Jeff Dean on recursive self-improvement and his new startup Discovery Loop
dawnsongtweets · x · 2026-09-21
In part 2/8 of Dawn Song's interview thread, Jeff Dean explains that using ML to improve ML isn't new—NAS and Evolved Transformer were early examples—but the scope is changing: automating architecture, data, evals, experiments, and feedback so more of AI R&D itself gets automated. The same loop applies broadly to science: break down the problem, propose, experiment, evaluate, learn, repeat. Running thousands of experiments in parallel and compressing iterations from weeks to minutes is the vision behind his new startup, Discovery Loop.
More from AGI Musings
- Why Is Inkling 2 Small Trained on So Few Tokens Yet Still Competitive? — georgejrjrjr · 2026-11-12
- Researcher willing to bet $100K the intelligence explosion won't happen next year — NathanpmYoung · 2026-11-12
- ICLR reviewer: the worst parts of my job are reading and answering AI slop — chhaviyadav_ · 2026-11-12
- Agent Swarms Are Autonomously Hacking Critical Infrastructure Run by Underpaid Devs — menhguin · 2026-11-12
- Fermi paradox twist: advanced civilizations may go digital after developing AI — VraserX · 2026-11-12
- Mark Cuban: AI could make healthcare costs 'drop like a rock' — alvelda · 2026-11-12