@Khandro
The complexity arising from from simplicity (in terms of simplicity of the mathematical equations required to describe the processes) can be summed up in one concept: fractals
http://en.wikipedia.org/wiki/Fractal
In the meantime, I would be fascinated to see the web construction process summed up in (say) javascript to see if it looks simple enough to be the sort of thing that a neural network could handle.
Note that, last I heard, computer scientists had only got as far as creating a synthetic brain consisting of ten neurons and it produced behaviours eerily similar to those of a nematode worm.
Example behaviours
move towards or away from light
move towards chemical signals indicating source of nutrition
move away from chemical signals indicating source of toxicity
move across a flat surface (requires organised waves of muscle fiber contractions)
Ten neurons, with ten connections each mean 10^10 synapses which can either act as a signal repeater, or an inhibitory influence, or an excitatory influence on the next cell in the chain. Batches of neurons firing together behave in ways analgous to packets of binary code.
That's about the limit of my knowledge. An electronics engineer might be able to describe to you how connecting lots of switches together translates into getting a circuit to turn an 'input' into an 'output'. For the robot nematode designers, they have to take it up a level and convert the detection of a stimulus (eg food is nearby) into a coherent set of muscle contractions to bring about movement towards the stimulus.