“One of the main challenges to improve LLM reliability is…” — Practical AI at 39:15
“One of the main challenges to improve LLM reliability is to replace autoregressive token prediction with planning, to your point, Daniel. Pretty much every top lab, and he names a few, is working on that, and some have already published ideas and results. It's likely that QSTAR is OpenAI's attempt at planning. They…”
0:00
47:07