I really REALLY wish people would stop assigning sentience to LLMs. > In other words, a hallucination is an error in (or a false) perception of something real

I really REALLY wish people would stop assigning sentience to LLMs. > In other words, a hallucination is an error in (or a false) perception of something real or concrete. an llm has no "perception" it doesn't "believe" or "think" that the answers it provides are "correct" or "true" or even "false". It's just autocompleting strings with the most probably next words. If we keep treating these things as if they're sentient entities that "want" to provide "correct" answers we're going to keep tripping over our false assumptions about their answers.