CMU Researchers Outline Theory of Capable Agents and a Path to Provable AI Safety
aran_nayebi · x · 2026-11-21
Aran Nayebi's group at CMU published a two-part ML@CMU blog series summarizing their recent work on the theory of AI and its safety (137 publications across 14 categories).
- Part I introduces the "Capable Agents" and "Contravariance" theories, arguing robots may inevitably become sentient as a consequence of how we build them.
- Part II works toward a science of provable safety, using capable agents as a case study for moving from barriers to actual guarantees.
The author frames the series in the wake of GRH being proven true by AI, aiming to lay theoretical foundations for provably safe autonomous agents.
More from Safety
- OnlyModels Model Fight 17: Qwen distills dominate, only 5/72 APEX claims held — Dave_Maynor · 2026-11-21
- Study: human auditors can't detect sabotage in AI safety research — ronbodkin · 2026-11-21
- ~$50K CAD unauthorized Gemini API usage on a partner-billed account sparks routing nightmare — Big-Coffee-8899 · 2026-11-21
- Signal blocks screenshots on iOS 27 and Android to keep AI agents out of chats — The Decoder · 2026-11-21
- Fireship: 178 sites hijacked via Optimizely, one Base64 string owned Java servers, 10-year-old Nginx bug hit 225M sites — Fireship · 2026-11-21
- House committees' 7-month probe flags Airbnb, Cursor over use of Chinese AI models — Formal-Grass-3173 · 2026-11-21