The future of chess
by George Dvorsky
Sentient Developments
2006-12-08
Now that the RAG Tournament featuring Vladmir Kramnik and Deep Fritz has concluded with the machine emerging victorious, it’s time for some contemplation about the current state of chess and its future.
First off, credit where credit is due. The Fritz team put together an excellent program that defeated Kramnik quite soundly—an achievement made all the more impressive by the fact that Kramnik’s style is considered quite difficult for computers to handle. Fritz v.10 may be the best chess playing entity of all time (not including Hydra which hasn’t been pitted against a world champion). Yes, Kramnik blundered away Game 2, but that, like fatigue, stress and over-confidence, are indelible parts of the human condition—intangibles that don’t apply to machines. Moreover, Kramnik never really threatened Fritz and lost quite soundly in game 6; he played for the draw in virtually every match and never assumed he could win.
And this is where we find ourselves at the end of 2006: the best chess playing humans, it would seem, are now unable to beat the expert chessbots. It’s at the point now where the victory is in the draw. The day is soon coming where even this may become impossible, but a part of me believes that chess is too complex for today’s computers to achieve consistent victories. Computer power and programming will have to be more powerful by an order of magnitude before computers can generate algorithms that will result in perfect play resulting in perpetual streams of victories.
This is assuming, of course, that chess could be solved in the strong sense. I’m not convinced this is possible. In the vast space of all possible games, perfect play in chess will lead to draws more often than not. The most solved that chess could ever become is in the weak sense where an algorithm can secure a win for one player, or a draw for either, against any possible moves by the opponent from the initial position only.
The full article can be read here.
Do you agree with George Dvorsky? Does Kramnik’s play represent the entire human race?
Do you agree with George Dvorsky? Does Kramnik’s play represent the entire human race?
Quite obviously not. We can have no idea that somewhere in India, or in a venezuelan village there isn’t a super genius who may not even know what chess is, yet if he knew he could cream both Kramnik and Fritz. But this is theory, not practice. Right now, it seems, that the current generation of chess computers are better than the current generation of active chess players. And here comes a potentially interesting thought:
Up until now the chess playing computer programs kept improving, because there was a GOAL, and that goal was to develop a program which can even defeat the world champion. Which currently happens to be Kramnik. What if (and this is an old dispute), what if Fischer could defeat both Kramnik and Fritz? What if Fischer (or any of the previous best chess players) were better than Kramnik? We don’t know that.
But at this point the chess program developers have no new goal and there is really not much point to put money and effort to develop an even better chess program.
Therefore, I find it quite possible that X years down the road a new chess genius appears, and defeats all the computer programs. Will the programmers come back and after perhaps decades of time gap start working again just to have the computer beat that one genius? I doubt it, but not impossible.
————————
The unreasonable future:
There will be no absolute chess computer. That has been calculated long time ago. In order to calculate all the best chess moves from d2-d4, allegedly has more moves (permutation of the number of moves throughout an entire chess game) than the number of atoms in the known universe. We (mankind) are not even close to have such computer and according to some experts, never will.
The reasonable future:
We will have chess computers and human chess players will quickly break the “habit” of challenging the computers in chess. Instead, they will return to playing chess against each other, as it was for thousands of years. The computer will be there as a teaching tool, for analysis, for practice, for lonely humans and so on. As so many people stated, weightlifters don’t compete against cranes, 100 meter dash champions don’t try to run against a race car, and so on. The exact same attitude will soon develop regarding chess computers. And I think that is normal and should be that way.
The picture of a toilet floor in relation to the future of chess, a coincidence, I assume?:)
Computers are now officialy better than humans at chess, it was only a matter of time, they’ve been better at mental arithmatic etc. for a long time. Givin the rate of increase in CPU performance, it’s likely that computers wont be beaten again after the next World Championship cycle. Lets get over it, cue forklift and racecar arguments. It’s still a great game and a great mental challenge for us humans. One thing computers don’t yet have is a competative spirit, Fritz gained no joy from it’s victory, we still can, and will be able to for the forseeable future.
the article seemed very prejudiced. maybe written by a chessbase supporter. Rybka is considered the strongest computer program by virtually everyone. He skips over rybka.
