Bubble Thought over a Neural Network

I want to join the effort of constructing a working thought engine.  Here are some starting thoughts on where to begin.


Lets use A Bayesian neural network  of relationship dependency as part of a neural network to form bubbles of thought.  A bubble thought is a imprint signal propagated over the network with instance signal linked to relevant variables.   The neural network can use pattern matching  resulting signal propagation over the complete network to achieve computation.  We then use output or new input to bubble continuing thought.