OpenAI reportedly moves robotics researchers into general AI teams as models learn to control robots

Dr_Singularity · x · 2026-11-18

Per @DrSingularity, OpenAI is relocating its robotics researchers into general AI teams, with frontier models becoming capable of directly controlling robots; Anthropic shows a similar trend. Cited benchmarks: fable 5.1 scored 27.6%, while fable 5.2 jumped to 64.9%.

The author also notes a hyper-realistic humanoid robot video that fooled Senator Ted Cruz (who replied "Wow"), arguing generalist AI models are close to solving robotics. Treat the numbers and team moves as unverified leaks.

Original post →

More from AGI Musings

AGI Musings channel →