1. I never said LLMs were not useful. They are, particularly with all the bells and whistles that are being added to them. I use them.
2. A robot-rich future can't be built with AIs that don't understand the physical world and don't anticipate the consequences of their actions.

→ View original post on X — @ylecun,