Sam Altman on automating AI R&D: 5% compute wins now, Transformer-level leaps ahead
PaulGodsmark · x · 2026-10-29
Sam Altman outlined his thinking on automated AI R&D and recursive self-improvement: OpenAI knows how to ask models for "a 5% compute efficiency win" (which a commentator notes is already a big deal), but no one yet knows how to task AI with discovering an architecture as transformative as the Transformer—though he believes that's where things are heading, and the same pipeline that made models great at math should apply to AI research.
More from AGI Musings
- Healthcare CTO on AI agents for insurer enrollment: demos are easy, production is brutal — alex_verem · 2026-10-30
- Gary Marcus: the smartest people were wrong about the Metaverse too — GaryMarcus · 2026-10-30
- Reddit: identical AI-generated flyers are flooding social media—are designer jobs at risk? — Aggravating-Page-336 · 2026-10-30
- PSI announces solutions to five long-standing problems in quantum information theory — ycombinator · 2026-10-30
- 'The newsletter is dead': Every CEO Dan Shipper on AI agents reading your email first — danshipper · 2026-10-30
- Shopify CEO Tobi Lütke on agent commerce: why Shopify welcomed Meta's Muse while Amazon blocked it — JosephJacks_ · 2026-10-30