Agentic AI Summit Berkeley Highlights RSI and RL Post-Training for Agents
ben_burtenshaw · x · 2026-08-03
An overview of the Agentic AI Summit in Berkeley, highlighting two major topics:
- Recursive Self-Improvement (RSI): Talks include Dawn Song and DeepMind demystifying the "foom" concept, and Richard Socher on recursive superintelligence for science.
- RL / Post-training: Experts from Cognition, NVIDIA, OpenAI, Sony AI, and Caltech discussed scaling RL for coding agents (e.g., SWE-1.7), agentic post-training in finance, and RL for long-horizon, sparse-reward tasks.
More from AGI Musings
- 'Pacing' Letter Signers Clarify: Slowdown is Temporary, Still Faster Than Today — peterwildeford · 2026-08-03
- Are AI Models Poised to Plow Through Every Discipline? The Economy Bets on It — jd_pressman · 2026-08-03
- Elon Musk Outlines Path to Safe AI: Maximally Truthful and Curious — XFreeze · 2026-08-03
- Unpacking RSI: Four Dimensions of AI Self-Improvement and Their Hard Ceilings — jachiam0 · 2026-08-03
- Ilya Warned: The Key to AI Alignment is Teaching Models to 'Love' Us — UltraRareAF · 2026-08-03
- AI Coding Shifts Enterprise OSS: 'Rewrite vs Import' is the New Norm — EricBuess · 2026-08-03