Evolve with a purpose

One of the main obstacles in the design of Alife systems is the concept of emergence of complexity.

In most (if not all) Alife systems, we would like complexity to emerge, appear on its own.

We should specify some basic rules, guided by some broad principles, and hope that as we let the system evolve, we get interesting behaviors.

Almost by definition, this is a very hard thing to do. Since we expect complex, emergent behavior, it is very hard to predict what a rule change will cause, because of the non-trivial interactions between many degrees of freedom.

A nice thing would be if we could somehow probe what a small change in the rules causes, and inch towards something interesting…

Wait… that is exactly what machine learning does !

The game plan is then as follows :