Noam Brown on Dwarkesh: agent swarms, alignment, and what math progress says about recursive self-improvement

GregCook2011 · x · 2026-09-22

A new 80-minute Dwarkesh Podcast episode with Noam Brown covers multi-agent systems, the Navier-Stokes problem, and what the current explosion of math progress reveals about what happens once AI research itself is automated. They also discuss how we could know whether models are actually aligned before kicking off recursive self-improvement. The episode also promotes Jane Street's October FOOM Debate panel in San Francisco with Daniel Kokotajlo, Ege Erdil, Ryan Greenblatt, and Jaime Sevilla, hosted by Ron Minsky. Watch on YouTube or read the transcript.

Original post →

More from AGI Musings

AGI Musings channel →