	Generational Dynamics
	Forecasting America's Destiny ... and the World's

\| HOME \| WEB LOG \| COUNTRY WIKI \| COMMENT \| FORUM \| DOWNLOADS \| ABOUT \|

Generational Dynamics Web Log for 6-Dec-06
Deep Fritz computer beats world chess champion Kramnik, 4-2.

Web Log - December, 2006

Deep Fritz computer beats world chess champion Kramnik, 4-2.

The world champion fought hard but was clearly outplayed overall in the six-game match.

Kramnik vs Deep Fritz 10 - final scoreboard

Kramnik vs Deep Fritz 10 - final scoreboard

Human world chess champion Vladimir Kramnik played remarkably well, often playing deeply intuitive and positional moves that frustrated Deep Fritz's attempt to defeat him on a purely tactical basis. Indeed, he may have had a winning position in game 2, but played an incredible blunder that permitted mate in one.

Games 1, 3, 4 and 5 were all hard-fought drawn games.

Human world chess champion Vladimir Kramnik <font size=-2>(Source: SourceBase.com)</font>

Human world chess champion Vladimir Kramnik (Source: SourceBase.com)

In Game 6, Kramnik went all out to try to win a game. As Black, he played the Sicilian Defense, an opening that provides lots of opportunities for a Black win, but is loaded with highly dangerous tactical opportunities for both sides.

Frederic Friedel, spokesman for technical team that developed the Deep Fritz software, gave a prediction early in the game during the online real-time commentary led by Grandmaster Yassir Seirawan.

He said (paraphrasing), "I'm going to make a prediction of what's going to happen -- not what I want to happen, but what will happen. Kramnik will play an even game with the computer, but sooner or later he's going to become aggressive and try to win. If he does that then he'll make a mistake, and Fritz will take the game."

In an ordinary human versus human game, it's quite common for both sides to make tactical errors, and the advantage often passes back and forth between the two players. But "it's very hard to win against Fritz," according to Friedel. "To win the game, you have to beat Fritz on every move in the opening, in the middle game, and in the end game. Just one slip, and Fritz will win."

That's pretty much what happened in this match. Kramnik's sophisticated powerful positional and intuitive play constantly kept Deep Fritz at bay, even gaining a small advantage in three of the games. But in the last game, wanting to win at least one game, he took a big risk and played a complex tactical opening. He made a couple of mistakes in the middle game, and Fritz was able to capitalize on those mistakes, finally exchanging Queens and transitioning into a clearly won endgame for Fritz.

Friedel made a couple of other comments of great interest.

He runs the ChessBase web site, which permits people around the world to play chess online with each other. He said that they've developed "secret" technology that allows them to detect whether a player is using a home computer to play his games on ChessBase. This technology watches several of the player's games and, if repeated use of a computer is detected, then the user is banished from the system.

The current version of the computer software is Deep Fritz 10, with many sophisticated enhancements to earlier versions of the Fritz chess program. Earlier versions of the software were a lot weaker chess programs than the current version.

However, Friedel pointed out that even earlier versions of Fritz are very strong today. "Fritz 3 was a weak chess player when it came out, but it's a very strong player today because computers have become so much more powerful."

This emphasizes the point that I made last week that the sophistication of the computer software is much less important than the speed of the computer. With an older computer, Fritz 3 may only have had time to look ahead 5 or 6 moves, but with newer, faster computers, it might look ahead 10-12 moves, making it a much more powerful chess player, even though the software algorithms are the same.

According to Friedel, chess software running on a slower computer almost always plays the same moves as on a much faster computer, but there will be perhaps 2 or 3 moves in the whole game where the software plays a different move on a faster computer -- and those 2 or 3 moves make all the difference between winning and losing.

This remark shows the way how computers and humans can play in the future. It had been thought that once computers got to be better than humans, then they would quickly become so much better that there would be no point in a human ever trying to play against a computer again. But this problem can be easily fixed by "handicapping" the computer, and giving it less time to "think." In the match just completed, each side had two hours to make 40 moves, averaging 2½ minutes per move, and with this timing the computer was able to defeat the world champion.

Next time, we could allow the human champion to have two hours to make 40 moves, but limit the computer to only one hour for 40 moves. The would presumably weaken the computer's play enough to make it an even match.

For many years, I've heard people say that a computer would never beat the best human chess players, because computers don't have "intuition." The current victory puts that argument to rest, especially because Kramnik is considered to be the best "intuitive" player in the world.

This match also pretty much puts to rest the argument that "a computer will never be as intelligent as a human being, because the human mind has potentially infinite capabilities."

Actually, the human mind has nowhere near infinite capabilities. A human being getting through the day has only a few choices to make at each point: "Do I drive to work, or take the bus?" A soldier in a war also has only few choices: "Do I shoot at one of those enemy soldiers -- and if so, which one? -- or do I retreat?"

Computer chess software today uses the same "minimax algorithm" that the first computer chess program used in the 1960s. This algorithm says, "I can make moves A, B, C or D. What will my opponent do in each case? If I do B then will my opponent do X, Y or Z? In each case, I want to MAXIMIZE my chances, and MINIMIZE my opponents' chances." This analysis can continue to any depth, depending on the time available.

A super-intelligent computer will be able to "think" using a similar minimax algorithm. "What will happen if I do A, B, C or D? If I do B, then either X, Y or Z will happen. What do I do next in each case?"

Of course there are more choices to make during a day than the number of moves in a chess position, but not infinitely more. Computers double in computational power every 18 months, and by 2010 or shortly thereafter, computers will be more powerful than the human brain.

By 2020, we'll have intelligent robots doing chores that humans don't want to do -- cleaning up waste sites, 24-hour nursemaids, and so forth. Within a few more years, intelligent robots will also be doing scientific research to develop improved versions of themselves, so that intelligent robots will eventually be far more intelligent than human beings. The point in time where intelligent robots are essentially in control of their own destiny is called "The Singularity," because there will be a bend in the exponential growth technology curve. There is no way to have any idea what's going to happen to the world after that point.

The development of a world champion chess playing computer lights the pathway that will lead to those events. (6-Dec-06) Permanent Link
Receive daily World View columns by e-mail
Donate to Generational Dynamics via PayPal

Web Log Pages

Current Web Log

Generational Dynamics Web Log for 6-Dec-06 Deep Fritz computer beats world chess champion Kramnik, 4-2.

Web Log Pages

Generational Dynamics Web Log for 6-Dec-06
Deep Fritz computer beats world chess champion Kramnik, 4-2.