If Shannon and Hinton collaborated, what would they build?

I was brainstorming sideproject ideas. I already implemented Shannon 1948 : A Mathematical Theory of Communication and Hinton 1986 : Learning representations by back-propagating errors.

I was curious to find out, if Shannon collaborated with Hinton after 1986, what would they make?

Would we have a wild neural network architecture? Would we be using data compressors instead of matrices to represent neural networks?

What do you guys think? I bet it's a silly question