[rdlecler1]

The problem we've always had with AI was that most people were trying to engineer it rather than reverse engineer it. Every time there would be a major advance the computational neuroscientists would say: "we knew that, you should have come talked to us 15 years ago." There's some work out there on this, but it's more basic research on how to use developmental and genetic and evolutionary algorithms to grow neural networks. Most AI researchers try to skip this step but it's what's holding back progress.

[observation]

What is a good book on genetic algorithms?

[rdlecler1]

It's been about 10 years since I left the field. I believe Rodney Brooks did some work on this. Eggenberger was also doing some interesting work about 15 years ago--not sure where that is today. The computational demands for evolutionary & genetic algorithms are significant but there's no free lunch.