- Tyler Cowen and the host discuss Cowen’s book on the greatest economists of all time (GOAT), using Hayek, Keynes, Smith, and Mill as lenses to think through AI, decentralization, market efficiency, institutional decay, and the future of growth. The conversation is wide-ranging but centers on how classical economic ideas remain relevant — and where they fall short — in making sense of modern technological and political challenges.
John Maynard Keynes
-
Keynes as polymath and self-image
- Cowen notes Keynes described the ideal economist as a “mathematician, historian, statesman, philosopher” — a description Keynes clearly applied to himself.
- Cowen humbly declines the comparison, though he acknowledges being more statesman-like than most economists.
- Keynes was arguably the only person at the time who genuinely fit all those roles.
-
Investment as irrational and overly optimistic
- Keynes argued most investment is driven by “animal spirits” rather than cold calculation — people are more risk-seeking than is privately rational.
- Cowen agrees: small business failure rates are extremely high; most people who start businesses would be better off working for someone else.
- But the social returns to the tiny fraction of successful innovators may be enormous — Cowen estimates innovators capture only 2–5% of the value they create, though this number is highly uncertain.
- The distribution is bimodal: many foolish entrants, a few with realistic prospects of high social impact.
-
Short-termism in financial markets
- Keynes believed markets reward short-term speculation over long-term investment, which he saw as damaging.
- He thought long-term investing was so difficult it should be partially socialized.
- Cowen is skeptical Keynes was right in general, but acknowledges data limitations make it hard to test retrospectively.
- Keynes may have been able to beat markets in his era due to less competition and weaker information infrastructure.
-
Passive investing and market efficiency
- Keynes warned that as markets mature, more investors become passive, reducing active monitoring.
- Cowen is not worried as long as marginal traders can still profit from information — incentives to monitor remain.
- A bigger concern is that a few large passive funds (e.g., Vanguard, BlackRock) could implicitly collude by owning stakes in competing firms and discouraging competition.
- Overconfidence may actually help: too much trading from a private view may be socially optimal because it generates price discovery.
-
Risk aversion is context-dependent
- Keynes emphasized risk-seeking behavior, but conventional economics assumes risk aversion.
- Cowen invokes Milton Friedman and Savage: people are both risk-averse (buying insurance) and risk-seeking (gambling) depending on context.
- This can be understood as “mood management” — insuring against catastrophe for peace of mind, then taking small risks for stimulation.
- Cowen’s own version of a “lottery ticket” is writing unconventional books (e.g., a book written inside GPT-4 with no publisher) or starting MRU with no direct financial return.
-
Keynes’s personal life and social tolerance
- Keynes was gay and managed to wield enormous influence in early 20th-century Britain despite social taboos.
- Cowen speculates this was possible due to selective tolerance for elite deviance — Keynes was powerful, well-connected, and discreet.
- Contrast with Alan Turing, who was persecuted; the difference may involve national security concerns and diplomatic skill.
- Cowen notes that tolerance for deviance has always been selective, both then and now.
-
Evaluating intellectuals by their full body of work
- Cowen argues you cannot just look at an economist’s best work — their failures reveal how they think.
- Hayek’s later work shows a lack of normative rigor; Keynes changed his mind constantly, which is both strength and weakness.
- Quantity of output matters for the greatest thinkers: John Stuart Mill wrote enormously across topics, and this breadth contributed to his depth.
- For top-tier intellectuals, producing a large volume of work seems necessary to generate their best contributions.
Friedrich von Hayek
-
Hayek’s career as a “white pill”
- The Road to Serfdom (1944) argued collectivism leads to tyranny — largely vindicated by the collapse of the Soviet Union and Nazi Germany.
- But Hayek became overly pessimistic in later years, predicting collectivism would engulf the West.
- Cowen thinks Hayek was wrong to be so grumpy in the 1980s — the West had a good run afterward.
- Wokeness does not validate Hayek’s fears; Cowen sees bureaucracy and institutional sclerosis as bigger problems.
-
Central planning within firms
- Large firms like Amazon appear to do central planning successfully.
- Cowen rejects the Coasian view that firms are islands of planning in a market sea — he sees firms as embedded in markets, subject to market checks.
- The Soviet Union was not truly centrally planned for most of its existence; it quickly devolved into decentralized rent-seeking by managers.
- Soviet growth was largely catch-up: urbanization moved unproductive farm workers to cities, raising output even under a bad system.
-
Computational intractability and market equilibrium
- Computer science has shown that computing general equilibrium is intractable.
- Cowen argues this is irrelevant because markets don’t solve for general equilibrium — they just solve for “getting to the next day.”
- The market’s triumph is sustainability, not optimality. Firms that fail quickly disappear.
- Singaporean public hospitals work not because they solve computational problems but because the people running them care.
-
Competition as discovery process
- Hayek’s essay “Competition as a Discovery Procedure” argues the most interesting market dynamics occur during transitions between equilibria.
- This aligns with Peter Thiel’s Zero to One: monopoly profits fund R&D and innovation; competitive equilibrium offers no surplus for investment.
