I believe the architecture for convolution neural networks were directly inspired by how vision works and some of the core design choices map onto real features of the visual cortex.