Math's loudest AI skeptic Daniel Litt now expects to lose his 2030 bet

ziv_ravid · x · 2026-09-16

The Information Bottleneck podcast will host Daniel Litt (@littmath, University of Toronto), one of the loudest skeptics of AI for math. In 2025 he bet AI wouldn't write top-tier math papers by 2030 — now he says he expects to lose.

His new essay "A beginning for mathematics" argues that once AI can produce a strong result for the price of a nice dinner, mathematics must stop rewarding papers and start rewarding real understanding. The interview will cover what that means for PhD students and why anyone would still want to become a mathematician.

Original post →

More from AGI Musings

AGI Musings channel →