Couldn't ANN's be modeled on any part of the body (not just the brain)? It's commonly said that ANN's are modeled after the brain, but as we know, neurons are also found elsewhere in the human body.
So if someone was being provocative and said that ANN's could equally well be modeled after the appendix, what are some good reasons to reject this logic?
In other words, how does the architecture of ANN's match the structure/organization of neurons in the brain above and beyond that of neurons found elsewhere in the body?
One example I can think of where the neurons are organized in the brain similarly to an ANN is with convolutional nets. Here the features learned in the different layers somewhat match the increasingly more abstract receptive fields that are seen in the visual system (eg. V1,V2,V3 etc...)
 A: At least where I live, suggesting that someone is thinking with a distal part of their digestive tract is not typically a compliment; I think the people working on ANN's would much prefer their work be compared with the operation of the brain.
That said, if you wanted to go outside the brain to include sensory/motor neurons in the spinal cord or periphery in a general architecture of a ANN, you could do so, but the interesting biological differences in those networks aren't really captured in a typical ANN - you're really just adding a new layer.
I think the simplest reason to reject a model outside the brain is that other parts of the nervous system (like the enteric nervous system) are not doing the types of tasks that most ANNs are intended to achieve (typically involving some level of classification or prediction). They are more about coordinating and synchronizing motor activity, like peristalsis of the gut. There is no reason you couldn't use a ANN modeled on gut function to do similar tasks, but there are much simpler algorithms available.
If you have a particular goal in mind, that might improve your thought experiment a bit, otherwise it seems like just a thought experiment.
