Raw LLM Responses

It's because a lot people doing it from what I see just don't understand the underlying technology and what it is and isn't, and/or they anthropomorphize it. Also when you say, "LLMs will write about their own consciousness" - no they aren't even writing about their own consciousness(and I am not making a dig at you btw I am just making a general point), because they don't have a consciousness. They are writing what they estimate to be the most likely answer(the next token) to your query via statistical probability based on things it trained on in its dataset. That's it.