AI and the Better Hand

Games have always been a good testing place for AI.  Now that Chess, Checkers have been professionally 'solved' with AI, other games with randomness and behavioral elements are being addressed.   Of course these games are still much simpler than 'games' that include randomness and continually changing context.   Essentially where the rules are constantly being changed.  The card game Poker is a kind of instructive intermediate example. 

AI Holds the Better Hand  By Don Monroe 

Communications of the ACM, August 2018, Vol. 61 No. 9, Pages 14-16

Although games of skill like Go and chess have long been touchstones for intelligence, programmers have gotten steadily better at crafting programs that can now beat even the best human opponents. Only recently, however, has artificial intelligence (AI) begun to successfully challenge humans in the much more popular (and lucrative) game of poker.

Part of what makes poker difficult is that the luck of the draw in this card game introduces an intrinsic randomness (although randomness is also an element of games like backgammon, at which software has beaten humans for decades). More important, though, is that in the games where computers previously have triumphed, players have "perfect information" about the state of the play up until that point.

"Randomness is not nearly as hard a problem," said Michael Bowling of the University of Alberta in Canada. In fact, "the techniques developed for perfect-information games don't work in imperfect-information games."

Perfect information games have largely been conquered using machine learning techniques that explore a huge number of ways the game could proceed to infer the best move for any situation. With poker, as in much of the real world, opponents know facts that you do not; the challenge is to infer something about this hidden knowledge from their actions, even as their actions are guided by their assessment of your hidden knowledge, and so on. Such recursive logic cannot be analyzed with simple machine learning.  .... "