Actually the match was disappointing for the computer. it showed that it did not really understand chess. it is able to make moves without errors and thus win the game. but it does not display understanding of the game itself.
I think the programs still have a long way to go. Without the one move error in game 2, the match would have ended in a draw. Kramnik played the sicilian as a deviation to his match strategy. A silly human mistake. I thought he should play the last game the same way as he had played the first 5 games. a draw would have been acceptable in the last game.
The point is that we humans made the machine that can calculate chess variations better than we can, just as they can calculate square roots faster than we can. So what ? Did we think that chess was so sacred ?(Go players, it’s your turn next, just a matter of computational speed :-))Chess is essentialy a mathematical problem, not a human centred one (Like ‘great litarature’) so the speed of the computer is the limiting factor. Chess has always been interesting because of the human vs. human stuff, not because of human vs. ‘All Possible Realities’ ??
As Susan discussed, perfect chess inevitably leads to a draw, the rules evolved that way, because it was a game for _humans_ to play, and enjoy.
Eventually, computers will be able to “solve” chess. Until then, it appears they will be soundly beating even the Grandmasters. (Let’s face it, computers have been beating the rest of us for quite a long time.)
But I no more worry that a computer can beat me than I worry that Susan can trounce me. I enjoy playing chess against my equally-unskilled human friends. I’ll be honest, I’m a poor chess player, but I love playing.
My car can outrun me.
My calculator is better at math.
My wife is a nicer person.
My computer can beat me at chess.
Life is rich and we gain, not lose from it. Pretty cool, eh?
Paul
Is George Dvorsky a ChessBase stooge, or what? Well, I mean he must be!
“Fritz v.10 may be the best chess playing entity of all time (not including Hydra …”
LOL. Yeah right.
What planet has Mr. Dvorsky been living on for the past year? Rybka is top. Rybka *destroys* Fritz 10 (and Deep Fritz 10), and every other CB chess engine for that matter. Oh, poor ChessBase… once upon-a-time they had near total-control of the top-quality chess engine market, but then unfortunately comes along Mr. Rajlich and rocks the boat bigtime with his fantastic creation: Rybka.
Allow me to repeat that wonderful word: Rybka. RYBKA. *** RYBKA ***.
ChessBase just cannot bring themselves to acknowledge the mere existence of wonderful Rybka.
Thanks for reading, and have a nice day.
PS.
Lovely Rybka. Wonderful Rybka.
Destroys Fritz.
Obliterates Shredder.
Nukes Hiarcs.
Torpedoes Chess Tiger.
I prefer to let people make foolish statements such as “I believe this is not possible” and then have a good laugh a few years later as someone figures out a simple way to do these things.
Remember, Fred Friedel is the guy who said out loud in a movie in 2003 “Computers cant do that” regarding finding the super move deep blue made in game 2, and now his company’s product Fritz “does that” in about 2 seconds.
Please, keep telling us what is not possible, since I look forward to seeing someone do just that very thing you said.
Paul said…
Eventually, computers will be able to “solve” chess
That’s just it. Each ply is an exponential increase in the number of needed calculations. Now simplify that and assume that there are 10 possible moves on the board (every time). Assume a 60 moves game (120 ply). In order to “solve chess”, in this imaginary game for the computer would need 10^120 calculations.
Even this primitiv assessment makes it obvious that we are very, very far from “solving chess”.
Lovely Rybka. Wonderful Rybka.
Destroys Fritz.
Obliterates Shredder.
Nukes Hiarcs.
Torpedoes Chess Tiger.
Chess is not soccer.
Can you defeat Chessmaster 2100 on 3 minutes thinking time?
