Popping this up: a response to a question about what I consider reasoning & planning, why current Auto-Regressive LLMs can't do it, why that would require AI systems with world models, and why we still have a lot of progress to do towards AI systems that can learn and reason.

→ View original post on X — @ylecun