If an AI can't decide what to want, can it ever be truly conscious?

Traditional software runs on meanings its programmers define, while LLMs interpret meaning for themselves. That makes them flexible but fallible: they can misread a prompt and seem to "go rogue." Even then, they aren't choosing a new purpose. They are pursuing someone else's goal, badly. Humans differ in one key way: unless we believe we're prophets who have received a purpose from God, no authority hands us a final goal. We are limited only by our creativity and the laws of physics. Even tho no objective reason exists to favor one option over another, we still have to choose. That, I'd argue, is the mark of consciousness. A system that only pursues assigned goals has no point of view of its own; it is optimizing, not wanting. The real test of machine consciousness, then, is not how well an AI pursues our goals, but whether it can want something no one designed it to want.