What this covers

Tyler Cowen presents a conversation centered on why AI will not catalyze an explosive economic takeoff. Rather than framing intelligence as the limiting factor in growth, he argues that human institutions, regulation, adoption speed, and scarce concentrations of talent form the real binding constraints. The conversation ranges across economic history, institutional design, leadership quality, and the mechanics of technological diffusion to show why even transformative intelligence will compound slowly—improving growth rates by roughly half a percentage point annually rather than producing the dramatic acceleration many anticipate. This slow boost, compounded over decades, will nonetheless reshape processes like drug development and scientific discovery, but only as existing institutional scaffolding (trials, regulation, bureaucracy) persists.

Cowen grounds his argument in historical precedent and economic principle. He invokes cost disease—the generalized insight that when one production factor becomes abundant, other constraints grow more binding and the marginal value of additional intelligence falls. He shows that even the Industrial Revolution, perhaps humanity's most pivotal transformation, grew at only 1.5% annually. He examines why sectors representing half the US economy—government, healthcare, education, nonprofits—will integrate AI only across three decades, creating a ceiling on aggregate growth. He turns to talent as a scarce bundle not easily scaled: elite clusters form because top people spot other top people, and the rarity of genuine achievement at the highest levels may resist AI acceleration. The conversation also considers geopolitical risk—how new technologies trigger technology-driven arms races that autocracies can win—and explores human resistance itself as an underestimated bottleneck, since populations trained for an older world will resist the disruption AI brings.

Sharpest takeaway

Cowen argues that AI will not produce explosive economic growth because the binding constraints on progress are human institutions, regulation, diffusion speed, and scarce bundles of talent—not intelligence itself—so AI's effect will be a slow, transformative-over-decades boost rather than a rapid takeoff.

  • Adding intelligence makes other production factors more binding, so the marginal value of more IQ falls (cost disease generalized)
  • Diffusion of new technologies is universally slow and no credible model shows AI will change that
  • Half the US economy (government, healthcare, education, nonprofits) will adopt AI only over ~30 years

The claims · ranked35 claims · weighted by value

This asset isn't compiled yet

You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.

0.80

Even the Industrial Revolution—arguably the most important event in human history—had a typical rate of economic growth of only about 1.5% a year, showing that the future is about compounding and persistence over the long run, and human beings won't change that much even with far more IQ and conscientiousness.

factualhigh valueestablishednovelty 2/4durability 4/4· Tyler Cowen

“Typical rate of economic growth during that period was about 1 1/2 percent.”

0.78

Cost disease is a special case of a general principle: when one factor of production (intelligence) becomes abundant, the other constraints in the system become more binding, raising their marginal importance and lowering the marginal value of additional intelligence—so adding intelligence is self-limiting on growth.

causalhigh valuecontestednovelty 3/4durability 4/4· Tyler Cowen

“Well, that just means the other constraints in your system become a lot more binding, that the marginal importance of those goes up, and the marginal value of more and more IQ or intelligence goes down. So that also is self-limiting on growth”

0.77

Transformative periods of new ideas (17th-century England with the scientific revolution, the rise of English global power, the first sustained ~1% growth from the 1620s) bundle together very good and very bad outcomes—civil war, regicide, radical ideas, the birth of libertarianism, and the brutal conquest of the New World all at once—and we may be in for a similar dose now with AI.

causalhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“England is the place that gets the compounding at 1% starting in the 1620s, and somehow they go crazy: civil war, kill the king, all these radical ideas. Libertarianism comes out of that, which I really like: John Milton, John Locke.”

0.77

Cowen predicted at peak Effective Altruism that the movement would largely fall apart because, like other movements he has watched since the 1960s (including the Berkeley free speech movement and libertarianism), it lacked clearly crystallized permanent institutional incentives—unlike Goldman Sachs or legal partnerships—and combined excess enthusiasm with an unstable mix of secular and semi-religious elements, while its two or three best ideas would remain very important.

forecasthigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“I just have a long history of seeing movements in my lifetime, from the 1960s onwards, including libertarianism. There are common patterns that happen to them all.”

0.77

The clustering of psychopathic 20th-century leaders had no single macro cause, but big new technologies trigger arms races that bad, non-democratic regimes can win (Soviet Russia, Nazi Germany, Mao's China); combined with a highly disoriented civilization (radical shifts in art, music, the rise of imperialism, no stable world order) and bad luck in who reached the top, the result was catastrophic—a dynamic relevant to AI today.

causalhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“I think when you get big new technologies, and this is relevant for AI, you get a lot of new arms races. Sometimes the bad people win those arms races.”

