EconTalk · Russ Roberts

AI, Employment, and Education (with Tyler Cowen)

March 30, 2026·1 hr 2 min·4 clips

Tyler Cowen proposes devoting a third of all college course time to using AI, arguing you can teach Homer's Odyssey and AI simultaneously without losing any content.

1. EconTalk host Russ Roberts interviews Tyler Cowen, economist at George Mason University and founder of Marginal Revolution, on his February 2026 views about AI's effects on employment and education. 2. This is Tyler Cowen's 20th appearance on EconTalk; his talk at the University of Austin (not UT Austin) directly inspired the episode's agenda. 3. The episode's core question is whether AI will cause mass unemployment and economic collapse, or whether it will create new jobs and wealth, and how higher education should respond. 4. Cowen dismisses the current wave of social-media doom about AI job losses, citing energy sector growth, biomedical trial management, and government lawyers needed to write AI legislation as near-term employment growth areas. 5. He uses the Industrial Revolution campfire thought experiment — noting that no one in 1820 could predict 'podcasters' as a future occupation — to argue that forecasting which specific new jobs AI will create is structurally impossible. 6. Cowen predicts more billionaires and mega-billionaires as small teams build large-revenue companies, but argues this creates new projects and new jobs rather than concentrating wealth without employment. 7. His biggest worry about AI is not unemployment but governance: AI may worsen political decision-making in ways that subsequently harm economies, and political science lacks the modeling tools economics has to reason through these dynamics. 8. He raises a second concern about the political displacement of upper-middle-class white-collar workers — consulting partners earning $1.4 million annually who might be pushed into energy-sector roles at $300,000 — describing them as politically influential in ways that could produce dangerous electoral consequences. 9. On driverless cars, Cowen notes that trucking jobs encompass much more than driving — loading, unloading, managing contact points — and that the full displacement will be slow, giving institutions time to adapt. 10. Cowen's central higher-education proposal is that roughly one-third of college course time should be devoted to teaching AI usage, arguing that teaching AI for Homer's Odyssey simultaneously teaches Homer's Odyssey at a higher standard — not a trade-off but a bundled skill. 11. He acknowledges the structural problem immediately: faculty do not know how to teach AI usage, and administrators know even less, suggesting students may need to lead the pedagogy. 12. On AI cheating, Cowen frames it as a willingness-to-enforce problem: a sampling approach that locks students in a room for 2-3% of their submission load would provide enough statistical signal to identify gross divergence, without requiring AI detection software or expulsion. 13. Russ Roberts pushes back on what 'cheating' even means — noting AI assistance exists on a spectrum from 'have it reorganize my draft' to 'write the whole thing' — and Cowen responds by proposing splitting courses into two tracks: one requiring AI use where the joint product is graded, one where writing is assessed as a proxy for thinking. 14. Cowen's experimental pilot proposal is that each college allow students to take one AI-taught class per year on a topic the institution doesn't offer, with a human coach who structures the experience and a second AI that evaluates learning at the end — invoking Adam Smith's competitive education principle. 15. He describes generating a Tudor England syllabus, quizzes, and grading rubric live during the University of Austin talk, and notes that seven or eight students in the room wanted that class, which the institution does not offer. 16. On the more radical end, Cowen proposes a college model where students design their own major entirely with AI tutoring, human coaches, and peer interaction — a 15-week one-on-one tutoring relationship with an AI that a second AI then evaluates. 17. Cowen notes that writing and thinking are deeply entangled — Roberts cannot think without writing, Cowen agrees writing is a path to thinking — raising the question of whether AI-assisted writing severs that developmental link. 18. Both Roberts and Cowen agree that some educational experiences, particularly group reading of texts like the Iliad with veterans who have combat experience, cannot be replicated by an AI, and that group seminars can be incorporated into AI-taught courses. 19. The episode's tone is collaborative and intellectually playful — both participants clearly enjoy the conversation; Roberts frequently challenges Cowen's proposals to sharpen them. 20. Economists, educators, and anyone planning curricula or career trajectories in a world of rapid AI adoption will find the episode most useful; listeners seeking empirical research on AI labor-market effects will find the discussion more speculative than data-driven.

As heard by us

AI becomes a classroom tool, sharpening the question of whether writing is a skill or a way to think.

Tyler Cowen and Russ Roberts treat AI in the classroom as a practical test of what education is meant to do. Students can now use machines to shape prose, but writing still matters because it teaches thinking.

Read the full review in PlayNext →

Why you'd press play

If you want an AI conversation that reaches the classroom, where writing becomes a way to think.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from EconTalk