Over the years I have seen increasingly complex board games 'brute force' or 'deep learning' attacked so that computers beat the best human players. Is it theoretically possible to produce a board game using a finite sized board that isn't amenable to such techniques? I don't THINK that infinite chess has been dealt with but the clue is in the title - the board is infinite. I suppose brute-forcing will eventually beat any human except, possibly, in a game where a player can alter any combination of pieces on the board. I'm imagining a setup like go but with a gameplay that allows every single piece to be reversed (e.g. black for white) and even then, it's only hard to brute force because if the board has 361 positions (as go) & potentially ALL 361 could be flipped, the solution would be large - possibly too large to brute force. I don't mean I HAVE such a game, but can a game be designed that would place computing solutions out of reach (I am using 10^82 positions as the minimum since this is the maximum estimate of atoms in the universe) BUT be playable by humans. Again, I look to Go because of it's simple board.
Sorry if this is a dumb question but I have looked but never found a discussion on this.
I totally agree that FINDING such a game is not a trivial task but there are limits on machine learning that logic cannot overcome. DL requires:
If 1 of those 3 are not the case, deep-learning cannot solve it. Brute-force is easily defeated because even a modestly large play area will ensure more possible combinations that can be stored.
I suggest that the definition of success is agreed upon by the players before each game. Maybe this is similar to 'magic' as HungryGoose suggested?
BTW 'infinite chess' as the name suggests uses an infinitely large board so I think that's an example of a game that DL cannot learn BUT I'm limiting my question to finite playing areas.