I don't believe anything is a "real answer;" it's still actually just responding to prompts. AI LLMs can't think for theselves, at least not yet. It is a bit unnerving though how it does respond in conversation to make it look like it's responding on its own. The AI is still following the same model it described, answering until the user indicates the "problem" has gone away.