The Professor Who Proved ChatGPT Can't Think — Subbarao Kambhampati

Machine Learning Street Talk · 92:14

Subbarao Kambhampati argues that o1 is *not* just another LLM but a genuinely different system — most likely an AlphaGo-style RL post-training phase over "pseudo-action" prompt-augmentation tokens plus inference-time...

Read the full summary on tuber

Redirecting...