teortaxesTex: RSI in the mundane sense started with RLHF — Western labs mean no-human-in-the-loop
teortaxesTex · x · 2026-11-18
Replying to StephenPiment, teortaxesTex adds conceptual distinctions:
- "Recursive self-editing" is a dumb LessWrong idea dating to the Emacs/Elisp era, somewhat rekindled by overhyped recent direct weight-editing successes.
- "Iterative AI improvement via AI" instead passes through data and training stages; by the most mundane standard, we've been doing "RSI" since the first RLHF iteration or earlier.
- The real dividing line: what Western labs call RSI means "AI self-improvement without a human in the loop" — still very far from the current Chinese level.
Together with StephenPiment's post, this forms one discussion with two complementary angles.
Related event: Debate over RSI: Chinese labs practice it but reject fast-takeoff framing(6 posts)→
More from AGI Musings
- Field experiment: Google AI Overviews cut publisher clicks 39.8% with no UX gain — paulnovosad · 2026-11-18
- Academia Has the Strongest Anti-AI Norms, Yet Grad Students Use AI Anyway — CatAstro_Piyush · 2026-11-18
- "College and the 9-5 path is over" — AI era anxiety goes mainstream — rand_longevity · 2026-11-18
- Models Will Be Commoditized: Burkov Backs Efficient Open-Weight Models, Doubts Overpriced IPOs — marlene_zw · 2026-11-18
- Commentary on Christiano/Dwarkesh: training-pipeline automation is desirable and safety-helpful — sebkrier · 2026-11-18
- "If AI brings intelligence and opportunity to billions, dying industries deserve no sympathy" — VraserX · 2026-11-18