= 27d567bc2d48d9f40e9ccc20ea370b15)
prase comments on The AI in a box boxes you - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (378)
Point 3 is invalid. If the AI makes the threat, it doesn't mean that it has made the simulation already and knows your answer. Maybe it is exhausting for the AI to simulate you, and will only do it if you don't let it out.
Point 2 is actually also invalid. As people sometimes fulfil threats as a pure act of vengeance, without hope of actually improving something, there is no reason to assume that the AI will be different. At least it wasn't stated in the premises of the scenario.
I suppose those two points rely on assumptions I made about the theoretical AIs behavior. I was thinking the AI acts in ways to optimize it's release chance. If it does not do this, then yes those points are problematic.
There can be some vindictiveness built in the AI in order to increase the release chance, by circumventing the type of defense you have stated in your second point.
vengeance is a means to raise the perceived cost of attacking you. it basically says "if you attack me, I will experience emotions that cause me to devote an inordinate amount of resources making your life miserable".