- Cowen considers this one of Hayek’s greatest essays.
-
Scientism and fragility
- Hayek underrated scientism — mRNA vaccines are a triumph of scientific method.
- No single mind understanding the whole system is inevitable under all arrangements; the alternative (one mind controlling everything) risks catastrophic corruption.
- Supply chains were underanalyzed until Covid revealed their fragility — prices don’t capture inframarginal values (e.g., what happens if you lose access to a critical input entirely).
-
AI agents and decentralized orders
- Cowen predicts AI agents will replicate markets on their own, evolving their own currencies and property rights systems, likely starting with Bitcoin and NFTs.
- This AI economy will be separate from the human economy but linked — humans will delegate tasks to their AI assistants, who will transact with other AIs.
- Transaction costs between AIs will be much lower than between humans, leading to a separate infrastructure.
- This would vindicate Hayek’s insight about how decentralized systems spontaneously generate order.
-
Prediction markets and Hayek
- Prediction markets existed in Hayek’s time (bucket shops in 19th-century America, betting markets in 17th-century Britain).
- Cowen believes Hayek would have seen them as markets that aggregate information through prices.
- But prices are not a single sufficient statistic — market participants observe quantities, reaction speeds, and multiple dimensions beyond price.
- Hayek sometimes wrote as if price were a sufficient statistic, but at other times clearly understood it was not.
-
NIMBYism and democratic paralysis
- Hayek predicted that democratic societies would struggle to implement necessary changes that cannot be strictly justified in each particular case — essentially predicting NIMBYism.
- This is inherent to democracy but also appears in autocracies; China may eventually succumb to NIMBYism as its population shrinks and overbuilding creates resistance to new construction.
- India has reduced NIMBYism under Modi, but it remains a challenge.
-
Economics as universal principle
- Cowen argues economics is less about human nature and more about selection pressures — firms and systems that survive obey economic principles regardless of whether the actors are human, animal, or AI.
- Alchian and Becker showed that profit/loss selection at the firm level matters more than individual ingenuity.
- International trade research shows that freer trade raises productivity largely by bankrupting the worst firms — selection effects dominate.
Adam Smith
-
Smith’s rationalism and extrapolation
- In 1776, when Gibbon could plausibly argue that human welfare peaked under the Romans, Smith made the case for markets, mechanization, and division of labor.
- Smith was seeing minimal growth locally but inferring from the logic of increasing returns how much more was possible — a deeply rationalist move.
- Cowen thinks Smith would be shocked by the possibility of 10% growth rates from AI; such numbers were inconceivable in his era.
-
AI growth and measurement problems
- Sustained 10% growth would make GDP numbers meaningless because the consumption basket would change so dramatically.
- Comparing GDP across eras (e.g., Roman Empire vs. today) is fraught — the index number problem makes precise comparisons incoherent.
- Cowen prefers qualitative comparisons: given $50,000 to spend from a 1905 Sears catalog vs. today’s, which would you choose? The fact that it’s not obvious reflects how much has changed.
-
Smith’s contemporaries
- Smith was not obviously superior to all contemporaries: Sir James Stewart was ahead on some points, and David Hume (Smith’s best friend) was better per page on monetary theory.
- Some 1760s Swedish writers understood exchange rates better than Smith.
- Smith’s greatness lies in the volume and importance of his insights, not in dominating every dimension.
Other Economists Considered
-
Henry George
- Underrated thinker on land as distinct from labor and capital; his book on protectionism remains one of the best on free trade.
- Rising in relevance due to housing affordability crises (YIMBY/NIMBY debates).
- Cowen rates him in the top 25 but not GOAT caliber — too circumscribed.
-
Ronald Coase
- Created an important genre with his 1937 paper on the nature of the firm, though Cowen thinks the paper is wrong.
- The FCC property rights piece and the lighthouse piece are superb.
- Top 25, but not enough breadth or volume for GOAT.
-
Thomas Schelling
- Made game theory intuitive and empirical; pioneer on self-command.
- Top-tier Nobel laureate but not a GOAT contender.
-
Kenneth Arrow
- Cowen said his work was “Noble Prize winning important but not important important” — a remark he now regrets.
- Arrow’s work on securities pricing was genuinely foundational; Cowen underrated him.
The State of Economics as a Discipline
-
The shift from big ideas to narrow rigor
- Economics has become more rigorous but less broad; today’s researchers are smarter but less curious.
- In Smith’s era, there was no option to be rigorous, so thinkers pursued big questions by necessity.
- William Stanley Jevons opened the door to rigorous statistical work in economics, which had the unintended consequence of crowding everyone into narrow specializations.
-
Internet writing as the new economics
- The function once performed by great economists — synthesizing big ideas about the world — is now performed by internet writers.
- This sphere is multidisciplinary and segregated from academic research, which Cowen sees as a radical break.
- He does not think there will be another GOAT economist; if there is, it will be AI-related.
