The LLM Is a Simulator of the Space of Text

The LLM is still sort of a simulator of the space of text. You could train it on basically like instances of walluigi behavior and then be like, stop it. Maybe we've got to start persecuting humans that try to jailbreak things by collapsing them into walluigi's. Plasibo Mansurs. Yes, exactly. Just like Eliezer said, if the terminators ever show up, he'll just be like, Oh, thank God, you were the robots who were sent here to protect me. That is the entirety of the post. I come out being slightly less doomeried than the post because I understand its arguments about walluigi’