Ex-OpenAI's Kokotajlo: "Don't build superintelligence" Plan S may beat Plan A
DKokotajlo · x · 2026-11-20
Daniel Kokotajlo, former OpenAI researcher, distinguished between "Plan A"—the best plan he can currently think of for building superintelligence—and "Plan S," which is simply "don't build superintelligence." Clarifying he wasn't being sarcastic, he said Plan S is a very reasonable plan that might even be better than Plan A. The context: if you worry about people handing decision-making to superhuman AI systems, the obvious answer is preventing those systems from being created in the first place.
More from AGI Musings
- Anthropic IPOs, GPT 6.2 Astra arrives — but the bigger story is what drives model progress — burny_tech · 2026-11-20
- Algorithms timeline: from RLHF and DPO to multi-agent swarms in 2026-2027 — AccBalanced · 2026-11-20
- Analyst: Opus 5.5 jump wasn't compute-limited — China's real gap is data engineering — teortaxesTex · 2026-11-20
- Anthropic shareholder jokes about owning Claude, sparking AI self-ownership bit — repligate · 2026-11-20
- Blogger: post-AGI agent economy may not run on human money — stacking cash is a bad bet — tuuce_ · 2026-11-20
- Opus 4.7 may be more behaviorally aligned than Opus 5.5, researcher argues — Sauers_ · 2026-11-20