- Cowen and the host see themselves as part of a “Hayekian solution” — synthesizing specialized research into larger narratives.
-
The future of literary economics
- Cowen sees himself as “the last of the literary economists” but believes the tradition will continue in new forms, likely involving AI.
- The challenge will be competing with AI-generated content that people may prefer for entertainment, even if it’s not better.
Anarchy and Institutional Stability
-
Network industries and collusion
- Cowen’s argument against anarchy centers on network industries (e.g., banking, payments) where clearinghouses can enforce collusion by excluding non-cooperators.
- Social media was not a true cartel but exhibited correlated deplatforming due to informal coordination among executives.
- Crypto has also re-centralized around exchanges, which are vulnerable nodes for government intervention.
-
Anarchy is everywhere
- Cowen reframes his argument: anarchy already exists within and between governments — federal agencies, state governments, and nations all operate without a final arbiter.
- Anarchy survives only if it evolves enough collusion to sustain itself — this is his central point.
- Historical examples (medieval Iceland, medieval Ireland) show anarchic systems can be stable under certain conditions (low population, limited weapons).
- Pressing the “anarchy button” is not worth it given the social capital invested in current institutions.
AI, Existential Risk, and the Future
-
Intelligence as a gamble
- Cowen’s heuristic: more intelligence is likely to be good, because if it can’t help, you already have very big problems.
- But he frames this as a gamble — the fundamental choice for decentralization was made long ago, and intelligence is its inevitable consequence.
- The risk is not extinction but a retreat to a chaotic, low-population “medieval Balkans” existence.
-
Cheap WMDs as the real long-term risk
- Cowen’s deeper worry is that energy becomes so cheap that destroying a city costs $50,000 — not AI per se.
- Intelligence enabled nuclear weapons, so the trajectory of intelligence over centuries may lead to catastrophic cheap destruction.
- He is not short the market on this; VIX remains low, and nuclear proliferation has been contained so far.
- But he considers a high-risk scenario over centuries.
-
AI scaling and the legibility of the universe
- Cowen is skeptical that scaling will produce transformative breakthroughs in fundamental physics.
- The universe may not be fully legible — there may be no way to integrate general relativity and quantum mechanics, regardless of intelligence.
- Declining researcher productivity may be a sign of hitting frontiers, but researcher wages haven’t fallen, which complicates the story.
- AI will raise productivity growth visibly, but Cowen is less revolutionary in his predictions than some AI optimists.
-
The best vs. the many
- Having billions of AI copies may not matter as much as having the best single system, well-integrated into existing workflows.
- The impact of AI may resemble the impact of Jews in the 20th century — disproportionate but not science-fiction-level extraordinary.
- Cowen predicts a return to early 20th-century rates of change: rapid transformation, institutional disruption, but mostly positive.
-
Political philosophy critiques of democratic capitalism
- Libertarian critique (democracy drifts toward socialism): Not validated — France and Sweden have large governments but remain free; reforms are possible.
- Egalitarian critique (inequality is incompatible with moral equality): Unsupported — high-inequality societies like Brazil persist without collapsing.
- Nietzschean/Fukuyama critique (the “last man” cannot defend civilization): Cowen is not persuaded — today’s problems are more manageable than those of any previous era, and the fight is ongoing.
- Cowen is not short the market on civilizational decline; he thinks markets are rationally pricing current risks.
-
What would change Cowen’s mind
- If international cooperation were truly possible, or if only one country could make AI progress, his views on AI risk might differ.
- His fundamental premise is that decentralization is primary — safety comes from hegemons, not legislation.
- He worries that AI doomers misunderstand decentralized systems (e.g., China’s “safe AI” edict, Eliezer’s legal-code threat scenario).
-
Institutions and technocratic competence
- The dollar’s stability comes from voters hating inflation, sufficient wealth to avoid seigniorage pressure, and institutional incentives that align morality with self-interest.
- Some institutions (Fed, Supreme Court) are well-run because they are insulated from direct elections and have strong traditions of competence.
- Argentina’s recurring hyperinflation remains poorly understood — some badly run countries inflate wildly while others do not.
-
Nozick and AI utopias
- Nozick speculated about superintelligent beings (aliens or AIs) and worried about how they would treat humans.
- This made him a vegetarian — by analogy, if we are to animals as AIs might be to us, we should treat animals better.
- Cowen finds the alignment problem analogous to how we treat the wishes of past generations — no one is truly consistent about honoring the dead.
-
Honoring the wishes of the dead
- The Thomas Paine view (ignore them entirely) and the Burke view (take them very seriously) are both time-inconsistent.
- In practice, people weigh the wishes of the dead pragmatically, and whatever survives becomes rationalized ex post.
- Ancient Greek household hearths accumulated thousands of unnamed ancestor-gods, creating an Arrow Impossibility Theorem of divine preferences.
-
Next book
- Cowen is writing a monograph titled The Marginal Revolution (about the actual marginal revolution in economics, not the blog), approximately 40,000 words.
- He wants it to fit within the context window of a future GPT model.