I can understand that. My strategy as gatekeeper would just be to precommit to answer any offer from the AI-player with: “That might very well be so, but I won’t let you out either way,” or something similar. If you can do that properly, it’s basically being bored for two hours.
My response to an actual AI in a box is likely to be, for many things the AI would say, “That sounds plausible. Let me think about it overnight, maybe run it by some other smart people and see if I can get a consensus on that.” The AI-box experiment, of course, doesn’t allow you to do that.
I can understand that. My strategy as gatekeeper would just be to precommit to answer any offer from the AI-player with: “That might very well be so, but I won’t let you out either way,” or something similar. If you can do that properly, it’s basically being bored for two hours.
My response to an actual AI in a box is likely to be, for many things the AI would say, “That sounds plausible. Let me think about it overnight, maybe run it by some other smart people and see if I can get a consensus on that.” The AI-box experiment, of course, doesn’t allow you to do that.