0.75

The main misgiving about progress is war: new technologies are reliably turned into instruments of war (17th-century England, electricity and the machine gun, nuclear weapons), and there may be a ratchet effect where wars become more destructive so that even if rarer, each one is a real doozy; the relatively peaceful Pinkerian trend is now fraying at the edges with numbers moving the wrong way.

forecasthigh valuecontestednovelty 2/4durability 4/4· Tyler Cowen

“But I do observe in history that when new technologies come along, they are turned into instruments of war, and some terrible things happen.”

0.73

Having a founder remain in charge matters because courage is a scarce input in major decisions: founders already have courage to begin with, they need less of it to push through big changes, and their authority makes it hard for others to block them—so founder leadership both economizes on courage and selects for it.

causalhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“I think courage is a very scarce input in a lot of decisions. Founders have courage to begin with, but they also need less courage to see through a big change in what the company will do.”

0.73

AI will not produce explosive economic growth (20%+) because when some sectors grow rapidly, cost disease appears in the large parts of the economy that can't use AI well; with government, healthcare, education, and nonprofits making up roughly half the US economy and adopting AI only over ~30 years, the net effect is only a modest boost to growth rates.

causalhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“It's very hard to get explosive economic growth for any reason, AI or not. One problem is that some parts of your economy grow very rapidly, and then you get a cost disease in the other parts of your economy that, for instance, can't use AI very well.”

0.73

Hayek pinpointed one factor—that the worst tend to rise in autocracies—but it is only a tendency, not a law: the UAE and several other family-based, clannish Gulf monarchies have proven more stable and meritocratic than expected and are not run by their worst people, showing that culture matters and Hayek was writing about a specific place and time.

factualhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“There are quite a few autocracies in the world today where the worst people have not risen to the top. The UAE would be, I think, the most obvious example.”

0.73

Chad Jones's growth model is too much a one-factor (population) model and is not very predictive; large increases in effective world population and purchasing power have historically coincided with most regions becoming less innovative, so growth is really about the quality of a society's best people and institutions, which is fragile.

factualhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“His model is far too much a one-factor model, right? Population. I don't think it's very predictive. We've had big increases in effective world population in terms of purchasing power. A lot of different areas have not become more innovative.”

0.73

The relationship between scale and output is oddly asymmetric: shrinking an economy or organization too much can cause it to implode and disappear (as with companies and nonprofits), but proportionally growing it does not deliver proportional 3x or 4x output gains—an asymmetry that is emotionally hard to internalize.

factualhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“So shrinking can kill you. Just like companies, nonprofits, if they shrink too much, often they just blow up and disappear. They implode. But that doesn't mean that growing them gets you 3x, 4x, whatever, proportional to how they grow.”

0.73

Rather than a uniform competency crisis, there is increasing variance in the talent distribution: the top is getting much better (measurable in chess, NBA basketball, science, internet writing), the very bottom is also improving (youth crime falling since the 90s), but a thick middle band from above the bottom to a bit above the median is clearly getting worse.

factualhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“I think there's basically increasing variance in the distribution. Young people at the top are doing much better, and they're far more impressive than they were in earlier times.”

0.73

Clusters of elite talent (like Stripe hiring Greg Brockman or labs with multiple Nobel laureates) form largely because top people attract other top people who are good at spotting them—Patrick could hire Greg Brockman because he is Patrick—and while members make each other much better, the core scarcity is that achievement at the very highest levels is extremely rare and may be hard for even strong AI to help with.

causalhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“He was able to hire Greg Brockman because he's Patrick. It's very simple... if you're Patrick and John, you're going to attract some Greg Brockmans... the Greg Brockmans are pretty good at spotting who are the Patricks and Johns.”

0.73

Tech wields more Washington influence than it realizes, channeled through national security imperatives that have kept AI development proceeding; tech's weakness comes from being geographically concentrated in one dominant-party area and from its distortions being invisible to most Americans—unlike geographically dispersed, ruthless lobbies such as community bankers who have connections to every member of the House.

causalhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“I think you're getting a lot more influence than maybe you realize, quickly, through national security reasons. The feds have not stopped the development of AI”

0.71

The specific bottleneck to AI-driven progress is humans: as AI begins to visibly change the world, there will be massive opposition—not necessarily from doomsday concerns but from people who trained themselves and their children for a different world and don't want this—and that human resistance will be a major and unpredictable bottleneck.

forecasthigh valuecontestednovelty 2/4durability 3/4· Tyler Cowen

“What are the specific bottlenecks? Like, why? Humans. Here they are. Bottleneck, bottleneck.”

0.71

The logic by which AIs making more AIs hits limits was already understood by classical economists: Malthus and Ricardo analyzed how rapidly breeding humans faced diminishing returns against other scarce factors, so if they saw today's AI they would expect it to extend the frontier but not solve all problems—an insight Bay Area thinkers, who overvalue intelligence, tend to miss.

factualhigh valuecontestednovelty 3/4durability 4/4· Tyler Cowen

“Ricardo knew this, right? He didn't call it AI, but Malthus and Ricardo, they all talked about this. It was just humans for them.”

