Ray Kurzweil reviewed the history of Neural Nets.
Ray knew the both of the two leaders of the competing Artificial intelligence factions (Symbolic vs. Connectionist) in the 1960s.
The connectionists were supporting neural networks. They could only handle single layer neural networks and had limited solution capacity.
A mathematic proof by Minski in the Symbolic school proved that neural nets could not handle particular basic problems. This limitation only applied to single layer neural nets.
Decades later 3-4 layer neural nets could be made but a mathematic limitation prevented them from going beyond 3-4 layers.
Another mathematical solution was needed to get beyond this limitation.
This has enabled the current age of deep learning with hundreds and thousands layer neural networks.
Neural nets with 15 layers are able to distinguish between dogs and cats.
Ray believes regions of the brain are key and understanding the neocortex.