View Question
Q: Humans v computers at chess. ( Answered ,   3 Comments )
 Question
 Subject: Humans v computers at chess. Category: Computers > Games Asked by: turtlesallthewaydown-ga List Price: \$10.00 Posted: 28 Oct 2002 01:48 PST Expires: 27 Nov 2002 01:48 PST Question ID: 90851
 ```How long will it be before computers beat every human player? I'd accept an answer to within a year with a mathematical proof. Please note the wording of the first sentence.```
 Subject: Re: Humans v computers at chess. Answered By: hailstorm-ga on 28 Oct 2002 04:35 PST Rated:
 ```turtlesallthewaydown, For your question, I will assume that by "human player" you mean one human acting alone, and that by "computers" you mean the most powerful chess playing computer available. I will also assume that the World's Champion of Chess is the greatest human player alive, and that no other player can defeat him (and even if he does, that player would then become the World's Champion of Chess) So, rephrasing the question with these assumptions in mind, how long will it be before the best chess program can defeat the World's Champion of Chess? If you want the answer to 100% certainity of a computer beating a human, the answer is never. The number of possible chess games has been estimated at 10 to the 10th power to the 50th power. Even a computer that could process a trillion trillion moves per second could never process all possible moves in a long match within the time limits of an official chess match. If you want a realistic answer of computer dominance, the answer would appear to be fairly soon. In fact, just recently the "Brains in Bahrain" battle between chess World Champion Vladimir Kramnik and the reigning most powerful chess program, Deep Fritz. In an epic eight match struggle, Deep Fritz fought back from an early deficit to tie Kramnik. To emphasize the importance of this draw, Kramnik himself was quoted as stating, "It is now clear that the top program and the world champion are now approximately equal." Now, chess was invented over 1,300 years ago, and humans have been devising strategies ever since. So much of chess strategy has been fleshed out that there remains very little left that man can do. In addition, his actual brain capacity increase since those early times has been very little. Compare this with computers. In a fraction of that time, computers have evolved to the point where they can make billions of calculations per second. Moore's Law states that computer processing time will approximately double every 18 months. Because this kind of raw processing power is something humans cannot do, it opens up all sorts of new stratagies that can be devised. So, while humans remain largely stagnant, computers are doubling in processing power and increasing their strategic horizons by at least double every 18 months. And currently, in a match between the best human and best computer, there is a 33 1/3% chance of computer victory, 33 1/3% chance of human victory, and 33 1/3% chance of a draw. In 18 months, the computer will be four times stronger, while the human will be roughly the same. Thus, the computer will become four times more likely than the human to win a match. Over an eight match series, the chances of a human player pulling off a draw, let alone winning the series, become microscopic. After another 18 months, the best human playing prowess will be 16 times less than the best computer. Although it is estimated that there are 10 to the tenth power to the 50th power possible games of chess that can be played, and computers will never be able to defeat a human with 100% certainity, for all intents and purposes victory for the human will become almost impossible at this point. Because we are almost at the end of 2002, I will say that, within a year of 2006, no human will stand a realistic chance of defeating the best computer chess program. Sites cited: Brains in Bahrain http://www.brainsinbahrain.com/ Chess from Math World http://mathworld.wolfram.com/Chess.html Silicon Showcase http://www.intel.com/research/silicon/mooreslaw.htm Moore's Law http://www.webopedia.com/TERM/M/Moores_Law.html History of Chess http://math.uww.edu/faculty/mcfarlat/177hist.htm Google search terms used: Moore's Law History of Chess Brains in Bahrain Deep Fritz```
 ```This is unquantifiable, as not every chess player will play against a computer. If you want to know when they "think" computers will always be able to beat the current -best- human being, then that is something there have been studies on. Maybe you'd like to reword your question? willie-ga```
 ```Yes, willie-ga is right . First there should had been a mathematical definition of what's a mathematical proof, then --perhaps-- the penalty of one star could have comne. Money isn't everything, right willie-ga?```
 ```I'm afraid I have to disagree with hailstorm. The improvement seen in a chess program due to an increase in processing power is not a linear relationship, and we are already at the point of seeing decreasing returns with faster processors. To say, "In 18 months, computers will be 4 times stronger." is frankly absurd. All the top chess programs use some more of an ever-deepening mini-max search to "see" so many moves into the future. The problem is, at each new level of depth, the complexity of the problem increases exponentially. After a certain depth, you can have a computer go 100 times faster and not be able to search any deeper -- there are simply too many combinations. If you don't believe me, simply take any chess program and run it on Pentium II computer and rate it some how. Now run it on a Pentium 4 (at least 4 times faster, probably 20) and see if your rating for it quadruples. I don't think so! Now, there are all sorts of strategies to continue to improve chess programs, but most of them won't come "free" just because the computers get faster -- they will come as the result of a clever programmer trying something new. In fact, we will probably have to make "shallower" searches to make the next big breakthrough in chess programming to better use all the processing power. Basically, you can't predict a date at which computers are going to be able to beat every human player and prove it with any scientific rigor. You can only make a guess. The original question is like asking, "When will there be no more war on Earth? Please provide a mathematical proof."```