Naomi Clark (@metasynthie) 暗悪直美@halfrobot.com3 MarThere are many devils in the dank, latent realm of LLMs. But the reinforcement-learning-human-feedback devil is one of the worst, most craven and conniving of these devils.31749
Naomi Clark (@metasynthie) 暗悪直美@halfrobot.comhegemon.substack.com/p/the-easy-p...hegemon.substack.comThe Easy Promise of Hidden RevelationsDon't blame LLMs. We have always yearned for slop.3/3/26, 9:25 pm·0 reposts·1 like