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

[philh]

I still disagree, but I think I’m going to try to get an executable model and see if that helps us hash it out :)