deficits, I would think, oh, the market's telling us something and then I would change my mind. Real interest rates are up a bit. It's probably mostly because of higher US government borrowing. A modest amount of it is probably from higher capex expenditures coming from the AI sector”

EU federalism with excessive harmonization prevents regional experimentation, unlike the US where Texan power grid is completely different from New Hampshire's, enabling competition and learning about what works.

“Although it's sort of sort of federal, there's so much harmonization, you can't really have entirely different regimes the way you can often find in in the US. You know, the Texan power system is completely different to the one in in New Hampshire in in Virginia. And and that provides for the kind of competition that allows us to to sort of see what works and and what doesn't work.”

European defense spending at 5% GDP (now likely for many countries) will probably be done bureaucratically, rewarding existing interest groups rather than being ready for the future, barely improving actual defensive capability.

“I don't doubt that many European countries will move to 5% GDP for defense spending. But they might do it in a quite bureaucratic way that rewards existing interest groups rather than being ready for the future. And they consider it another form of fiscal policy and in essence barely end up being safer.”

Cowen doesn't feel tempted to adopt Amish-style AI rejection; he wants to live to 97 and benefit from modern medicine, and few people will choose AI rejection once they see the benefits—more like the Amish or Christian Scientists, who are dwindling in number.

“I've visited the Amish multiple times. I'm not tempted. I want to live to 97 and whatever else. I don't think many people will do this. I think it will be like the Amish where yes some people do it but they will be regarded as highly unusual outliers.”

No one is ready for fundamental economic change historically; Britain had an industrial revolution with various policies (poor laws, trade unions) that arguably mattered but didn't really matter much compared to the overall scope of the change.

“I think we learned from history that no one is ready for a fundamental change. Britain had an industrial revolution. Different things were done with the poor laws, trade unions. Arguably those were a good idea, but they really didn't matter that much compared to the overall scope of the change.”

Cowen is experimenting with AI without downside risk because he's at an age where he has no plans to retire; younger people (20-25 years old) would face high uncertainty about whether their career plans will remain viable.

“It's easy at my age. I have no plans to retire, but if I had to retire, I could just do so. So I don't feel I face any risk. I can experiment with it, try new things without a truly significant downside. But if I were say 20 years old, 25 years old, I would be highly uncertain about my future career plans.”

Even if someone tells young people they'll have more wealth and live to 97, it's still psychologically uncomfortable to face uncertainty about career plans; the logical reassurance doesn't reduce the felt discomfort of uncertainty.

“I wouldn't be sure what to do. I would do what you probably advise people to do, which is to get to know the AIS and stay flexible and then just see. But that's a somewhat uncomfortable position to be in. Even if someone tells you, well, the world will have a lot more wealth and you'll live to be 97, so don't worry. It's still an uncomfortable position to be in.”

Cowen and Avy Balwit's Free Press essay was co-authored with each writer contributing sections; Balwit believes AI change will occur much faster than Cowen; the published essay represents a weighted average of their views, with the more alarmist content about identity crisis and religious transformation reflecting Balwit's faster timeline.

“That's a co-authored essay and we each wrote it. And Abatital thinks the pace of change will be much faster than I think it will be. So the essay you could say is a weighted average of our two opinions.”

Cowen is not tempted by the Amish model because he wants to live to 97 and achieve other life goals; few people will choose cognitive Amish status; most will adopt a 'barbell' strategy: minimal AI some of the time (hiking, vacations) but AI-intensive most of the time.

“I've visited the Amish multiple times. I'm not tempted. I want to live to 97 and whatever else. I don't think many people will do this. I think it will be like the Amish where yes some people do it but they will be regarded as highly unusual outliers.”

The Gulf (UAE and Saudi Arabia) is emerging as a major AI power center alongside the US and China; Gulf leadership is building data centers, infrastructure, and universities with committed long-term investment; the region could become the 'third or fourth AI power' reshaping global power structures—analogous to England's rise via coal in the Industrial Revolution.

