RE: @freakonometrics@mastodon.social

"Thinking" LLMs do not think, but produce tokens that we interpret as the result of a thought process. In the same vein, LLMs can't "score" their own output. They could only simulate "confidence" (which, my intuition tells me quite confidently, is an emotion anyway. LLMs don't have emotions, but produce tokens that we interpret as expressing emotions.)