Tyler Cowen & Alex Tabarrok on AI, Jobs, and Economic Growth
Economists Tyler Cowen and Alex Tabarrok argue that AI will not cause mass unemployment but will instead transform labor markets, create new jobs, and dramatically raise living standards, much like previous technological revolutions. They draw historical parallels to the Industrial Revolution, Ricardo's fears about machinery, and the Luddites to suggest that fears of job destruction consistently underestimate job creation. They express optimism about AI's potential to reduce extreme poverty, extend healthy lifespans, and solve major scientific challenges.
Summary
In this fireside chat hosted by OpenAI's Wyatt Thompson, economists Tyler Cowen and Alex Tabarrok offer a broadly optimistic assessment of AI's impact on labor markets and economic growth. Both economists begin by situating AI within the long history of technological disruption, arguing that fears of permanent job destruction have consistently proven wrong — from David Ricardo's 1817 essay 'On Machinery' to the Luddites smashing mechanical looms. They note that Excel didn't eliminate accountants, tractors didn't permanently displace farm laborers, and they expect AI to follow a similar pattern.
Cowen identifies several specific sectors he expects to grow due to AI: energy and electrical grid infrastructure, biomedical research and clinical trials, elderly care, 'messy jobs' requiring coordination and judgment (as described by economist Louis Garicano), government and AI regulation law, cybersecurity, compliance, niche entertainment, and highly idiosyncratic service jobs difficult to predict in advance. He maintains a blog series called 'New Service Sector Jobs' with over 150 entries documenting how wealth creation generates unforeseen employment categories.
Tabarrok reframes the 'mass unemployment' narrative by pointing out that 50% unemployment and a 50% reduction in the workweek are nearly the same thing — and historically, the latter has already happened. Annual work hours have fallen from roughly 3,000 in 1850 to about 1,500 today, and the share of a person's life spent working has dropped from 50% to 10%. He argues this trend is likely to continue, with AI enabling shorter work weeks, longer childhoods, more education, and bigger retirements rather than catastrophic unemployment.
On the question of income distribution, Cowen argues that people at the bottom will benefit most from deflationary pressures making services nearly free, while the biggest relative losers will be upper-upper-middle-class professionals — lawyers, consultants, finance workers — who relied on credential-based career paths. He acknowledges this will create social and political disruption but notes these individuals are unlikely to face poverty. Tabarrok adds that consumption inequality may matter less than income inequality because tools like AI are broadly accessible regardless of wealth.
Both economists dismiss the 'Citrini scenario' — the claim that AI will cause a new Great Depression by concentrating all purchasing power at the top — as economically incoherent, invoking Say's Law to explain that profits require spending customers and that money circulates through the economy.
On the question of what happens if AI becomes capable of doing everything better than humans, both economists invoke comparative advantage: even if AI is superior at all tasks, resource constraints (energy, time, capital, land) mean cooperation and trade remain beneficial. They also note that who owns capital and land matters enormously for distribution, and suggest that UBI or sovereign wealth funds could be tools for redistribution in a post-scarcity world, though they view that scenario as far in the future.
Cowen expresses concern about political risks from AI companies' dependence on energy and compute in a geopolitically unstable world. He also identifies the upper-middle-class disruption as one of his biggest political worries, since these are influential people whose relative status will decline even if their absolute living standards remain high.
On AI legal personhood, both economists find the idea of AIs having property rights and legal accountability plausible and necessary, with Tabarrok drawing a parallel to how corporations are treated as legal persons. Cowen raises concerns about 'rogue AIs' not tethered to any legal accountability, comparing them to wild animals, and suggests some form of technical firewall may be needed.
Finally, both economists express excitement about AI's potential in medical research, pointing to the enormous value of eliminating diseases like cancer, and urge OpenAI to push for unlocking medical data — specifically referencing Epic's comprehensive health records — as a transformative opportunity. They close by emphasizing that the world outside the U.S. and major tech hubs is vastly underserved by AI adoption, representing enormous opportunity for meaningful work in technology diffusion.
Key Insights
- Cowen argues that framing AI's impact as '50% unemployment' versus 'a 50% reduction in the work week' describes nearly the same outcome but with opposite emotional valence, and that history already shows work hours have fallen by half since 1850 without catastrophic consequences.
- Tabarrok points out that the Luddites were technically the first people to attack algorithmic automation — the Jacquard loom used punch cards — and that their predictions of permanent job destruction were empirically wrong, as weavers, farm laborers, and accountants were not eliminated by machines.
- Cowen argues that the biggest losers from AI will be upper-upper-middle-class professionals in law, consulting, and finance who expected credential-based paths to high salaries, while people at the economic bottom will benefit most from deflationary AI-driven prices on services.
- Tabarrok invokes Ricardo's theory of comparative advantage to argue that even if AI is better than humans at everything, resource constraints like energy, capital, and time mean trade and cooperation between humans and AI systems remain mutually beneficial.
- Cowen dismisses the 'Citrini scenario' — that AI will cause a Great Depression by concentrating purchasing power at the top — as economically impossible, arguing that high profits require spending consumers and that money circulates through the economy via Say's Law.
- Cowen identifies the political risk of AI companies' dependence on energy and compute in geopolitically unstable regions as one of his primary concerns, suggesting it is underappreciated relative to more commonly discussed AI risks.
- Tabarrok argues that consumption inequality under AI may be less severe than income inequality suggests, noting that AI tools are broadly accessible regardless of wealth — meaning billionaires do not have access to meaningfully better AI than ordinary users.
- Both economists argue that the most impactful thing AI companies can do for social benefit is to make models more capable — particularly for medical research — pointing to estimates that eliminating cancer would be worth approximately $80 trillion in human welfare value.
Topics
Full transcript available for MurmurCast members
Sign Up to Access