Richard Socher says modern language-model training includes multiple stages, including reinforcement learning that enables models to take actions.

“There's so many more clever things that people are doing. They're like different stages of training.”

“you have the whole rl training and you can take actions and like all of these things where that can go really far and then the folks that come from the neuro symbolic direction say oh this will never work because they can't do neuro symbolic reasoning it's like”