Artificial neural networks (ANNs) were conceived to replicate the way that living creatures with brains learn. Extensive work has been done on optimizing the training of these neural networks. It is easier to perform experiments using artificial neural networks than natural ones because the former require no living subjects.

I am currently training a natural intelligence. Among other things, he can transport himself both horizontally and vertically over a wide variety of surfaces, and can identify various classes of objects (dogs, cats, cars, etc.) easily using visual means.

Are there any ANN-derived insights that might be used to expand the training range of a NI?