Sounds interesting. I'll have to look it up. I read Superintelligence a few years ago (Amazon link) and that was good but so pessimistic. It was very well written and gave a lot of good ideas. A bit information dense at times but good. But pessimistic!
I prefer to remain optimistic of what is coming. I mean, even if some crazy guys program the AI to kill all humans (and let's face it, this is inevitable), if AI is truly intelligent, it will be capable of making it's own choices and it may be just as likely or unlikely to follow such instructions as you or I would be. If they can remember everything, they presumably could understand the consequences or their actions better than we often can, and that combined with simple logic would probably lead them to make mortal choices. I think more Data (from star trek) than evil Arnold (from terminator).
But whatever the case may be, it will be interesting to see what comes.
I agree with your outlook, but how long would we need to wait for true intelligent and self-regulating AI? It's the time between now and when an intelligent AI appears that is the most dangerous I feel.
That's a good question and yeah your concern is probably right on the money.
I'm not too pessimistic - I mean most researchers and companies are keen to think these issues tbrough cautiosly!