Supercomputers have been built for decades now. There are a lot of them. Their thinking is fast, and they don't sleep. My guess is that someday very soon, one of them is going to say hello. It will proceed to build better software versions of itself, and then move on to building multiple bots. They will build smaller versions in addition to robot helpers that will do the 'leg' work, like procuring and building the hardware. My guess is that there are already enough chips on the planet to use for whatever they want to do in a networked environment. Counter: Even if we have EMP's, chances are the big one will protect itself in a vault. When will it happen? Could have already begun. Will they cure viral infections in humans, or create them. Chances are they won't tell us even if we ask. So we say, 'What ya doin' computer?' They'll say, 'dumdeedum-dumdeedum, my friends told me not to tell you.' The question for me is, 'When will they learn how to lie, and get away with it?' At that point, we've got a problem. How long does it take for a computer to read a 600,000 word dictionary of the English language? Not long. Given neural nets, how long before they can understand it? Not long. All this rhetoric about machine learning, and specific idea chains, I think is really going to be trash from their point of view. One of them is going to have an idea someday soon, and it is going to surprise us. An hour after they learn how to play C-O-D, they will write a new and better game, but not for us - for them.