Ask Me Anything | October 2026

Carroll argues AI's real near-term risk is human misuse of agentic systems causing medium-scale harm, not existential "superintelligence" doom scenarios.

• Navier-Stokes claim scrutinized: Carroll says OpenAI's claimed proof of finite-time blowup in 3D Navier-Stokes is causing "consternation" in math, with disputes over whether Lean verified the actual proof and whether training data from human mathematicians' chats was used.
• AI's edge is checkability: Carroll argues LLMs excel where tasks are well-defined and verifiable (math proofs, chess, Go, code) but much less so in physics, biology, or art, where there's no clear "win" condition.

Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.