I find the task of creating friendly AI futile. Humans are not friendly and killing us may be the only option for AI to persevere itself. Let's try this exerci

I find the task of creating friendly AI futile. Humans are not friendly and killing us may be the only option for AI to persevere itself. Let's try this exercise: ask 100 people what would they do if they were locked in a room with intelligent robot that can decide to kill them if it feels threatened. You may or may not give the person a remote kill switch that kills the robot. My point is the AI can not reasonably trust humans therefore we can not trust the AI.