“The Gulf stands a good chance of being one of the most important places in the world. And I don't just mean for its oil, uh, for its data centers. And it's possible the smartest entities within a few years will be based in the Gulf. Now, who controls them, who has plug pulling rights, what leaks out to China, it's all very complicated, but there's a very good chance that will happen. And you asked for historical analogies. I would say one is England. As you well know, England was a backwater in world history for most of the millennia. Number of reasons why that stops. But coal is a big part of those reasons. And England then becomes arguably the most important nation. So instead of coal, maybe now it will be AI. And it could be the world will have a third or maybe fourth AI power. US, China, Saudi, and UAE.”

Silicon Valley AI practitioners underinvest in understanding social sciences and history, leading them to underestimate institutional resistance and overestimate the speed of world-level change; Cowen has debated these people for years and believes his historical/sociological view prevails in debate.

“when it comes to the actual world, I feel they have underinvested in the social sciences and history and I've had long extended back and forths with these people for now a few years and I'll stick with my current view. I feel I win those debates.”

Currently, opting out of AI is easy and has low cost; but over 10-15 years, AI will become so integrated into websites and services that opting out becomes as difficult as not owning a mobile phone—possible but highly inconvenient and socially isolating.

“I think it will be difficult over time. Now it's super easy. But what we now call websites, I suspect will be configured to work with the AI more than work with you. For a long time, there'll be some kind of dual system, but it will be like not having a mobile device, which you can do today still, but it's highly inconvenient. and your friends expect you to have one.”

The UK has less federalism than the US (simplifying infrastructure approvals) but faces a larger fiscal crisis, a talent crisis of unexplained origin, and administrative state constraints that impede the kind of growth Silicon Valley envisions despite theoretical governance advantages over the US.

“Now, in the UK, you have far less federalism. You have a bigger fiscal crisis. You have a talent crisis for some set of reasons I can't explain. You just don't seem to be able to get your act together to do a bunch of things. I don't know why.”

Good immediate future regarding AI: China does not win the race, and whatever infrastructure is erected in Saudi Arabia and UAE is secure, orderly, and works for everyone—these are hard things to manage and unprecedented events where no one knows what they're doing.

“a good immediate future is that China does not win this race and also that whatever we erect in Saudi and UAE that it is secure and orderly and and works for everyone that's a hard thing to manage. This is these are unprecedented events like none of us know what we're doing. There are not really experts you can consult in a way.”

Market prices (currencies, equity prices, real interest rates, volatility levels) show the world looks pretty normal for the US and UK; there is no radical overturning reflected in markets, and people with real money at stake agree with the position that AI's economic impact will be modest.

“The very best estimate we have is that which is embedded in market prices and there are plenty of market prices that indirectly give signals about what will happen. I would say if you look at currencies, equity prices, real interest rates, levels of volatility, really anywhere you care to look, the world looks pretty normal, at least for our two economies...People who have real money at stake are essentially agreeing with the position the two of us have set out.”

Silicon Valley is moving faster and more febrile than at any point since 1997-1998, but now with much better knowledge of how to build companies, iterate, and test; the speed of learning there is remarkable.

“the speed of iteration within the valley. I haven't seen it as febrile and excited since the 97 1998 period but based on much much more knowledge of how to build companies, how to iterate, how to test things. uh and the the speed of learning there is really singing to San Francisco's ability to as to oblomerate all of this incredible talent.”

One positive for Europe is it's woken up to hard power and energy independence, which might drive infrastructure construction and capital investment in factories—historically good for economies.

“I suppose that the sort of one positive that that Europe has is in a way it's woken up right it's woken up to hard power. It's woken up to needing to have energy independence and energy sovereignty. And that might mean, you know, shovels going into the ground and things really being built and factories being built, which historically has been quite a good thing for for economies.”

