AI Summary
5 min readAlex Imas, a University of Chicago professor of economics and applied AI, joins Odd Lots hosts Tracy Alloway and Joe Weisenthal to explain why AI's labor market effects might differ from past technologies. Drawing from his research and experiments, Imas critiques simplistic exposure metrics and highlights overlooked economic mechanisms like task interdependencies and demand responses.
Economists' Views on AI's Impact
Economists studying AI, per a recent survey by Kevin Bryan, Basil Halperin, and others, expect substantial capability gains but moderate effects: extra 2-3% productivity growth by 2030-2050, with limited unemployment. This aligns technologists' views, surprising Imas. Surveys rank knowledge work as exposed, but Imas notes these measure only if AI handles 50% of a task—not full replacement, and jobs comprise multiple tasks.
Past technologies like the player piano displaced jobs but created others; AI's generality—from essays to forecasts—sets it apart from narrow tools like Go-playing AI. Pre-LLM AI targeted specifics; large language models (LLMs) enable agents that act (e.g., building spreadsheets), shifting economics.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
What you'll learn
- 1 (01:53) **Intro Skepticism on AI Job Impacts** - Hosts discuss economists' historical analogies for AI's labor effects and frustration over vague new job predictions
- 2 (04:38) **Guest Introduction: Alex Imas** - Chicago professor on economics and applied AI shares early ChatGPT realization
- 3 (06:42) **AI Generality and AGI Origins** - Explains leap from narrow tools to general intelligence, coining of AGI term
- 4 (09:32) **Economists' Modal AI Forecasts** - Survey shows alignment between economists and technologists on moderate growth by 2030-2050
- 5 (11:36) **Limits of Job Exposure Measures** - "GPTs are GPTs" paper maps tasks AI does 50% of, but ignores full job context
- 6 (15:03) **Task Complementarity Explained** - Interrelated tasks mean weak links fail whole job (e.g., bad seasoning ruins meal)
- 7 (20:28) **Unemployment Scenarios** - Full automation, inelastic demand, or one-task jobs could cause job loss
+ Full timestamped outline available in the app
Show Notes
Everyone knows that new technologies can be really disruptive to the labor market, but eventually new jobs emerge and things come back into balance. And there is a sense in which many view AI with the same lens. Yes, there will be pain in some sectors, but then there will be productivity gains and new sources of demand and new opportunities for labor that we can't conceive of yet. But could it be different this time? Could AI be disruptive in a manner that, say, the steam engine was not? On this episode we speak with Alex Imas, a professor at the University of Chicago focusing on economics and applied AI. We talk about his work on the AI and labor question, how to think about which jobs may be most at risk, and why the sheer speed of AI development could make it categorically different than prior general purpose technologies that came before it.
Subscribe to the Odd Lots Newsletter
Join the conversation: discord.gg/oddlots
See omnystudio.com/listener for privacy information.
More from this podcast
Odd Lots →