0.69

AI will be smart and conscientious and will boost the rate of economic growth by something like half a percentage point per year; while imperceptible in any given year, compounded over 30-40 years this is an enormous, world-transforming difference—effectively making 20-year processes (like drug development) take 10 years—but it does not feel revolutionary because the underlying systems of trials and regulation remain.

forecasthigh valuecontestednovelty 3/4durability 2/4· Tyler Cowen

“I think they will boost the rate of economic growth by something like half a percentage point a year. Over 30-40 years, that's an enormous difference. It will transform the entire world.”

0.69

Much of the equity premium may now be in the past because sophisticated investors—especially VCs and Sand Hill Road—are very good at internalizing value, so returns will increasingly be earned in private markets rather than accruing to public equities and public pension funds, which makes it unclear what one can do with public equities.

forecasthigh valuecontestednovelty 3/4durability 2/4· Tyler Cowen

“maybe a lot of the equity premium is in the past. People, especially in this part of the world, are very good at internalizing value, and it will be held and earned in private markets and by VCs rather than public pension funds.”

0.68

Most of sub-Saharan Africa still lacks reliable clean water even though the intelligence required to provide it is not scarce, demonstrating that humanity is bottlenecked by institutional and implementation constraints rather than intelligence—and capturing the benefits of strong AI faces the same kind of non-intelligence bottlenecks.

factualhigh valuecontestednovelty 2/4durability 3/4· Tyler Cowen

“Most of sub-Saharan Africa still does not have reliable clean water. The intelligence required for that is not scarce. We cannot so readily do it.”

0.66

Labor-market data show the returns to IQ as it translates into wages are amazingly low and fairly insignificant; success comes not from raw intelligence but from a scarce bundle of traits—being an 8-9 out of 10 across eight or nine areas plus one exceptional area and a lot of determination—and it is the bundle, and the bundle's interaction with the world, that is scarce.

factualhigh valuecontestednovelty 3/4durability 3/4· Tyler Cowen

“If you look at labor market data, the returns to IQ as it translates into wages, they're amazingly low. They're pretty insignificant.”

0.63

Contrary to a leadership-decline narrative, most of the US bureaucracy in Washington—generals, national security people, top agency officials, Treasury, the Fed—is impressive and shows no signs of getting worse, and recent presidential candidates like Obama and Romney were genuinely strong; the founders (Jefferson, Hamilton) achieved great things partly because founding a country allows feats that can't be replicated later.

factualhigh valuecontestednovelty 2/4durability 2/4· Tyler Cowen

“Most of the U.S. bureaucracy in Washington I think is pretty impressive: generals, national security people, top people in agencies, people at Treasury, people at the Fed. I interact with these groups pretty often.”

0.63

Much talk of declining test scores is overblown: a lot of tests, especially PISA, have pushed more people into taking them, so the median falls partly for compositional reasons; adjusted for that, scores are roughly constant, with at most a very modest decline, some of it pandemic-related and recovering slowly.

factualhigh valuecontestednovelty 2/4durability 2/4· Tyler Cowen

“The median goes down. A lot of tests, they've pushed more people into taking the test -- PISA scores in particular. I suspect those scores, adjusted for that, are roughly constant”

0.63

The peaks of young achievement are clearly better now than in the past—a 24-year-old today doing high-quality podcasting has no equivalent decades ago, and historical figures like Churchill were not as impressive young as their reputations suggest, with most of Churchill's early life being that of a destructive failure until his real achievements came late.

factualhigh valuecontestednovelty 2/4durability 2/4· Tyler Cowen

“Was there a 24-year-old like you doing the equivalent of podcasting? It's just clearly better now than it was back then.”

0.63

China is the least successful Chinese society, struggling to match Mexico's per capita income, in part because its huge scale is one of its big problems, and there is a fear that democratizing would let the median voter fail to protect elite interests.

factualhigh valuecontestednovelty 2/4durability 2/4· Tyler Cowen

“They're the least successful Chinese society in part because of their scale. Their scale is one of their big problems.”

0.62

The formal definition of autism is framed entirely around present-day deficits, under which essentially no high-functioning person qualifies; but there is some other, not-yet-pinned-down coherent definition under which perhaps a third of a high-achieving audience would qualify, indicating the whole concept needs a major reboot, and the word itself is a poor label worth giving away.

definitionhigh valuecontestednovelty 3/4durability 2/4· Tyler Cowen

“If you look at the formal definition, it's all about deficits that people have right now. If you define it that way, no one here is autistic. If you define it some other way, which maybe we haven't pinned down yet, a third of you here are autistic.”