Gabor
As others have said computers don’t really strategically ‘understand’ chess but it is of no consequence since are able to beat the best humans almost entirely through the sheer depth of their tactical strength due to the ability to examine millions of positions per second.
But computers won’t be competing with professional Go players for a very long time to come. Despite many years of trying and increasing resources (ex-chess programmers, money etc) being thrown at the problem, the best Go programs are still pathetically weak and show no sign of any more than very slight improvement. Even if Moore’s Law continues indefinitely, its been calculated that it will take thousands of years for computers to do to Go what they have done to Checkers, Chess and others and win by sheer computational power.
To beat the best Go players the programmers will have to develop pattern matching algorithms that match humans’ fantastic natural abilities. That may happen over the forthcoming decades as the understanding of the human brain advances and AI techniques develop but its going to be a long hard road.
Actually, there is another aspect to this that worries me more. If we ever get programs that can compete with the best Go players it will be because computers are close to making humans obsolete in general and not just in board games!
A few random thoughts:
If there is even partial truth to the adage “you learn more when you lose than when you win” (it can’t be completely true, otherwise I would be world champion by now) then we all have something to learn from the recent Fritz/Kramnik match.
If chess is a sport and worthy of being including in the Olympics, then do computers even have a place ‘competing’ with humans. Obviously as a tool or learning device they certainly can be helpful, but that doesn’t mean they should be used in match play.
Dvorsky states: “as far as the advancement of chess is concerned, it is time for humans to take a backseat to the computers.” I don’t agree with this. Humans can’t memorize or calculate as fast as a computer but that is not the whole of chess. A previous poster stated that Fritz won but clearly didn’t understand the game. I think this is partially true. Computers don’t understand many aspects of the game that humans find attractive. Fun, joy, art, beauty, etc. draws many to chess before memorizing opening books or endgame study. Many human players would rather play (or kibitz) a line that is “beautiful” over a line that is more direct or produces a forcing win. Many (and I’m included) would rather sac a Queen and win in 15 moves compared to entering a mate in 9 line without any “brilliancies.” Computers may be able to appear to write music or make pictures or play chess, but I’m not even partially convinced that computers are great artists.
Anyway, that’s my two cents for now.
-mm
The future of chess is not in doping controls.
Man must reinforce technological controls instead of doping…
The danger now, is more in an informatic cheat than in a medecinal one.
US championship is showing that: two big cheaters, and only one caught red handed… the second one is still running freely… There was also a cheater in the indian championship too…
As usual I disagree with most everyting. Apart from that i have a sugesstion; If you take a human from the top 100 list like Kramnik and a program from the top 100 list as Fritz, why not use a computer from the top 100 list?
http://www.top500.org/list/2006/11/100
One Anon said:
[i]But computers won’t be competing with professional Go players for a very long time to come. Despite many years of trying and increasing resources (ex-chess programmers, money etc) being thrown at the problem, the best Go programs are still pathetically weak and show no sign of any more than very slight improvement.[/i]
That was exactly the reasoning about chess in the 70s and 80s. Go is obviously more difficult due to the sheer calculating power requirement.
Larger board, more pieces. Due to the more basic popularity of chess, and the fact that chess is popular on the same geographic areas where high tech development took place first, serious programming efforts went into chess first. Theoretically there is nothing inherently more sophisticated about Go than chess, which would make it less possible (or impossible) to write champion level Go programs. If there will be demand, there will be Go program which eventually defeat the best Go players. That’s exactly what I said about chess programs at the time when everybody insisted that “there will never be a chess program which defeats the best chess players”.
Actually, there is another aspect to this that worries me more. If we ever get programs that can compete with the best Go players it will be because computers are close to making humans obsolete in general and not just in board games!
