[barbarian]

"Super Intelligence" made a nice point that AI will likely be very stupid until it's suddenly very, very smart. I thought the AI worries I read about in the media were a bit hyperbolic till I read this book - it makes some good and sensible arguments as to how a human dangerous AI might come about. It doesn't claim general AI is necessarily very likely, but more that, in the event it does come about, it will be very sudden, very swift - it won't be the gradual curve of innovation and improvement we've been used to elsewhere - and so our time to react to it's birth and implications will be very short.