I think he argues that it's the actuators that we need to be concerned about, because there are plenty of intelligent entities around who will do bad things wit

I think he argues that it's the actuators that we need to be concerned about, because there are plenty of intelligent entities around who will do bad things with actuators that allow that. One more entity intent on doing a bad thing does not make a great deal of difference. Neither humans or AI's should be given the tools to do great harm to other humans or AI's. He cites drones as an example, but of course nuclear weapons are the obvious gotchas for humanity. One trident sub is perfectly capable of killing a hundred million people, I doubt whether anyone so vaporized would give a toss whether these were launched by a googlebot or a mad captain.