⟳ Репост от Mx Maddison Stoff
it really is fascinating, it’s not like the problem never happens with other versions but there was a while there where it was like, is this all LLMs or just 4o and it turns out the answer was “4o specifically destroyed people’s minds at a massively disproportionate rate”
laurensure hope there are some smart people studying that one version of chatGPT that was apparently just tremendously good at making people insane