Daron Acemoglu on Liberalism, Automation, and the Educated Elite

Episode Summary

Executive Summary: Daron Acemoglu argues that liberal democracy must be rebuilt around meaningful individual freedom, community consensus, and a more pro-worker political economy rather than technocracy or contractarian abstractions. He criticizes elite-driven governance, warns that automation and AI can centralize power and depress labor’s role unless paired with new task creation, and defends free speech while favoring social-media regulation. He is cautiously optimistic about AI, immigration, and longevity, but only if institutions and incentives are redesigned.

Main Topics: Rebuilding liberalism beyond social contract theories (Priority: 5/5): Acemoglu says social contract and contractarian theories dodge the hard problem of how societies build consensus around shared moral values and priorities. He prefers a liberalism grounded in individual freedom plus community-level agreement, not an imposed philosophical 'book' of ultimate principles. Critique of elite technocracy and working-class liberalism (Priority: 5/5): He argues modern liberalism has become overly dependent on educated elites and hierarchy, undermining broad freedom and social legitimacy. 'Working-class liberalism' means political and economic arrangements that empower people who live from labor, not just capital or inherited status, while still respecting expertise. Automation, new tasks, and the labor share (Priority: 5/5): He distinguishes automation from broader technological change. Automation can reduce labor share and harm specific worker groups, especially in manual jobs, but shared prosperity depends on balancing automation with the creation of new tasks and opportunities. AI: productivity gains, decentralization, and pro-worker design (Priority: 5/5): Acemoglu is more optimistic about AI than his earlier estimates, but still expects moderate rather than transformative productivity growth. He worries about centralization, winner-take-all dynamics, and uses 'pro-worker AI' to mean tools that expand worker capability rather than just replacing labor. Free speech and social media regulation (Priority: 4/5): He claims to be a free-speech absolutist in principle, but says social media algorithms create a new problem because platforms weaponize speech through targeted amplification. He distinguishes unrestricted individual speech from algorithmic manipulation by powerful companies. Immigration, education, and public capacity (Priority: 4/5): He supports skilled immigration but worries policymakers use it as a substitute for educating Americans. His concern is not immigration itself but elite resignation about improving broad educational outcomes and reconstituting institutions like teacher unions. Demography, longevity, and Armenia (Priority: 3/5): He is less alarmed by fertility decline than many economists because induced innovation and longer healthy lives may offset some demographic pressures. He also discusses Armenia’s constraints and opportunities, noting its human capital and geopolitical vulnerability.

Key Arguments: Liberalism should start from meaningful individual freedom, then allow societies to build consensus around shared priorities rather than assuming a universal moral contract. The crisis of liberal democracy is partly internal: left liberalism empowered educated elites too much and failed to maintain broad legitimacy. Non-interference is insufficient without non-domination and real opportunities; people need protection from concentrated power and the ability to act on their freedom. Automation is not synonymous with progress; it often reduces labor share at the firm and sector level unless matched with new task creation. AI should be judged by whether it complements workers and expands their capabilities, not simply by aggregate productivity growth or corporate profits. Social media is different from the printing press because algorithmic feeds exploit vulnerabilities at scale, making regulation of platforms compatible with free speech. Skilled immigration is positive, but it should not become an excuse for abandoning domestic education and social mobility. Demographic decline does not automatically imply stagnation if innovation is induced and people remain healthier and productive longer. AI’s biggest risk is centralization; more domain-specific, decentralized models may better serve liberal society and worker interests.

Data Points: Manufacturing employment share in the U.S.: 8% to 9% - Used to illustrate that the classic industrial working class is no longer the dominant social category. Private-sector union density: Very small - Mentioned as part of the decline of the traditional industrial working-class model. Population receiving labor income: Vast majority - Acemoglu defines the working class broadly as people who earn mainly from the labor market rather than capital or inheritance. Labor share example: 62 to 60 - Cowen cites a small labor-share decline to question whether automation is truly alarming at the macro level. Share of U.S. labor market supported by new tasks: 40% to 50% - Acemoglu says a substantial portion of current U.S. human work is related to new tasks created over time. Early AI productivity estimate window: Next 10 years - Acemoglu says his prior AI estimate was for the 2024-2034 period. Revised view on AI productivity: Upward revision, but not radically - He says current model progress would lead him to revise his 2023 estimate modestly upward. Probability of an alternative AI stack scenario: ~30% - He estimates a rough probability that chips, applications, or other layers could capture more value than foundation models. H-1B cap: 85,000 workers per year - Raised in discussion of skilled immigration and whether it substitutes for domestic education. Life expectancy aspiration: 100 - Acemoglu suggests a well-educated young person who avoids accidents may plausibly live to 100. Countries mentioned as experiencing fertility decline: Brazil 1.36 fertility rate (approx.) - Used to argue fertility decline is broadening beyond rich countries. Likelihood of AI productivity boost similar to late-1990s IT boom: Half a percentage point or more is imaginable - Cowen argues a productivity gain of that scale seems plausible; Acemoglu calls it amazing but still modest relative to hype.

Pivotal Quotes: "we need to build consensus around shared moral values and shared priorities" — Daron Acemoglu: Explaining his objection to social contract and contractarian theories as foundations for liberalism. "I think what brought liberalism's crisis, in two words, liberalism crisis is related to the rise of a post-industrial economy... and a philosophical shift that gave more and more priority and power to the educated elite" — Daron Acemoglu: Summarizing his diagnosis of why liberal democracy is in trouble. "I do not believe that we need to have a regulator that says, oh, only pro-worker AI is allowed. Absolutely not." — Daron Acemoglu: Clarifying that his pro-worker AI agenda is about incentives and norms, not direct state approval of every system.

Implications: The conversation points toward a liberalism that is less elitist, more decentralist, and more attentive to worker capability. For AI, the key issue is not just speed of innovation but who benefits, whether power concentrates, and whether institutions can adapt.