Computer Questioners
By Joel Benjamin   
February 15, 2007
Hey Joel,

Now that the computer has again beaten the world champion of chess (Fritz-Kramnik), isn't it time that we reevaluate this whole man vs. machine thing? If a human opponent is caught during a game using a computer to assist his play then that is cheating. So isn't the computer the ULTIMATE CHEATER?

If a computer can refer to databases on the hard drive and use multiple processors to calculate a move then we as humans should be able to get some assistance when playing against a computer. Instead of a one on one battle against the machine (which really isn't fair), why not the machine playing against the world champion plus two seconds (two helping grandmaster or masters), and an analysis board. if you think about it, the world champion would be the main processor, the two helpers would be the secondary processor doing analysis, and the analysis board would be like the ram in the computer. Man vs. machine should not be an individual man but rather a team of people, because in real life, we as humans can only achieve the ultimate goals of life when we work together as a team

Tony Wong

and

Dear GM Joel,

I just read through the rules for GM Kramnik's match with Deep Fritz. It reminded me of Kasparov's complaints and demands for the Deep Blue match as chronicled in the "Game Over" docudrama. It seems that Kramnik is getting even more from the Fritz team than Kasparov asked for. What do you think about the rules for the match?

Sincerely,
Don Brooks

The Kramnik-Fritz match has refocused attention on man vs. machine competition, particularly the issue of "fairness." In answer to Tony, I would start by arguing that these matches are meant to be, to some extent, "unfair." Of course humans cannot match the awesome calculating power of the computers. But if you go back ten years, the human advantages of experience, world knowledge, and pattern recognition were actually considered more valuable. Whereas man vs. machine was once a test of programming skill, it is now clearly man fighting to triumph against the odds. We may now need fine-tuning to even out the chances, but if the human doesn't play alone the match seems pointless.

Issues of fairness have historically erred on the side of the human. This has been the case largely because the companies behind the programs benefit so much from these matches while the human grandmaster puts his reputation on the line. Even when humans were still favored, programmers were expected to give the grandmaster a version of the program to practice against.

In Kasparov-Deep Blue, IBM took the position that fair meant leaving both sides free to bring their strengths to the table. Since Kasparov was meant to be the favorite, it didn't seem necessary to give him any competitive advantages. Their philosophy was logically defensible but not very popular. Kasparov repeatedly objected before, during and after the match. Most fans seem to feel IBM should have made a concession somewhere. Since the public sympathizes with the human player, it makes sense for the computer side to be as accommodating as possible.

Humans have increasingly become viewed as the underdog since Kasparov's loss to Deep Blue. So the grandmaster can ask for more and more. The recent Kramnik-Deep Fritz match gave peculiar rights to "mankind." You are absolutely correct, Don - Kramnik got much more than Kasparov even asked for! I will summarize the most important regulations:

1) The Fritz team was limited in the alterations they could make to the program's book (predetermined opening moves). Any such alterations had to be made in the presence of the match arbiter. Kramnik was also given special access to the computer's "thinking." I quote from the match rules:

As long as Deep Fritz is "in book", that is, playing moves from memory and not calculating variations, Mr. Kramnik sees the display of the Deep Fritz opening book. For the current board position he sees all moves, including all statistics (number of games, ELO performance, score) from grandmaster games and the move weighting of Deep Fritz. To this purpose, Mr. Kramnik uses his own computer screen showing the screen of the Deep Fritz machine with book display activated.

2) Kramnik had special rights concerning Fritz's use of endgame tablebases which allow a computer to play perfectly with five pieces (including kings) or less on the board. If a position arose in the game that appeared in the tablebase, the Fritz team would have to inform Kramnik and tell him the evaluation of the position. If the position were a theoretical draw, Kramnik would have the right to accept a draw. In some cases this could be an enormous advantage. For instance, Kramnik would have been safe in rook vs. rook and bishop. Most positions are theoretically drawn but often lost in practice. Here Kramnik would not have to risk making the decisive error.

3) If a game lasted six hours or 56 moves, Kramnik had the right to adjourn and continue on the next rest day. This stipulation is particularly bizarre, because Kramnik could consult any computer program for help - perhaps even his opponent!

4) The Deep Fritz team had to provide Kramnik with a printout of the computer analysis of the game upon completion.

These concessions are in one sense decidedly unfair. Should a player be allowed to claim a draw because his opponent knows exactly how to win? Should he be allowed access to his opponent's opening plans during the game, and see the rest of its analysis afterwards?

Yet these rules are fair in the sense that they still provide for a competition with reasonable chances for both sides to win. Rule 1 is not easy to take advantage of. Rules 2 and 3 are big concessions, but only if the game goes far enough. Rule four references the infamous "logs" from Kasparov-Deep Blue. As someone who has the logs from that match, I can tell you that they are unlikely to help very much.

I'm sure that in future matches the rules will be even more skewed towards the human competitor.

Joel Benjamin