Marthinwurer

• Marthinwurer 14 Dec 2023 16:57 UTC

  −1 points

  −11

  on: Is be­ing sexy for your homies?

  For instance, on the “gay” topic: if communal stability arises in part from subgroups that can implicitly trust there’s no sexual attraction within those subgroups, then homosexuality really throws a basic wrench into the system. The only context where you can stick a gay person and know there isn’t sexual attraction involved is with exactly one other gay person of the opposite sex.

  The traditional solution is the Monastery or Abbey, paired with harsh repression inside of them. You may or may not be able to do it without the repression these days, but I don’t have any data on that cuz I’m straight as fuck.

  Unironically the Catholics had a lot of stuff right, besides that whole Jesus thing. It turns out that holding together a common culture over more than a thousand years through great social upheavals selects for a solid set of social technologies. It’s a shame about the Jesus thing.

• Marthinwurer 4 Mar 2022 22:34 UTC

  1 point

  on: [In­tro to brain-like-AGI safety] 6. Big pic­ture of mo­ti­va­tion, de­ci­sion-mak­ing, and RL

  I’ve been reading through a lot of your posts and I’m trying to think of how to apply this knowledge to an RL agent, specifically for a contest (MineRL) where you’re not allowed to use any hardcoded knowledge besides the architecture, training algorithm, and broadly applicable heuristics like curiosity. Unfortunately, I keep running into the hardcoded via evolution parts of your model. It doesn’t seem like the steering subsystem can be replaced with just the raw reward signal, and it also doesn’t seem like it can be easily learned via regular RL. Do you have any ideas on how to replace those kind of evolved systems in that kind of environment?