chess doesn't have an infinite set though, that's what IBM's "deep blue" was doing: it calculated the entire tree of all possible games and chose the winning paths
I know, but to the average human mind, it sure does look like infinite.
Maybe the endgame is like that, but for chess at least.. I know the good players kinda memorize all the openers and "counters" to those starts.
The mid game is where the infinite possibilities exist. Yes. Every serious chess player has openings that they prefer as white and defenses for many of those openings as black, but after say 5-15 moves into the game, at some point, one of the players makes a move that presents a position on the board that has never been seen before. From that point on, it's a new game that's being played for the very first time.
Bullshit. Deep blue lost 2 of 6 games in the kasporov (who is not the world's greatest chess player) rematch and 3 of 6 games in their initial series.
The possible combinations are near infinite and the AI must assign a point value to each move and run through 20-30 move combinations, depending on the time given for each move, and then selects the best move based on the points system it uses. Human players can, and often do, come up with unpredicted moves which then throw the AI off making what were high scoring moves before the human's unpredictable moves into blunders that ultimately lose the game.
Even today's best AI computers still often succumb to the creativity of the world's best chess players.
"near infinite"
the set of all possible chess games is most definitely finite, and computable. never said deep blue was unbeatable, just what it's strategy was -- it's entirely possible to be driven into a region of the tree having no winning results
You said,
it calculated the entire tree of all possible games and chose the winning paths
Which is not true. It calculated possible moves to a certain number of moves out and assigned a point score to each path. Today's chess computers go to 22-25 moves out. No idea how far out deep blue went, but it was never "all possible". If it had done all possible, it would win every time, because it would exploit the opponents failure to make the best move in every game.
And, two computers going 30 moves out, or even more, still does not lead necessarily to a draw.
(post is archived)