Tenth Anniversary of Kasparov-Deep Blue
May 2007 is the 10th anniversary of the epic man versus machine chess match throw-down, when Grandmaster Garry Kasparov went head-to-head against IBM supercomputer Deep Blue.
In a series of six games played between May 3-11, 1997, Kasparov was beaten by the machine, 3.5-2.5, the first time a computer had ever defeated a reigning world champion in a match under standard chess tournament time controls. Kasparov had played an earlier iteration of Deep Blue in 1996, winning handily by a score of 4-2, but IBM engineers had heavily upgraded the computer for the rematch; furthermore, the rules allowed IBM to modify Deep Blue's programming between matches, to adjust for weaknesses in the computer's game that became apparent as the match progressed. Kasparov was also denied access to Deep Blue's earlier games; the Deep Blue programmers, on the other hand, had access to hundreds of Kasparov's matches.
After the loss, Kasparov claimed he had detected characteristics in the computer's game that led him to believe human chess players were responsible for the computer's moves. IBM denied cheating, but refused Kasparov's demands for a printout of the computer's log; they also declined his offer of a rematch and instead retired Deep Blue.
Ten years later, many programs designed to run on an average PC can defeat most master players under standard tournament conditions, while the best commercial programs have surpassed even world-champion-level players at blitz and short time control games. Programs like Rybka, Shredder and Fritz now consistently compete well against top human players, while the development of chess-playing software has become something of a competition in itself, with numerous computer chess tournaments now being held. Some observers believe that within a few years, computers will consistently defeat even the best human chess players.
Until his retirement in 2005, Kasparov continued to challenge himself against computer opponents, with moderate success. In January 2003, he played a six-game match against Deep Junior that ended in a draw; in November of the same year, he faced off against a version of the chess program Fritz with similar results. In June of that year, Mindscape released Kasparov Chessmate for the PC, Mac and Palm OS.
I just finished reading The Singularity is Near by Ray Kurzweil, who goes on a bit about Deep Blue. Supposedly there exists a chess-playing algorithm now that is so much more efficient than Deep Blue that, running on something like just ten off-the-shelf PCs, it is ranked similarly to Deep Blue. It has something to do with being better at "pruning" the bad moves, so that it can see farther down the consequences of good ones, and faster, with less hardware. So it's not just that computers are getting beefier - the programs themselves are improving.
Articles I've read say the same thing: unlike the days leading up to Deep Blue, when playing a better game of chess meant building a more powerful computer, the "brute force" method has increasingly fallen by the wayside in favour of better, more efficient software. Given that it's a hell of a lot easier and cheaper to develop better software than to build a bigger supercomputer, it's no great surprise that we're seeing development head in this direction. Unfortunately for Garry and his fellow Grandmasters, however, the increased ease and availability of that development path means that the end of their superiority to chess-playing computers is going to arrive much, much sooner.
At what point does being a good chess player begin depending on one's ability to beat a machine? John Henry comes to mind.
I'm conflicted between the skill apparent in the engineer and the programmer who devised a software capable of beating Kasparov, and the skill of a trained human mind capable of keeping up with machines for so long whose sole purpose is to play chess. Eventually man is going to lose every time, but it was a good effort up until that point.
Does this take away from how incredible Kasparov's skill is? Does it remove impetus for anyone to try and be the best they can at (insert game of choice here)? Sure, it's a game, but there's got to be something we're good at. Why try so hard, when you always know that a computer can do it better, and faster? I guess for mental exercise? For pure pleasure? As long as your goal is never to be the best in the world.
The other question I have is, at what point do players emerge who are horrible against other people, but are able to pick the most computationally expensive strategies against a chess-algorithm to ensure a win? That's a whole new sort of skill.
In the end, all of this furthers the progress of programming and engineering, and it raises new challenges for the human mind to tackle.
After all, if we're to succeed against our eventual robot overlords, we need to learn a new skillset for exploiting their Vulcan-ish logic-only ways.
Unlike the algorithms for playing most other board games (especially go), chess-playing programs aren't all that different from a human player, with one difference: they are not susceptible to the psychological aspect of play, and they cannot (yet) exploit this weakness in an opponent. The only advantage that being a human can give over a machine is if one is extremely bad at psychological intimidation, or extremely vulnerable to it... both of which constitute just a lack of a disadvantage.
Not like that's any reason to stop playing chess, even competitively. Just because a car can go ninety miles an hour doesn't mean people have stopped running.
Very true, people continue to try and be the fastest people on earth. And, to continue your point, just because jets go Mach 3.3 doesn't mean that people have given up on land speed records. I guess the shift is just against that which we measure ourselves against. Kasparov and his ilk have to give up being the best players of chess in existence, and resign themselves to being the best biological/organic players of chess in existence.
Masters of the universe, as long as the universe is appropriately defined.