In my opinion, a simple AI (Artificial Intelligence) can be modelled by a Finite State Automaton (or a Finite State Machine, FSM), that includes nondeterminism.
Part of input used for choosing a transition between states can be random data.
That way, for example, a Computer Game AI players can be more challenging opponents.... human players won't be so sure about AI players' decision making.
then we can, for example, each of such finite state automatons, run as a separate process / thread.
each of them being a 'neuron' responsible for something.
they can communicate & cooperate.
we can give / balance priorities & resources to such.
for example, we can have processes responsible for security that can be given a higher priority & more resources at times of danger.
programs can detect higher priority / more resources and react, for example by activating more functionalities, passing these resources to other processes. i call this a 'tiered functionality' that is activated by a priority level and available resources.
See also: Neural Networks, Stitie Space, Neural Network Learning.