It's one thing to act like a rocket scientist and another to actually be one. AI will fail, although it will provide a few toys for the amusement of children and other consumers of entertainment product.
I tend to agree in the short run. Nature tends to solve problems with genetic or evolutionary algorithms, and this is very difficult to emulate in current hardware. Efforts are, however, beginning to be commercially useful, and this will start driving research.
Interesting. I have friends whose son is working with AI and some interesting progress is being made.