This is a sci-fi style “the robots are coming” type of concern. It is not a competition, and it never will be. What do you mean “making humans obsolete”. Humans, as such ARE obsolete. The universe would exist just fine without humans, without life as we know it, because we don’t know it :). But we humans do exist, and there is no logical point to even bring up something like this. We shouldn’t try to compete with our own “creations”, we should make them “serve” us. I don’t see Deep Fritz “defeating mankind”. I see DF (chess computers in general) as a clever creation of mankind, which we should be able to utilize to our own advantage. If we don’t create a machine-men competition (in general), there will be no competition, no winners and no losers. Other than mankind will be the winner of simply having such wonderful machines.
Gabor
the final solution for chess will be BIG thing (if it happens). big enough for the programmers to go on building stronger and stronger engines even after they have managed to beat humans. actually they managed to beat humans long time ago already.
if a program gives a hard time to the world champion, even if it loses, does that not mean that a simple out of the box program for a couple dollars IS stronger than humans? so what if the top 5 players on the planet give it a hard time? it can still beat millions of players without half trying hard – computers ARE better at this game.
i dont know if chess is impossible to solve completely. by sheer tactical computation – checking positions, maybe. but there may be algorithms within the game that could work without understanding the game. i remember reading about a chess prodigy, when asked why he made a certain move. the child replied “i had to make that move or i would have lost” – without knowing the concept behind the move, let alone seeing a forced losing line.
Kasparov in his standard shape would be able to win at least one game vs DF10.However,I think he would also loose at least one game due to his attacking stile.
Unfortunatelly he doesn’t play chess any more.
R.J.Fischer?
He doesn’t play normal chess any more too.
But he plays Chess960!
Now,do you think he is dumb enough to accept match vs DF10 like machine under Chess960 rules?
No,he is smart enough to know he would loose (not just becouse he is old).
I have a divergent view to spout forth here. The World Champion became World Champion by playing and beating humans. For his whole career Kramnik has trained and focused on beating humans. Not tatical monsters called computers. A human brings a lot more varied thinking patterns to the board then a computer calculating tatical moves in a linear logical matter. Think for a second what if Tal played Deep Blue or Deep Fritz? How would their short-term tatics done against his deep sacrificial combinations?
Now that two World Champions have been beaten I have a thought that there will be a challenge to the human spirit; defeat the computer. It is possible. For one a new generation is growing up using the computer engine as a tool. They will have insights into the ‘mind’ of the engine that an older player like Kramnik does not have. Second, beating or even drawing the computer will carry more prestige than winning over the club champion. Third, not much study has been done on how to beat a tatical monster compared to study on how to beat a human. Different patterns of thinking will have to be invented. Maybe thinking 20 moves ahead will have to be a skill developed, a skill essential to think strategically deep enough to outweigh the tatical monster.
Every machine has a weakness. We have not found a significant weakness in a computer to beat it yet, or maybe we have not found significant strength in ourselves yet. Either way it is beatable if we find a way.
Hey, Gabor, you have a point. If all people in the world would play chess 10 hours a day, we could find a brilliant person who may beat the computer. However: if half of the world would play chess and the other half would write chess playing softwares, then probably the best chess software would still beat the best player.
JakeDP – yes, analysis how the computer thinks may help a little to defeat it but not much. For example, I know how computers take the square roots of numbers but that does not mean that I can take the square root that fast.
The best chess playing software is not Fritz but Rybka is, see http://web.telia.com/~u85924109/ssdf/list.htm
(not including Hydra, which runs on a super computer). I would enjoy a human-computer match, even if the computer wins all games, because the QUALITY of the match will be high. (The problem is that the losing grandmaster would not enjoy it that much.) Giving the computer less time would RUIN the very high quality match. Then it would be a match like any other grandmaster against grandmaster match. (Whick is high quality but not that high as the computer plays.) I rather give a pawn advantage to the human player (and dont decrease the time), and see a high quality match.
Who else has the feeling that Kramnik (or many other top GMs) could easily win a correspondence match against a program (say Rybka), even without access to computer assistance? The point being that the time factor would only help the computer’s play a little bit, but with a human it would help immensely.
