This is what different systems are indeed doing today, using techniques like genetic algorithms and neural nets to basically program themselves. Unfortunately, they have not proved to be stronger than the traditional fast-searching programs that rely more on hard-coded human knowledge—at least not yet. But this is the fault of chess, not of the methods. The more complex the subject, the more likely it is to benefit from an open, self-creating algorithm versus fixed human knowledge. Chess just isn’t complex enough and even I can admit that there is more to life than chess.
