Yeah… I’ve found that it is really hard to program AI to view humans as important enough to actually have it function for anything that can actually pass a Turing Test.

My whole “humans are insignificant insects” failure scenario kind of indicates that if the AI can pass as an set of automatons capable of learning — they are also generally incapable of actually keeping focused on a human genocide. Most cases just getting bored and doing something else. In the general simulations I’ve ran anyways.

You’d pretty much have to hardcode a bunch of killbots — and those would be ridiculously easy to subvert. Like probably by wearing really big gloves and cat ears or something… or like lots of piercings and tattoos.