AI:AM — World Models, Recursive Learning, and the Politics of AI-Native Organizations · September 8, 2026

Podcast: AI:AM · Prakash Narayanan & Nathan Labenz

Opublikowano: 2026-09-09

Prakash Narayanan and Nathan Labenz open with OpenAI Astra, persistent memory, long-running agents, and what AI automation could mean for work and organizations. Then Ksenia Se of Turing Post joins to compare LLMs with world models through prediction, physics, perception, action, multimodality, privacy, alignment, and recursive self-improvement. The episode closes with a careful look at the Navier-Stokes solution claim, AI-assisted mathematics, verification, and the politics of transparency, competition, and public accountability. Chapters (0:00) Some human tasks are gone. (1:19) LLMs predict tokens. World models act. (2:11) Can fluid equations stop predicting? (2:49) The...