Re:An interesting application? (Score 1)

Well, that's mostly true. But I think you guys are getting stuck in the details. There are a number of interesting ways to use NN's to simulate intelligence. An easily paralleizable (??) method is known (I believe) as voting: in this scenario, instead of having one large Net with many interconnections, you have many smaller nets, each of which comes to a decision on an output and "votes" on the total output. There's been some work done on this (can't find the references anywhere) and IIRC it seems that this works -- in general -- better at lower numbers of total nodes than a monolithic NN.

In any case, I just wanted to point out the an NN-like scheme can be devised that works well in a parallel environment.

stevec