Mike

misaligned.markets

17 October 2026 19:51

I'm sure @deontologistics.bsky.social would have a lot to add here. Really liked this post from him, but I know he's written much beyond these older thoughts on critiquing post-Searlean critiques of LLMs/naive views on reference.

William Gillis 🏴

rechelon.bsky.social

17 October 2026 19:47

Yeah, I have no interest in delving into some people's brains. But also alas "interestingness" doesn't equate the same thing as moral importance around getting an accurate map correct. It matters extremely to get one's model of the prison guard correct, but the dude's shit is uninteresting.

William Gillis 🏴

rechelon.bsky.social

16 October 2026 18:27

Sure, I would just also quibble with the word "roots" -- sometimes the most steelmanned version is not even remotely the causal origin of their position or even a future attractor pulling them in, and what they're throwing out that can be mapped towards it is just chaff they don't believe.

William Gillis 🏴

rechelon.bsky.social

15 October 2026 23:11

I would nuance that the most important thing is accuracy not necessarily charitableness. It's important to ALSO understand and trace the steelmanned versions of certain things, but it's at least as important to understand when someone isn't saying the steelmanned version but an utterly venal version

boarders.bsky.social

15 October 2026 22:38

the rule of corecursion club is the first rule of corecursion club