You may want to take a look at the Shannon number, we would need quite a large neural net to solve chess in this way.