This was supposed to happen, like, ten years from now. In some sense it's no more like a human than the computer that beat Kasparov at chess twenty years ago, but it's not just the old game-beating algorithms with more processing power; the algorithms themselves are better. Until recently, a lot of AI skeptics were saying "beat top humans at Go and then I'll be impressed".
It comes from the same team that does DeepDream, which has mostly just been treated as a funny novelty that puts out weird acid-trip pictures, but is also a huge leap forward in terms of AI, and has only been improving in the months since it was first announced. Guys, I know that remaking non-dog images out of dogs isn't exactly high art, but it's also the kind of thing that was flatly impossible for computers to do until recently. DeepDream isn't a lost Photoshop filter, it's a major step towards human-like creative thought. The starving artists are going to be starving harder when they're also replaced by computers.
Do I think that AIs are going to become self-aware soon? Probably not, depending on how you define soon. But I do think that they're going to become more intelligent than most people anticipate, and as a direct consequence, more people's lives will be threatened than you assume here. What happens to the wealthiest businesspeople when the best computers can run large corporations more efficiently than they can? What happens to our economy in that case? What happens to our culture when popular media is literally designed by algorithms that optimize products like films and books to succeed? What will be the first nation where the human government decides to act simply as troubleshooters for a governing machine?
All of this could cause disaster very gradually, but a disaster could happen suddenly, too - not if a powerful AI attains self awareness, but rather if a powerful AI metaphorically forgets that it dropped a grenade and then stands in place.