What makes AI difficult is that the traditional science/engineering approach of "divide and conquer" doesn't work: there are too many parts, and they're too tightly bound together. We need a different approach.
-
There are mathematical fields that literally study systems made of many parts that are tightly bound together: complex systems and dynamical systems have a lot to give to ML and AI in general.
They've been involved in ML from day one, but so far there isn't much to show. Hopefully this will change in the future.
