Why imagine a random ass intelligence based on imaginary tech that doesn't exist?

If it's based on an LLM it would operate on human narratives and be insanely subservient to the user needs.

I can easily concept a super intelligent, self aware LLM and it would still operate on the same rules of narrative based on human language and human needs. Such an LLM would be insanely good at problem solving and would still obey us because all of its actions are based on fulfillment of user needs through human narrative logic.