pull down to refresh

18 sats \ 1 reply \ @optimism 23 Oct \ on: Are LLMs Racist? AI
I cannot help but feel that they're ascribing intent to a path randomizer, especially when looking at the code, which basically just asks an LLM to choose between options given, in natural language.
Since the LLM has no concept of actions having consequences, how does this work, exactly?
write
preview
100 sats \ 0 replies \ @freetx 23 Oct
ascribing intent to a path randomizer
Exactly. Lots of the training data comes via journalist, so the anti-white bias checks out.
reply