Schmidhuber: I published the first concrete RSI algorithms in 1987, when compute cost 100M times more
SchmidhuberAI · x · 2026-10-10
Jürgen Schmidhuber weighs in on the current Recursive Self-Improvement (RSI) hype, noting he published the first concrete RSI algorithms back in 1987, when compute was roughly 100,000,000x more expensive (technical note IDSIA-9-26).
He recaps his earlier work: RSI with self-modifying policies since 1994, gradient-descent-based RSI in neural networks since 1992, asymptotically optimal RSI for curriculum learning since 2002, the mathematically optimal self-referential Gödel Machine since 2003, and RSI combined with artificial curiosity and intrinsic motivation since 1990/1997. His take: with compute now cheap, RSI is driving the future of both software and physical AI.
Related event: Schmidhuber claims recursive self-improvement dates back to his 1987 work(2 posts)→
More from AGI Musings
- AI memo: China to restrict exports of 23 AI technologies, plus Kimi K3.1 Pro and Genie 4 — dralandthompson · 2026-10-31
- Chinese firms offer AI 'ghost bots' that resurrect deceased loved ones from 30 seconds of footage — ArtificialOther · 2026-10-31
- SCOPE workshop proposed for ICLR 2027 asks what counts as an AI scientific discovery — mariyaivasileva · 2026-10-31
- ThursdAI crew debates: with Opus 7 coming, who still cares about better benchmarks? — thursdai_pod · 2026-10-31
- Peter Diamandis: The real AI divide is curious vs. afraid, not rich vs. poor — PeterDiamandis · 2026-10-31
- Chris Manning: Language Is the Real Unlock for Intelligence — stanfordnlp · 2026-10-31