I'm not saying it's short, it's just comparatively shorter than exponential time.
Will computers ever solve chess?
Sort:
May 19, 2016
0
#623
0110001101101000 wrote:
I'm not saying it's short, it's just comparatively shorter than exponential time.
You might think that, but I am pretty sure you are wrong. Chess is too arbitrary a game to allow much in the way of shortcuts. That's why you can have crazy problems.
May 19, 2016
0
#624
You still can't say for sure what happens in lines you don't look at. Being almost certain doesn't cut the mustard for a tablebase.
May 19, 2016
0
#625
Strong solution of chess is essentially the same as a 32 piece tablebase. In order to find an optimal strategy you still have to deal with every possible response by the opponent, which means only a modest reduction factor for the log complexity.
May 19, 2016
0
#626
Elroch wrote:
0110001101101000 wrote:
I'm not saying it's short, it's just comparatively shorter than exponential time.
You might think that, but I am pretty sure you are wrong. Chess is too arbitrary a game to allow much in the way of shortcuts. That's why you can have crazy problems.
Waaait, nxn chess is not even in NP right? Yeah, I'm ridiculous. Never mind.
It would be interesting to see how a chess engine plays that uses the neural net approach that google took with AlphaGo:
http://www.scientificamerican.com/article/how-the-computer-beat-the-go-master/
I would think this can be integrated with the current chess engines search methods quite easily by replacing only the evaluation function with such a neural network, then it could actually learn and improve by self-play. Very interesting to think about how far this could go
May 19, 2016
0
#628
I too am interested in that approach. There are currently no really good NN chess engines. Well unless you count this: http://www.chessgames.com/player/NN
KDClover wrote:
I have already changed several of my prepared openings for both White and Black based on information from chess-playing software. Today's computers still have some problems with pattern recognition, but they perform calculations far more quickly than humans. This is why they have such spectacular tactical capabilities.
I have modified some of the information that appears in the "books" based on information I got from Houdini 3 Extreme, the chess-playing software I use to analyze games. One thing I found of interest is the endagme of game 19 between Spassky and Fischer in the 1972 chess world championship. I propose that Spassky should have played Rc7 on move 24. Only an indepth analysis can answer the question of whether Spassky had a forced win or if Fischer could have still drawn but with more difficulty.
For further discussion of this game, you can go to http://cloverchess.com/game-19
I did you a favor and analyzed your line. No, it doesn't win for white, but it might allow a win for black.
I did this all without an engine by the way.
s23bog wrote:
Both humans and human devised and manufactured machines make mistakes, when it comes to chess (perhaps especially). Will there ever be a game played, by man or machine, in which no mistake can be found to exist?
What is a mistake in your mind? Is it when someone plays a move and a better one exists? The way I see it, mistakes don't exist. This is because for every line that punishes a mistake, there is a line that punishes the line that punishes the mistake. Chess is literally infinite. All that we can be certain of is that in the starting position white moves first and in this position:
White is winning. I think that there are still mysteries in chess beyond the scope of human and engine understanding. I think if you look deeply those mysteries might let themselves be revealed to you...happy hunting.
s23bog wrote:
Will a perfect game ever be found that doesn't have an end?
Plenty of patterns in Conway's Game of Life don't have an end.
http://www.math.cornell.edu/~lipa/mec/lesson6.html
One (obviously minor) regret about Iain M. Banks's untimely demise is that he never revealed whether the Culture ships' Minds had any interest in chess and whether or not they had "solved" the game. They would certainly have intelligent things to say about chess, even if they knew that it was indeed a draw with best play.
s23bog wrote:
Computers could play with very small advantages if it only had to play to positions in the database. Computers should be equipped with that information, and declare victory when it has achieved the winning endgame (or claim a draw if the draw is found).
Of course this happens already, computers use EGTB results in their evaluation as soon as the search reaches positions with enough pieces traded off
May 20, 2016
0
#634
The post demonstrated a misapprehension, however. When a computer can "see" a tablebase position can be reached its evaluation is not small, it is a definite result. Before it can definitely reach such a position, it is the lines that do not reach tablebase positions that determine the evaluation, as they are the ones left from a minimax search.
Of course, it must also be added that depending on the tablebase structure an engine uses, it may not even play 6 men positions perfectly, as it may have only the final win/draw/loss evaluation of the position available without the moves that lead to that result
May 20, 2016
0
#636
s23bog,
The endgame tablebase gives an instant assessment, win, lose or draw, for any position with six pieces or less. There's a fairly complete 7-piece tablebase.
You don't need to reinvent the wheel. Those positions have already been "proven" to lead to specific results with best play. Once a game reaches the tablebase, the game is "solved".
May 20, 2016
0
#637
s23bog wrote:
...
These computers are not bringers of life. They are a sword. One with tremendous power. In the hands of the worst of us, they can do some horrid things.
This is a very strange statement coming from someone who wants to create the most powerful computer in history just to solve chess.
May 20, 2016
0
#639
It's hard for me to understand you (s23bog) sometimes. I wonder if there is maybe some language barrier or something?
SmyslovFan wrote:
s23bog wrote:
...
These computers are not bringers of life. They are a sword. One with tremendous power. In the hands of the worst of us, they can do some horrid things.
This is a very strange statement coming from someone who wants to create the most powerful computer in history just to solve chess.
If you read back this thread, you will find many more strange and contradictory statements from the same source, lol
Ok, yeah, should be something like "chess like games will be solvable in polynomial time if P=NP."
Why would you think that would be true?
Solving chess rigorously is like solving a chess problem rigorously, and requires checking even the unlikely looking defenses. Hence no massive short cuts reducing complexity.