Raw LLM Responses

To the layperson, I think a Terminator AI is more viscerally compelling than a Monkey's Paw AI. For one thing, most people tend to think their ideas about how the world *should* work are internally consistent and coherent, and they probably haven't really had to bite enough bullets throughout their lives to realize that figuring out how to actually "optimize" the world is a hard problem. They also probably haven't done enough CS work to realize how often a very, very smart person will make mistakes, even when dealing with problems that aren't truly novel, or spent enough time in certain investment circles to understand how deep-seated the "move fast and break things" culture is. And then there's the fact that people tend to react differently to agent and non-agent threats - e.g. reacting more strongly to the news of a nearby gunman than an impending natural disaster expected to kill hundreds or thousands in their area. Obviously, there are a lot of things that are just *wrong* about the "Terminator AI" idea, so I think the really interesting question is whether that narrative is more harmful than it is useful in gathering attention to the issue.