How would an AI calculate the percentage of sensory-memory sameness?

Say I show the AI a sock, how will it react with "sock"?How would an AI calculate the percentage of sensory-memory sameness?
This is a kind of pattern recognition problem that neural network (NN) based AI is readily adapted to solve. It is simply a matter of 'training' the NN to recognize a 'sock' and be able to reliably distinguish it from a 'non-sock' in whatever visual, auditory or tactile context one chooses for the experimental design of the AI system. Here is an example, a web java applet showing how a simple NN can be 'trained' to recognize simple hand drawn characters. Try it!



AI and NN's are a fascinating field of study. If I were starting over again in my EE career, this is definitely the field that would hold greatest interest for me. Enjoy!



http://www.sund.de/netze/applets/BPN/bpn鈥?/a>How would an AI calculate the percentage of sensory-memory sameness?
These problems were often contemplated when I was in graduate school, and contrary to some people's opinion, it was not a simple matter.

First, there was the brute force method. A number of optical sensors were trained on the object and the percentage of sensors stimulated by the object was compared to some model. Obviously, this method required that objects be the same size and have the same physical characteristics. Mathematical transforms could be performed on the image to change the size to match the model. But then you would need at least two models, tube socks (which are straight) and regular socks with a bend in them.

The second method which seemed more fruitful was to look at the relationships of the parts.

A third method was to define the object in terms of function.

There was also some discussion that parallel processing might perform this task better than sequential processing.

The purpose of this research was to try to make the machine recognize objects and label them as humans do.How would an AI calculate the percentage of sensory-memory sameness?
Oh, maybe the AI will find the other sock that is missing. The AI can't calculate the smell of the sock, or the warmth the sock can provide, or the softness of the sock, but maybe the AI can remember where the other sock is.