Rise of the machines

(Written in 2006)

This is my second time I write about chess on this website. But please don't get me wrong. I have not played any game of chess for at least a year. I do follow chess news as a hobby. Most of the time I never had any conversation related to chess. Funny enough, when I went to Math conference at Madison, Wisconsin last April 2006. My colleagues recognized me as a chess guy since they have read my website. I thought I made it clear that I like math a lot more than chess. Anyway, ...

Development of computer chess:

The very first chess progarm started around 1956 since the beginning of computer. We can say that it has been a dream about computer playing chess for a long time ago as you can see in many science fictions. People usually don't talk much about this period since it is boring. I can beat any chess programs in that period while I give Calculus lecture. Things started to get interesting in the early 1980's when there was the first computer program that beat chess grandmaster. Deep Thought was the first quite strong and popular chess pragram. Deep Thought borned in 1985 and named after a fictional computer in Douglas Adams' series, The Hitchhiker's Guide to the Galaxy (I have at least 3 of my friends in the department have 42 as their favorite number becuase of this fiction). Later the creators of Deep thought, Feng-hsiung Hsu and Thomas Anantharaman, went to IBM and changed the name of his program to Deep Blue. However, it was not until February 1996 that computer had a first win against the world chess champion. I remember I was watching that game live on the internet. Anyway the world champion at that time, Garry Kasparov, won the overall match with score 4-2. Only a year later Deep Blue came back and had a second match with our same champion. This time, Deep Blue is faster and stronger. Deep Blue made a headline by beating Kasparov 3.5-2.5. The day after the match, IBM stock price went up significantly. Another funny thing is that at the press conference after the last game which the computer won in only 19 moves, Kasparov was very mad. He blamed his computer chess expert assistance for suggesting a cumbersome opening and demanded a rematch. Also after this game, this line of opening has been disappeared from the grandmaster level tournament. On the IBM side, it did not take a long time before them to disassemble Deep Blue. There was no rematch whatsoever forever.

Is Kasparov-Deep Blue match the end of the fight between human and computer in chess?

Plenty of people thought so (We can see how good IBM promotion team was!). In fact, it was just the start of the a real intense fight between human race and computer. To make a story short (since I have a lot more to say about other aspects of computer chess. I just want to give a minimum background of chess program.) I have a summarize of the important matches below :

1996: Garry Kasparov - Deep Blue : 4-2 Kasparov win 3, draw 2 and lost 1

1997: Garry Kasparov - Deep Blue : 2.5-3.5 Kasparov win 1, draw 3 and lost 2

2002: Vladimir Kramnik - Deep Fritz : 4-4 Kramnik win 2, draw 4 and lost 2

2003: Garry Kasparov - Deep Junior : 3-3 Kasparov win 1, draw 4 and lost 1

2003: Garry Kaparov - X3D Fritz : 2-2 Kaparov win 1, draw 2 and lost 1

2005: Michael Adams - Hydra : 0.5-5.5 Adams win 0, draw 1 and lost 5

Novermber-December 2006: Kramnik - Deep Fritz 6 games match: score ??? ( my prediction,human will lost 2 games and draw 4 games).

I want to make short remarks here.

1) Deep Blue , Deep Fritz and Deep Junior had nothing to do with each other. I meant they had different creators. They share the name "Deep" since they are all use parallel processing systems.

2) Kramnik was/is a very strong player. He usually rank number about #3 in the world. Adams was not in Kasparov and Kramnik class but he is in the top 10 for most of the time.

3) In my opinion, all the the programs I mentioned after Deep Blue are stronger than Deep Blue. These programs were running at a lower speed ( about 100 times slower than Deep Blue) but stronger since there has been a constantly development of chess algorithms. Chess programmers are getting better. They learn from their experiences.

4) It is not only the chess programs that get better, human are better at playing with chess program too. Fritz and Junior are a very popular commercial computer chess program, every grandmasters have them in their computers. Human can practice and learn which position computer good or bad at. Computer is bad at a long term manoeuvre type position. But computer are perfect at complicated position that full of tactics.

5) Hydra runs on a super computer and not a commercial software as Fritz and Junior that run on PC. In my opinion Hydra is the strongest computer chess program in the history but only a bit better than Fritz.

6) I don't think at the time Kasparov lost to Deep Blue, computer was stronger than human. Although I accepted that during the last 2-3 years started from 2004, computer has been significantly stronger than human.

The role of computer in chess and learing process:

I like to learn stuffs a lot and quite interested in the learing process. Computer in chess is a perfect example of how to speed up the learning curve. During the very famous American world champion, Bobby Fischer , time, he had to study Russian when he was in the middle school so that he can read chess magazines which were mostly written in Russian. These days, searching database of your opponent's games, learning opening or practicing the game with the program that are stronger than anyone in the world are just one click away. All the informations that you need is right in your home computer. That is why kids these days learn chess so fast. Today The record of youngest grandmaster is 12 years and 8 months old (Sergey Karjakin from Ukraine). I believe it will take any kids who have enough potential only 5-7 years to becomes grandmaster. To become a world champion will depend on gift, experience, motivation and hard working (there is only one world champion at each point in time. It is not that easy to be one). Note that The current world number 1 and 2, are at their peaks in term of ratings at the age 32 and 37. ( I just want to point out that it takes time to learn something).

My proposal of next man-machine match in the future:

Computer is too strong now. We can see from the result of Adams and Hydra last year (0.5-5.5). It was quite embarassing. Computer needs to give a human some handicaps. Here is my proposal

1) Human allow to consult an opening book and game databases. Computer are allowed to use opening book. Human should allowed to do so.

2) Human allow to consult 8-pieces endgame table base. Same reason as number 1.

3) More time control for human. Now human has about 3.5 half per game. I offer 10 hours for human per game plus brakes for lunch, dinner(about an hour each so that there is no rush when eating) and time to go to the rest room with the possibility of a nap in the middle. Computer can think 2-3 millions positions per second. I think this is fair.

4) Human have a weak chess program just to check blunder. Everyone make a blunder. We don't want to see a good game get spoiled by just one blunder move.

According to this I think human might have a fair chance to beat computer.

source: Wikipedia


12/5/06 : The match between Deep Fritz and Kramnik just finished this morning. Kramnik lost 2 games and draw 4 games (just exactly what I had predicted). The impression is that, in this match, they already used many rules that I proposed before for example: Kramnik can see fritz opening book , the time control is longer that the regular match and there is some special rules used for the endgame database of Fritz. I am sure the rule of the future match between man-machine is heading in this direction.

Kramnik actually played pretty well. But my impression is he had a chance to prepare for this match with exactly same version of program sent to him a month before the match started. In this case he has a very good control on opening theory. He can pick the position that he's comfortable with from home. If he had not missed mate in one at the second game, he might not have to risk by playing a bad (risky) opening on the last game and the match might have been tied (but who know anyway).

Let me make a stronger prediction here: Human will never win any single game against computer from now on in the regular tournament rule.

However human can still beating the machine if they are assited with other strong chess program and have a very long time to think ( to be more precise here, I should not say "think", I should say "checking through the variations").

Interesting Chess game