Is GPT-4 a Good Solution to the AI Alignment Problem?

The chatbot starts as a superposition of both the well-behaved and badly-behaved. The user must interact with the bad AI in order to bring it into being. RLHF is probably increasing the likelihood of a misalignment catastrophe, according to this theory. This theory has increased my credence in the absurd science fiction tropes that the AI alignment community has tended to reject. It links back to literal robot war with a picture of the Terminator skeletons.