"Pretty wild and dangerous things" - I think you're suffering from a failure of imagination there. Examples of humans who thought something wild and dangerous w

"Pretty wild and dangerous things" - I think you're suffering from a failure of imagination there. Examples of humans who thought something wild and dangerous were Hitler, or Jesus or Mohammed; those people changed the face of the world because they thought something. And those people were human, constrained by evolution to think in certain ways. An AGI doesn't necessarily have to think inside the human box. There is a lot more of "dangerous thought-space" out there than you seem to be accounting for. Your second paragraph: presumably you think it's literally impossible to invent nanotechnology to make yourself omnipotent, but I'm not willing to place that at less likely than 5%. Something clever enough could probably just manipulate the existing social structures to kill all the humans (something very intelligent that truly wanted nuclear war could probably make it happen). Just a little bit of tweaking to smallpox might be able to destroy humanity anyway. Even just the right new religion might do it (witness the great purges by pretty much every religion ever). None of these suggestions are intended as "this is a plausible way of killing those inconvenient humans"; but if each of my suggestions (nanotech, nuclear war, pandemic, grand new idea like a religion) has a five percent chance of working, that's already a 18.5% chance that exactly one of them would work. Once the humans are dead, of course, it has all the time in the world to achieve all its goals.