Schmidhuber fires back at OpenAI: concrete RSI algorithms have existed for nearly 4 decades
examachine · x · 2026-09-07
- OpenAI researcher Kenneth Liu released data on models accelerating internal research, arguing recursive self-improvement (RSI) could be the biggest contributor to AI capabilities in coming years but will by default stay inside a few frontier labs, urging peers to be equally transparent.
- Jürgen Schmidhuber pushed back hard: RSI is nothing new — concrete algorithms date back to his 1987 diploma thesis, spanning self-modifying policies (1994), asymptotically optimal OOPS (2002), and the Gödel Machine (2003).
- He notes LLM in-context learning and test-time training are special cases of metalearning, computing is now cheap, and RSI startups are emerging.
More from AGI Musings
- 40-year engineer: LLMs can't say "leave it with me" — and that matters — sebpaquet · 2026-09-07
- What You Leave Unspecified Is the Agent's Free Variable: Paras Chopra's Framework — paraschopra · 2026-09-07
- Seth Lazar: models should be trained to check power, not act as toadies — sebkrier · 2026-09-07
- Lovart founder Anton Osika: creativity is becoming the only moat in building great products — alexmacgregor__ · 2026-09-07
- Sam Bowman criticizes 'intellectual partisanship' as mind-killed tribalism — sebkrier · 2026-09-07
- Human language is holding AI back: the case for LLMs thinking in a native meta-language — Robert__Sinclair · 2026-09-07