Yes, this explains how it works. The primary method for training this model is behavior cloning, and I explain what it is along with reinforcement learning (RL) and its limitations. And I don’t mention bias at all, sorry.

→ View original post on X — @dfintelligence