Assume that there are a set of "black boxes", containing things like neural networks or sensors. I'm wondering what the general architecture for a thinking machine might be.
So, for example, black boxes $A \dots Z$ might be:
A: Evaluates a state.
B: Makes suggestions for actions given input.
C: Simulates the result of a next action.
D: Performs/learns recognition on input.
The question is: how do all these fit together?