Dwarkesh PodcastTyler Cowen — Hayek, Keynes, & Smith on AI, animal spirits, anarchy, & growth
CHAPTERS
- 0:00 – 5:26
Keynes as the all-around economist: animal spirits, investing, and market psychology
Cowen and Patel open with Keynes’ self-conception as a rare “complete” economist—mathematician, historian, statesman, philosopher—and use it to frame Keynes’ biggest ideas. They dig into animal spirits, why investment is often driven by optimism rather than calculation, and how that shapes real-world returns in entrepreneurship and venture capital.
- •Keynes’ view of the master economist and why it plausibly described Keynes himself
- •Animal spirits as a driver of investment booms despite frequent disappointment
- •Skewed return distributions: a few winners vs many “projectors” who do poorly
- •How Keynes’ intuitions relate to VC, M&A, and entrepreneurial risk-taking today
- 5:26 – 12:36
Speculation vs long-term investment: efficient markets, passive capital, and finance’s true size
The conversation turns to Keynes on speculation: it may be easier to profit by predicting crowd behavior than by making true long-term forecasts about fundamentals. They discuss data limits in early 20th-century markets, whether passive investing is dangerous, and how to interpret the size of finance (wealth vs GDP).
- •Could Keynes-era markets have been inefficient enough to allow consistent outperformance?
- •Why historical EMH testing is hard: poor data, spreads, execution prices, low frequency
- •Passive investing concerns: monitoring incentives vs common-ownership/implicit collusion
- •Finance share of GDP vs finance share of wealth as the more meaningful metric
- 12:36 – 15:11
Risk preferences are context-dependent: Friedman–Savage, mood management, and “lottery tickets”
Patel challenges the tension between “humans are risk-averse” and Keynes’ “risk-seeking” animal spirits. Cowen invokes Friedman–Savage: people insure and gamble depending on context, and risk-taking can function as mood management. Cowen also describes his own non-pecuniary “bets” (e.g., publishing experiments).
- •Risk aversion vs risk seeking isn’t global—people switch by context
- •Insurance and gambling coexist as a behavioral regularity
- •Risk-taking as boredom relief / mood management rather than pure irrationality
- •Cowen’s examples: unusual projects as personal “lottery tickets”
- 15:11 – 18:12
How to judge “GOAT” economists: weighting peak work, failures, and quantity
They step back to methodology: should greatness be judged by a few top contributions or by an entire body of work? Cowen argues failures reveal thinking style and limitations, and that volume can be an ingredient in breadth (especially for thinkers like Mill).
- •Top contributions matter most, but failures are diagnostically important
- •Hayek’s drop-off and what it signals about rigor and normativity
- •Keynes’ frequent mind-changes as strength and weakness
- •Why writing a lot across domains can create multi-perspective insight
- 18:12 – 23:00
Controversy and social tolerance: Keynes, Turing, and how public punishment works
Patel asks why Keynes avoided career-destroying scandal despite his sexuality in early 20th-century Britain, contrasting him with Alan Turing. Cowen suggests “selective tolerance” existed and that personal diplomacy, power, and context mattered—paralleling modern cancellation dynamics.
- •Keynes’ political power and social skill as protective factors
- •Selective tolerance in elite circles vs harsh punishment for others
- •Turing’s vulnerability: diplomacy, security context, and incomplete historical understanding
- •Modern analogy: why some people survive taboos and others are sanctioned
- 23:00 – 25:02
Hayek after Road to Serfdom: collectivist instincts, bureaucracy, and what the real problem is
Shifting to Hayek, Patel notes Hayek’s “white pill” legacy as collectivisms collapsed, while Cowen argues Hayek became too pessimistic later. They debate whether modern cultural trends vindicate Hayek’s envy/collectivism story, versus Cowen’s emphasis on bureaucracy and institutional sclerosis.
- •Hayek’s later pessimism: built-in collectivist instincts and envy
- •Whether ‘wokeness’ supports Hayek vs Cowen’s claim it’s not the main issue
- •Bigger modern problems: bureaucracy, kludge-ocracy, sclerotic institutions
- •Hayek’s model vs contemporary political dysfunction
- 25:02 – 32:25
Planning vs markets: firms, the Soviet ‘decentralization,’ and competition as discovery
Patel presses whether big tech firms show that central planning can work; Cowen rejects the firm/market dichotomy and reframes firms as market-embedded contracting structures. They discuss why Soviet outcomes weren’t “true planning” for long, why urbanization drove growth, and highlight Hayek’s ‘competition as discovery’ as a core insight (including its tensions with scientism).
- •Cowen’s anti-Coase stance: firms are part of the market, not ‘islands of planning’
- •Soviet Union as decentralized rent-seeking with bad incentives, backed by party power
- •Why Soviet growth happened: reconstruction and especially urbanization catch-up
- •Hayek’s ‘competition as a discovery procedure’ and links to innovation/monopoly debates
- 32:25 – 47:41
Supply chains, tacit knowledge, and AI agents building markets (Bitcoin/NFT-like property rights)
They explore fragility in decentralized systems: supply chains can fail when prices don’t reveal infra-marginal values, as seen in crises like COVID. Then they pivot to AI agents: Cowen predicts they will spontaneously recreate market institutions—currencies and property rights—likely starting from crypto primitives, and that human productivity will bifurcate based on delegation skill.
