Schmidhuber gives Oxford talk on recursive self-improvement, a theme he traces to 1987
mmbronstein · x · 2026-11-13
Deep learning pioneer Jürgen Schmidhuber delivered a talk at Oxford centered on his long-standing "Recursive Self-Improvement" framework, with slides reading "RCS since 1987." He referenced his 1987 meta-learning RNN work ("linear time, fully parallelizable") and argued that AI-generated AI keeps getting more complex, tying it to his Gödel-machine line of research.
More from AGI Musings
- Gary Marcus declares November the beginning of the end of the AI bubble — GaryMarcus · 2026-11-13
- Beyond p(Doom): Dan Faggella proposes a 'cosmic' p(Bloom) framing for AGI — danfaggella · 2026-11-13
- After Humans Fade, Which AGI Future Is 'Better'? A Three-Way Debate — danfaggella · 2026-11-13
- AI Agents Are the New AI Companions — and May Hit the Same Retention Wall — vaibhavbetter · 2026-11-13
- AIFlow_ML claims centralized AI is data harvesting, questioning SF labs' funding — AIFlow_ML · 2026-11-13
- Reddit Post: AI Chats Feel Like Lucid Dreams and May Breed a 'Genius Illusion' — ResonantFork · 2026-11-13