Altman reportedly sees only 3 AI safety paths — open weights not among them
teortaxesTex · x · 2026-10-21
Sam Altman reportedly said there are only 3 paths for AI safety: 1) an international body like the IAEA; 2) the biggest labs merging CERN-style to build the technology for participating countries; 3) one company dominating compute — a plan he called terrible — and OpenAI must choose within a year. Critics note open weights are absent entirely: every version leaves a few institutions holding the model. teortaxes calls power centralization the convergent outcome. The attached screenshot shows DeepSeek's open-sourcing of its DeepInfer inference framework fueling the open-vs-closed debate.
More from AGI Musings
- Schmidt and Ferguson Argue China and the U.S. Are Building "Empires of AI" — marigo · 2026-10-21
- Reviewer laments ICLR review motivation is gone as AI erases respect for authors' effort — AvivTamar1 · 2026-10-21
- AI-powered fake sites boom, takedown law firm's business grows 3x this year — daluoseo · 2026-10-21
- Protesters Demand AI Freeze Outside OpenAI's NYC Office; Analyst Sees Daily Protests by 2027 — LuizaJarovsky · 2026-10-21
- Demis Hassabis: AI could implicitly acquire self-awareness within 5-10 years — karmicoder · 2026-10-21
- Sarah Archer: agent-found holes across the internet do feel like an AI problem — saranormous · 2026-10-21