UAE is buying ChatGPT Plus subscriptions for all its citizens, ensuring the political base will be well-versed in AI capabilities probably ahead of the United States.

“And I'm sure you know that UAE is buying chat GPT plus subscriptions for all of its citizens. So the political base there will be very well-versed in what's going on probably way ahead of the United States.”

Websites will gradually become configured to work with AI more than with humans; for a long time there'll be dual systems, but eventually not having AI integration will be like not having a mobile device—still technically possible but highly inconvenient.

“What we now call websites, I suspect will be configured to work with the AI more than work with you. For a long time, there'll be some kind of dual system, but it will be like not having a mobile device, which you can do today still, but it's highly inconvenient. and your friends expect you to have one.”

Western democracies including the US have seen a shift in fundamental national tempo and direction over the last 6-7 months, reflecting deep-seated expression of pressure; leaders reflect the moment rather than shaping it.

“we've seen this across a number of western democracies including of course uh you know the US that sort of change in the you fundamental tempo and direction of the nation over the last sort of six or seven months reflecting a you know sort of a deep-seated expression. So that that feels to me like a really tough place for us to bring our conversation to a close”

Young recent college graduates entering the labor market for the first time will experience the strongest negative AI labor market impact, concentrated in places like the Bay Area where graduates want to stay and expect certain job types.

“I think parts of each of our countries where you have a lot of young recent college graduates entering the market for the first time that's where the negative impact will be the strongest. I have a decent sense where that is in America. You would know better about the UK, but that's exactly where I think it will show up.”

Cowen prefers to host dinner with AI engineers rather than economists currently because the field of economics has become boring while AI engineers represent the frontier of interesting ideas and capability.

“The economist right now is boring. I would rather have the AI engineer. And these are not hypotheticals. If you look at where I go and with whom I spend my time, I can vouch that I'm telling you the truth.”

Leaders reflect the people who choose them and 'we need to do better'; recent Western history shows drift toward selecting leaders who reflect immediate pressures rather than long-term clarity, as visible in shifts in 'fundamental tempo and direction' of nations over 6-7 months.

“leaders ultimately reflect the people who choose them and we need to do better. Unfortunately, people also find leaders who reflect the pressures of the the moment and we've seen this across a number of western democracies including of course uh you know the US that sort of change in the you fundamental tempo and direction of the nation over the last sort of six or seven months reflecting a you know sort of a deep-seated expression.”

Young people use AI differently than older people: older users like Cowen use it as super-Google search and thinking support; younger cohorts (early 20s) use it as perma-companion for emotional and relational guidance, which is more profound change than productivity speed-ups.

“Sam Alman said something uh a few weeks ago where he he described how different generations uh were using chat GPT and you know people like me use it as a super Google search plus uh sketching out on bits of paper as I do my thinking. And he described that, you know, the the younger cohort, which I guess was people in their sort of early 20s, use it as a perma companion for, you know, not to not to figure out Turkish economic history in the 13th century, but to figure out the status of their relationships or make sense of their of their feelings. That to me feels like a more profound change than speeding up a particular bit of a company's processes.”

The Gulf (UAE and Saudi Arabia) is building out infrastructure and capacity from energy to data centers to universities; this could create a new semi-central power pole in world geopolitics alongside US and China.

“at the same time we have this sort of enormous pole emerging in the Gulf uh between the UAE and Saudi um and and building out infrastructure and capacity and capability really from you know the energy to the data centers to the universities upwards.”

The European Union faces the biggest risk: active regulatory hostility toward AI, lack of homegrown AI models, numerous laws making AI illegal or impossible, insufficient governance coordination due to quasi-federal structure, and inability to compete technologically—EU cannot progress on AI under current constraints.

“The biggest problems might come in the European Union where there's active hostility. At least your government will make noises about wanting to do things. I'm not sure that will be followed through on as it goes through the bureaucracy. Uh but European Union, I don't see progress there. They don't have their own significant models. There are numerous laws that make AI much harder to do or even like technically illegal. They need to get their act together. There's a number of uh institutional challenges in the European Union.”

