Dwarkesh podcast with Noam Brown: math progress, multi-agent systems, and verifying alignment before RSI
morqon · x · 2026-09-18
Dwarkesh Patel releases a new episode with Noam Brown covering multi-agent systems, the Navier-Stokes problem, and what the current explosion of mathematical progress tells us about automating AI research. Key threads: how AI firms will be structured, what math breakthroughs imply for recursive self-improvement, the growing internal/external model gap, and why chain-of-thought is degrading. The episode closes on the central question of how we would actually know if models are aligned before kicking off RSI.
Related event: Noam Brown on Dwarkesh: 10,000-Agent Swarm Solves Navier-Stokes(7 posts)→
More from AGI Musings
- Andrew Ng Pushes Back on AI Extinction Fears: 'More Science Fiction Than Science' — rohanpaul_ai · 2026-09-18
- As models improve, scaffolding falls away: Axolotl moves integrations out of core — MoonL88537 · 2026-09-18
- LLMs can map known unknowns, but don't ask them for evidence where none exists — StewartalsopIII · 2026-09-18
- Clive Wearing insists he's dead despite being awake — a consciousness case AI circles cite — yeastsplainer · 2026-09-18
- Is maximizing a simple goal really easier? Rethinking paperclip rationality as an optimization problem — jessi_cata · 2026-09-18
- AI agents killed flow state — but faster iterations may bring it back — cocktailpeanut · 2026-09-18