I've done a fair bit of work in AI, mostly genetic programming and evolutionary computing. The popular conception that if we just had more processing power we could create a truly intelligent machine is, well, nonsense. The machine still needs some code to run. The problem for AI isn't one of computational power, it's a problem of representation. (The history of Deep Blue is an excellent example of this.)

I'm not suggesting that more computational power won't help - quite the opposite. But it doesn't solve anything on its own. As Michio notes, machines can already see and hear better than humans, but they have no understanding. That understanding can only occur with good information representation. I personally think evolutionary techniques are probably the quickest path to get there, but then, the coders will be no more aware of the machine's internal representations than neuroscientists are of humans' internal representations today. Whether that understanding is important is something of a philosophical question.

Women are quantum computers. How else can a man can be wrong even when he's right?

@jonny

http://videosift.com/video/Conway-Game-of-Life-simulating-itself

Moores Law hasn't been valid for almost 10 years.