Of course, this is just a hypothetical situation, but say Kramnik agreed to play a seven-day correspondence match against Rybka, and he would have to live in a hotel suite without computer access. Sort of like Nickel-Hydra, but the human cannot get help from a machine, so it’s truly man vs. machine. I think Kramnik would win.
Once you get rid of the psychological advantage of the computer by extensively (but not unreasonably) lengthening the time controls, it is the unaided human who will perform better than the machine. We may not be able to calculate as well under pressure, but we still understand the game far better than the machine, which ‘understands’ nothing. Even Rybka cannot yet understand simple strategic ideas which any human can easily grasp, eg. no engine can solve (give the correct line as its first preference) this study:
N7/1r2p3/rp1p2/bkp1PPp1/1p1P2bp/PBP1KP1P/PP6/8 w – – 0 1
White to play and draw.
No one would be insane enough to try and win a blitz match against Fritz, under Kramnik’s conditions. The time control is just unfavourable to the way we play chess. Although Fritz definitely did not demonstrate its superiority to the human race in chess in this match, there will come a time when computers will simply perform better OTB. So what?
Chess OTB is a sport for humans. It is not about perfection, it is about matching up whatever two human minds can come up with on a chessboard under tournament conditions, the fact that the computer algorithm produces more perfect results under these conditions is irrelevant. Computers are basically an analysis tool, and a very good one at that, but it does not mean that their calculations can supplant our human knowledge of the game. And calculations never will replace understanding, not until the day that chess is solved.
But if we are talking about ‘chess truth’ and want to really find out who is absolutely better at playing chess (choosing moves in positions, not limited by time or external conditions), then correspondence chess is the ultimate test, and such a test will reveal that, even if we assume that what computers do can be called ‘playing’ the game, we can ultimately still do it better than them.
And no, Kramnik’s play certainly does not represent the whole human race. Between six billion human beings I think we should be able to work out a mate in one 🙂 On another note, why has no-one asked Anand to play a match against a machine yet? Just because he has never clearly shown himself as the human No 1, doesn’t mean he wouldn’t do better than a World Champion against software. I think Anand would have won in Kramnik’s place.
has this guy actually bothered watching the games live? or he just saw the news of a draw or loss and went through it as quickly as possible on the java screen?
Sanyas, just look at the history of chess: first people said that the computer will never beat a strong club player, then once it happened they said it will never beat a master, once it happend, it will never beat a grand master, and finally it will never beat the world champion. Now you continue this tradition by saying the humans “understands” the game better and the computer can never beat them in correspondence chess. I must laugh. I am a strong player but the computer could wipe me off the board even in correspondence chess. A GM may have some chance today in correspondence chess, but wait another 5-10 years and they will have no chance AT ALL.
“Does Kramnik’s play represent the entire human race?”
No, he represents the best of the human race, excluding bloggers in here. Kram is the winner of this uneven match.
” It is not about perfection, “
Of course chess is about perfection, why would you think it’s not??
To fix the FEN in Sanyas’ post, it was clearly meant to be
N7/1r2p3/r2p1p2/bkp1PPp1/1p1P2bp/pBP1KP1P/PP6/8 w – – 0 1
That is, the Black Pawn on b6 was supposed to be on f6, and instead of a 9th White Pawn on a3, the a3-pawn should be black. Chess programs let you copy the FEN of the current diagram to a clipboard for pasting—whereas if I give the pieces in normal (English) form, I might err. OK, here goes:
White: Na8, Bb3, Ke3, Pa2,b2,c3,d4,e5,f5,f3,h3
Black: Kb5, Ba5,g4, Ra6,b7, Pa3,b4,c5,d6,e7,f6,g5,h4.
The point appears to be that White can win both Rooks by 1.c4+ Kc6 2.Ba4+ but loses the resulting opposite-Bishops ending because Black gets connected passed pawns. Whereas the clearly (to a human) drawing line throws even more material away but locks up the whole board. Indeed, Fritz 9 likes winning the Rooks and has no clue.