I came across this page from my University. It is the slides for a graduate level systems design engineering course on how to model neurobiological systems. The professor is a really smart guy who is cross listed in both engineering and philosophy departments and is well aware of all the theoretical issues concerning artificial intelligence.

http://compneuro.uwaterloo.ca/courses/c ... index.html

I don't really understand the math, but gather that it uses basic neuroscience as theory and then models the encoding and decoding of neurons using algebra and calculus. I would assume to model any kind of thought process which may use millions to billions of neurons for any length of time would require supercomputers. I think at this point there is a lot of guesswork among researchers, and they are only just beginning to learn how to model these systems.