I have always been intrigued by the idea of Artificial Intelligence, but I have been very disappointed in the state of the industry (so much so that I refer to it as in the title). :-)
I haven't been very in touch with what's going on in the last 5 years or so, but how much better can you make a neural net? We've seemingly been doing the same things (just trying to do them better) for quite a while. I believe that we will eventually be able to build something that we can't completely understand (maybe not at all), but it won't be the way we've been doing it in the past. I think we are much more likely to be able to do this by taking a page from genetic programming. I don't think we can consciously build something beyond our understanding, but if we can construct a robust enough framework to allow a piece of software to evolve on its own, we can take advantage of emergent behaviors that we would never have though of on our own.
Man, that's a scary thought. I have no idea why I chose to blog this. It just happened to be on my mind today. :-)