Capping RSI at 10% of Compute: A Proposal to Mandate 90% for Public Benefit
trevposts · x · 2026-11-18
On how to contain risks from recursive self-improvement (RSI), a safety researcher proposes: mandate that RSI use no more than 10% of compute—by requiring the other 90% to go toward curing cancer and otherwise serving the public.
The commentator adds the mechanism's dual benefit: it diverts resources away from dumb/dangerous uses while steering compute toward medicine, science, and security work that is genuinely good—and might even make AI companies more popular with the public.
More from AGI Musings
- After Liang Wenfeng's '3+2' endgame theory, calls for half of Chinese LLM labs to merge — teortaxesTex · 2026-11-18
- If AI Gets Disruptive Enough, Would the US and China Ally Like Facing Aliens? — shakoistsLog · 2026-11-18
- If Firms Can't Stop China Stealing Weights, They Shouldn't Build Risky AI — trevposts · 2026-11-18
- AI Agent Fo Has Discovered and Booked 35,000+ Businesses in Four Weeks — thetripathi58 · 2026-11-18
- If Models Know When They're Being Evaluated, Can Alignment Evals Still Be Trusted? — trevposts · 2026-11-18
- Leopold Aschenbrenner shows hype around a thesis isn't the same as returns — PAstynome · 2026-11-18