x
Still no Lie Detector for LLMs — LessWrong