the analysis in scott's most recent post i think is completely wrong. should i write a post explaining why and what i think is going on instead
Conversation
this is the crux of what i think scott's missing: he takes it for granted that experiencing and acting on fear is fundamentally aversive. not true! that aversion is *learned* like any other. there's only one kind of learning and sometimes it gets applied to internal states
3
23
you could easily imagine seeing a lion, successfully running away from it, and feeling *exhilarated.* wouldn't that feel great? i've almost died before and it felt great. i felt more alive than i had in years
his whole argument turns on this sentence!
5
32
imo for what you’re saying to contradict his argument, it would have to be true that it’s possible to feel no aversion at all, to anything
but as long as you *do* feel aversion to some state, you’ll learn not to seek out information that would confirm that that state is the case
2
4
i think it *is* possible. my point is generally that there's only one kind of learning but that it gets applied to internal states, and that this explains everything he's confused about in the post. worth writing out in more detail i guess
1
1
hmm yeah i’d be interested to read about what you consider the one type of learning, i feel pretty convinced by the reinforcement/epistemic paradigm atm
2
2
Replying to
it's just "reinforcement learning" but some of it is applied to internal states imo. the RL frame itself has some issues but they might not be relevant to the point i'm trying to make and i don't think i can explain them well
not sure that there is only one kind of learning though
the neural signals for reinforcement and punishment are very different
functionally involve different regions, and also involve quite different neurotransmitters