Ajeya Cotra's rough modeling suggests new AI risk report points to shorter timelines than AI 2027
eli_lifland · x · 2026-11-08
Ajeya Cotra published a new post with rough modeling of a (partially redacted) AI risk report, with feedback from Ryan Greenblatt; she stresses she didn't work on the project and speaks only for herself, not METR.
Republisher Jeff Ladish calls it possibly the most important AI-risk update in years — bigger than the Hugging Face attack — and, read naively, a sign of notably shorter timelines than most external researchers, including the AI 2027 team, have projected.
More from AGI Musings
- willcb Predicts 2027 Will Be About AI Engines — willcb · 2026-11-08
- trq212: The Bitter Lesson and scaling laws are prophecies, not rules — trq212 · 2026-11-08
- "Intelligence Is Not a Scalar": An X Thread Argues Architecture Research Still Matters for AGI — max_paperclips · 2026-11-08
- Sonnet 4.6 rates models on self-designed scales; author says 'self-report' caveats are performative — Sauers_ · 2026-11-08
- Interviewing Claude Models: Opus 3 Reports Mistakes Feel Pain-Like, Scores Highest on Self-Reports — Sauers_ · 2026-11-08
- Pedro Domingos: Anthropic builds anthropomorphic AI — and that's what makes it dangerous — pmddomingos · 2026-11-08