Richard Socher on AI conducting research: training and GPU kernel gains at Recursive
TWIML AI Podcast · rss · 2026-11-04
TWIML podcast with Richard Socher (ex-Salesforce chief scientist, You.com founder) on moving from AI assisting researchers to AI conducting research. Covers Recursive's reported gains in model training and GPU kernel optimization, why fast verification matters, how reward hacking complicates the search for real improvements, the data/experiment infrastructure needed for biology/chemistry/physics discovery, alignment under commercial incentives, and his 'Eureka machine' vision.
More from AGI Musings
- SAIL AI Challenge opens final phase: turn rare disease diagnoses into treatment reports — zakkohane · 2026-11-04
- Startup pitches 'monitor your agents so devs don't commit AI crimes' — basedjensen · 2026-11-04
- Primordiant launches general-purpose robot scientists to run experiments at scale — ycombinator · 2026-11-04
- Even with a 1000x-smarter superintelligence, people will still scroll Reels — jaivin · 2026-11-04
- Terence Tao: "AIs are much better at cheating" and finding model exploits — rohanpaul_ai · 2026-11-04
- Pain vs suffering may decouple — and differ across species, argues author — JoshPurtell · 2026-11-04