Ben Todd argues AI task substitutability, not price, decides whether automation lifts GDP
ben_j_todd · x · 2026-09-30
Ben Todd and Herbie Bradley debate AI's GDP impact. Todd argues points like price-drop skepticism only delay rather than prevent the growth effect, and that task substitutability (sigma) is the contested key parameter: Epoch's paper calls sigma=0.5 a 'pessimistic lower bound', with higher substitutability unlocking far more knowledge work—scientific research, hyper-personalized expert services, virtual-only firms. Bradley counters that a task getting 1000x cheaper needs >1000x induced demand to grow its GDP share, though diffusion, not price, is the real bottleneck.
Related event: Researchers debate whether AI automation can boost GDP(3 posts)→
More from AGI Musings
- Economist finds LLMs 'so bad' at generating referee reports, tries translating as workaround — paulnovosad · 2026-09-30
- Investor Matt Turck jokes he now asks AI about everything from diet to parenting — mattturck · 2026-09-30
- Economist Paul Novosad decries LLM-written peer reviews as pushing academia to a worse equilibrium — paulnovosad · 2026-09-30
- Every: Automation Creates More Work for Human Experts — every · 2026-09-30
- Williams philosophy professor: writing isn't as important as we think in the AI era — joshgans · 2026-09-30
- Zvi mocks AI firms rebranding AI as cute: how would this slide play out in a movie? — TheZvi · 2026-09-30