Clayton Christensen's 'Innovator's Dilemma' concept has shaped a generation of managers to expect and embrace disruption, capital markets are more fluid, consumer demand pull is strong, and governments are building infrastructure, all of which could accelerate AI adoption beyond the historical 20-year precedent.

“could it be the case that a generation of managers who grew up in the shadow of Clayton Christensen's book the innovators dilemma for whom the word disruption has been part of their career. Capital markets that move more freely. The scale of the demand pull from consumers for this. the fact that governments may be doing many stupid things but also seem to be resigned in some cases and excited in other cases to build out the infrastructure as quickly as we can. Does that provide a tailwind that might see this moving quicker than that 20-year period?”

One EU positive: recent geopolitical shocks (Ukraine, energy independence, hard power) have awakened European consciousness to the need for physical infrastructure and capability-building; defense spending increases to 5% GDP could benefit broader economic capacity—but only if done efficiently rather than captured by incumbent interests.

“the sort of one positive that that Europe has is in a way it's woken up right it's woken up to hard power. It's woken up to needing to have energy independence and energy sovereignty. And that might mean, you know, shovels going into the ground and things really being built and factories being built, which historically has been quite a good thing for for economies.”

US managers and capital markets are the best in the world but still contain substantial mediocrity; the US has been unable to quickly permit and build new nuclear power plants—taking 10 years if lucky—demonstrating that managerial quality and capital fluidity are insufficient to overcome structural delays in self-imposed regulatory and institutional burdens.

“I don't think our managers and consultants and capital markets are also great. They're probably the best in the world in the United States and there's still plenty of mediocrity there. Uh, when was the last time the US built a new nuclear power plant? What are the prospects for that in Britain? In both countries, it's looking up, right? That's good. Probably at the moment, if we start one today, it takes us 10 years if we're lucky.”

Cowen himself now 'writes for the AI'—he writes things intending that AI will read them and know him, treating AI as his primary audience rather than humans, and this reflects a fundamental shift in how he thinks about writing and knowledge production.

“And I now do things I call it writing for the AI. I write things because I want the AI to know it about me or know it about the world. And very often they're my number one audience, not the humans.”

Cowen uses AI as an 'infavore' (presumably informational counselor/advisor), asking it to recommend restaurants where Tyler Cowen would want to eat; the AI knows his preferences from his extensive online writings on food, creating a sense of being known.

“I very much use them as an infavorore would. And it also knows me. Like I was in Mexico recently. I've written a lot on food and dining out and I say pick a restaurant for me where Tyler Cowan would want to eat and it knows what I want. Like it I'm delighted it knows me. That for me is just a thrill.”

Cowen and Amodei co-authored essay arguing AI could fundamentally challenge what it means to be human by 2030, but Amodei believes pace of change will be much faster than Cowen does; the essay is a weighted average of their two opinions.

“That's a co-authored essay and we each wrote it. And Abatital thinks the pace of change will be much faster than I think it will be. So the essay you could say is a weighted average of our two opinions. I'm fine with doing things like that. Uh but she does think it will be faster.”

TV series 'Humans' captured this dynamic a decade ago: a teenage girl named Matty has a humanoid robot that's better at everything (doctoring, empathy, diagnosis, care), making her ask 'why study?' because the robot will always be better; this encapsulates the psychological challenge.

“There was a a TV series called Humans about a decade ago. In humans, we had lifelike, very realistic humanoid who would help us around the home. And they were so good at everything. And one of the uh characters was a girl in a teenage girl called Matty, very bright. And the mother said, 'Why aren't you studying?' And Matty goes, 'Well, what's the point? The robot will always be a better doctor than me and more patient and better at diagnosing and better at delivering care and better at the empathetic side.'”