- •Why prices can fail in crises: missing infra-marginal value signals when access drops to zero
- •Transparency/legibility isn’t always good—corruptibility of a single ‘controlling mind’
- •AI agents as market participants: emergent currencies and proto-property rights
- •Human comparative advantage shifts: who can delegate effectively to AI assistants
- 47:41 – 52:41
John Stuart Mill: women, children, and the anti-Hayekian case for rational moral progress
Mill’s ‘Subjection of Women’ is used as a case study where universal social practice masked contingent oppression—and where rational unpacking beat tradition. They extend Mill’s logic to children’s rights and education, discuss Mill’s own strict upbringing, and weigh whether modern democracies can still channel high-quality advice even if great thinkers aren’t elected.
- •Mill’s argument: observed universals can be contingent products of force, not wisdom
- •Applying Mill’s lens to children and coercion in schooling
- •Mill as both Hayekian (free speech) and anti-Hayekian (rationalist critique of tradition)
- •Political quality: Mill in Parliament vs today’s advisory ‘through the cracks’ model
- 52:41 – 58:30
Adam Smith on growth and measurement: from near-stagnation to AI-era bottlenecks
Smith’s brilliance is framed as extrapolating massive potential from modest observed growth, anticipating increasing returns and division of labor. They debate whether AI could produce sustained 10% growth, and Cowen argues growth statistics lose meaning as the consumption basket changes—illustrated by catalog-era comparisons and index-number problems across eras.
- •Smith’s rationalist extrapolation: small local growth but big theoretical possibility
- •Why 10% growth for decades may be conceptually misleading as bundles transform
- •Index-number ambiguity: why ‘X% richer than Romans’ can be nonsensical
- •Bottlenecks: law, institutions, and human constraints as checks on AI-driven growth
- 58:30 – 1:08:07
Other GOAT contenders and where big ideas live now: George, Coase, Schelling, Arrow, internet writing
Patel proposes additional GOAT candidates (Henry George, Coase, Schelling, Arrow), and Cowen ranks them highly but outside the very top—often due to limited breadth/quantity. They then argue that today’s ‘big ideas’ migrate to multidisciplinary internet writing, creating a new meta-discipline distinct from academic specialization.
- •Henry George’s revival via YIMBY/NIMBY politics; humane moral force and renewed relevance
- •Coase’s foundational genres despite disagreements; why lack of macro matters for GOAT status
- •Schelling’s role making game theory intuitive; Arrow’s underestimated importance in finance theory
- •Internet writing as the new carrier of grand syntheses, blurring disciplinary boundaries
- 1:08:07 – 1:13:16
Anarchy, cartel dynamics, and networked enforcement: social media, banking rails, and ‘anarchy is everywhere’
Cowen revisits his arguments about anarchy and protection agencies: network industries can evolve collusive equilibria, especially around vulnerable nodes like payment clearing. Applied to social media, he sees partial coordination (deplatforming cascades) but emphasizes banking/payment networks as the deeper risk; he also reframes the debate by claiming anarchy exists within governments and internationally.
- •Soft cartel behavior in social media via inference/coordination rather than explicit collusion
- •Clearinghouses/payment systems as the key collusion-enabling chokepoints
- •Crypto/Web3 as waves of centralization and decentralization rather than one stable structure
- •Reframe: anarchy within federal layers and among nations; anarchies that persist develop collusion
- 1:13:16 – 1:23:18
Cheap WMDs, decentralization as destiny, and AI governance: hegemons, regulation cycles, and national security realism
They connect existential risk to ‘cheap energy’ making mass destruction accessible (the ‘$50,000 nuke’ scenario), yielding a fragile long-run trajectory. Cowen argues decentralization is an ancient, irreversible choice; therefore safety often comes from relatively benevolent hegemons rather than idealized global coordination, and governments tend to underreact until a headline incident then overreact.
- •Core worry: destructive capability gets cheap; civilization risks collapse into chaotic low-population equilibria
- •Decentralization as a foundational constraint on what governance can achieve
- •Policy implication: prefer ‘better’ hegemons leading AI over rivals; skepticism of naive cooperation stories
- •Regulatory pattern: little action until an SBF-like incident, then overreaction
- 1:23:18 – 1:42:21
Democracy’s critiques, technocratic islands, and AI scaling skepticism: legibility limits and diminishing returns
In the closing stretch, they evaluate critiques of democratic capitalism (drift to socialism, inequality, ‘last man’ decadence) and Cowen insists on concrete falsifiers like asset-price signals. They then address AI scaling: Cowen expects meaningful productivity gains but doubts unlimited ‘dial-turning’ intelligence, emphasizing possible illegibility of reality, legal/regulatory bottlenecks, and the importance of workflow integration over sheer model size.
- •Why some institutions (Fed, courts) can be high-functioning: norms, incentives, reputations, civic virtue loops
- •Handling ideological critiques with concrete indicators (e.g., what would show up in markets/VIX?)
- •Scaling debate: technical scaling may continue, but epistemic bottlenecks and illegible domains may cap returns
- •AI impact forecast: end of stagnation and 20th-century-like acceleration, not sci-fi singularity
