Only the first one seems to mean anything. (And the zeroth.) The third one he admits is a joke. The second one can't apply to something that can't make decis

Only the first one seems to mean anything. (And the zeroth.) The third one he admits is a joke. The second one can't apply to something that can't make decisions autonomously. The only orders you can give it are very specific ones it's programmed for. And if you've programmed an interface to do something and it doesn't do it, that's so obviously a problem that we don't need to discuss it. 0 and 1 mean more, though. Interfaces should protect the user from their mistakes, so long as they don't prevent the user from accomplishing their tasks by doing so. This is trickier than it sounds since that much power is hard to bubble-wrap. Protecting the entire web from bad decisions is even tougher. Designing a protocol is hard. You practically have to be clairvoyant to foresee all the possible problems.