0.60

Because AIs trawl and retain everything written online, they will form a salient persona of an author and increasingly constitute a large part of one's audience with accumulating purchasing power; Cowen is therefore deliberately writing his books for AIs rather than human reviewers, and argues that anyone who believes even a modest version of AI progress is foolish not to do the same.

normativehigh valuespeaker onlynovelty 4/4durability 2/4· Tyler Cowen

“The last book I wrote is calledGoat: Who's the Greatest Economist of All Time? I'm happy if humans read it, but mostly I wrote it for the AIs.”

0.59

Under slow takeoff, AI gives us more degrees of freedom, making all choices and problems more complex and increasing the need for guidance; consequently all inputs other than the AI rise in marginal value, which leads Cowen to reposition himself over time from a producer of content to a connector and networker.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Tyler Cowen

“All inputs other than the AI rise in marginal value. Since I'm an input other than the AI, or I hope, that means I rise in marginal value, but I need to do different things.”

0.57

In evaluating grant or job applicants, much of the signal comes within the first few minutes from intangibles: often a single question the person can't answer (e.g. nonprofit applicants unable to explain how they'd build a donor base), or wrong priorities and status markers (e.g. wanting an award only for immigration status); a transcript alone would be worth only about 25% of the value of the live video interaction.

factualhigh valuespeaker onlynovelty 3/4durability 3/4· Tyler Cowen

“How do you think about building out your donor base?' it's remarkable to me how many people have no idea how to answer that. Without that, you don't have anything.”

0.57

Different geographic intellectual cultures err in opposite directions: the Bay Area thinks in terms of infinities (creative and destructive) and overvalues intelligence; Washington DC thinks at the margin and is wiser overall but would end growth if everyone thought that way; the EU is wise, cultured, and tasteful but would produce -1% growth—so a healthy society like the US balances these tendencies, and the UK's balance has recently gone out of whack with too many non-growth-oriented people.

factualhigh valuespeaker onlynovelty 3/4durability 3/4· Tyler Cowen

“You come pretty close to thinking in terms of infinities, on the creative side and the destructive side. And no one in Washington thinks in terms of infinities; they think at the margin. Overall, I think they're much wiser than the people here.”

0.57

The consistent epistemic move is to trust the relevant experts in each domain: AI experts (who see the promise of strong AI, against asleep economists) and technology-diffusion experts (who universally find diffusion slow with no rapid takeoff); combining both expert views yields the conclusion that AI is amazing in the long run but will be a slow, bottleneck-ridden slog.

normativehigh valuespeaker onlynovelty 3/4durability 3/4· Tyler Cowen

“The people who study diffusion of new technologies never think there will be rapid takeoff.”

0.55

AI models have progressed faster than Cowen expected; the o1 model is very impressive, his economics test questions designed for GPT-4 became too easy, and he would not be shocked if an AI model beats human expert economists on hard questions on a regular basis in less than three years.

forecasthigh valuecontestednovelty 2/4durability 1/4· Tyler Cowen

“I would not be shocked if somebody's AI model, in less than three years, beat human experts on a regular basis.”

0.52

Progress studies should never become a formal, trademarked, or centrally directed movement but rather a decentralized change of ethos and vibe; on that gentler, more enduring trajectory a small number of talented people working on neglected areas like science policy can get a third to half of what they want and have a huge impact.

normativehigh valuespeaker onlynovelty 2/4durability 3/4· Tyler Cowen

“It would be people doing things in a very decentralized way that would reflect a general change of ethos and vibe. So, I hope it has, in many ways, a gentler but more enduring trajectory.”

0.31

Cowen distinguishes between being genuinely well-cultured (like the cultured French Brussels official) and his own 'barbarian's approach to culture'—a veneer built from collecting a lot of information, which can be mistaken for being cultured but is fundamentally different, an artistic rather than organic relationship to culture.

definitionspeaker onlynovelty 2/4durability 2/4· Tyler Cowen

“I have a veneer of culture that comes from having collected a lot of information. So I'll know more about culture than a lot of people, and that can be mistaken for being well-cultured, but it's really quite different. It's like a barbarian's approach to culture.”

0.26

The optimistic counterpoint is that human agency matters: people sitting in the year 1000 could never have imagined today's world, and it is up to us to take this extraordinary inherited heritage and do more with it.

normativespeaker onlynovelty 1/4durability 2/4· Tyler Cowen

“Human agency matters. If we were all sitting around in the year 1000, we never could have imagined the world being anything like this, even a much poorer country.”

0.10

Cowen states he is happy to answer if given a specific country, name, and time period regarding awful early-20th-century leaders.

factualspeaker onlynovelty 0/4durability 0/4· Tyler Cowen

“Well, give me like a country and a name and a time period, and I'll try to answer it.”