james.lucassen comments on [Intro to brain-like-AGI safety] 5. The “long-term predictor”, and TD learning