I decided to write a little bit of code to emulate a basic version of ‘selection’. Not sure where I am going with this just sketching a simple environment to see how hard it is and therfore how hard it might be to create a more elaborate setting.

The link below will run, in a separate window an arena  filled with 500 ‘creatures’, each with a unique colour. The brain (or neural network)  for each of these creatures has two instructions. The first is to agitate randomly to an ajacent position and the second command is for 10% of the moves they will move, consistantly,  in one of the 4 compass directions (N, S, E, W).

At each side of the arena are two zones, these are the ‘feed zones’. Every time one of the creatures finishes up on the feed zones for the next cycle or generation 2 creatures with exactly the same parameteres including colour and direction bias will be randomly spawned in the world. The life time of the creatures is 150 moves.

I have introduced a couple of slider controls to increase/decrease the number of moves that determine a life time and the second to increase/decrease the ‘move bias’.