Slashdot Mirror
Kasparov Draws Game 4 and Match Against X3D Fritz
jaydee77ca writes "Garry Kasparov survived opening danger and played very precise, technical chess to draw Game 4 with black against X3D Fritz. The final match result is a 2.0 - 2.0 draw, proving yet again that the day of the machines has not yet arrived."
proving yet again that the day of the machines has not yet arrived
Sigh. Such an obviously human-biased conclusion to what is indisputably one of the great achievments of computer chess. The fact that Fritz, running on rather modest hardware, drew Kasparov, is an incredible feat. The obvious followup is that the days of a human world champion are numbered. And most likely that number is most conveniently expressed in months, not years.
Running on an Intel Xeon server with four 2.8 GHz processors.
proving yet again that the day of the machines has not yet arrived.
Didn't that already happen a few year back when he lost to Deeper Blue?
I really hate Dan Patrick.
Why man vs. machine is so important to us is a little baffling. While it might be nice for our egos, what does this really do for the game of chess? Does the challenge make people better chess players? Maybe. Should we consider this any more interesting than a normal game between grandmasters? The Terminator mentality somewhat bothers me, that we feel so insecure about ourselves that we have to congratulate people when they can do something better than a tool can! (Personally I root for the block of silicon ;-)
While the computer can be programmed to "look ahead" for N moves, the computer must also be programmed to pick a move eventually, what is programmed to be the "best move". And all this programming is done by humans. Voila!
Slashdot: Failed Car Analogies. Amateur Lawyering. Anecdote Battles.
It's saying a whole lot that it beat him? I would hope that a machine calculating trillions of moves would be able to. Like a lot of articles I've read, the machine can often pick excellent moves at any given time, but it lacks an understanding of the overall flow of the game, and big-picture strategy. Those kinds of things are hard to figure out for a machine without a soul, even with near-infinite cycles to spend. Until the machine can prove the game and calculate a way to draw every time no matter what moves the other person makes, I think a human will always have a good chance at beating the machine.
My sig is blank, I typed this by hand.
As long as the best chess playing computers are still made by humans, I'll feel confident in the superiority of our species.
It's when the best human-made chess playing computers are routinely beaten by the best computer-made chess playing computers that I'll be worried.
The easiest way to describe why Kasparov loses to a computer is because he is human. How often does he play his best chess? Not often - he's human.
The computer, on the other hand, always plays its best chess. So we are often comparing the computer's best vs. Kasparov's weak or mid-level chess, i.e. *mistakes*.
I don't think that Kasparov playing his best and making no mistakes would have any trouble with current computers. But *with* mistakes and fatigue and such...sure.
So the question really becomes, is it as fun to have the computer win when Kasparov makes a mistake? I don't think so. I think the real fun comes when he plays the best he can, is sure he can win, and has the computer do some wicked shit that no one has ever seen. When they staring thinking like humans - only better.
That doesn't seem to have happened yet. They simply have gotten good enough to be able to pounce on GMs that make mistakes, but not on good GMs that don't.
Hell, that's just my observation - I'm no chess or chess AI guru.
dmiessler.com -- grep understanding knowledge
Theoretically, if both sides play perfectly then every match will end in a draw. So what if Kasparov plays perfectly? Obviously he's lost before, so he doesn't all the time, but it's certainly possible that he could, at least for one game (people play less-complicated games, like Tic-Tac-Toe, perfectly all the time). If so, then no matter how good the computer played it could only draw him. So really, I think chess isn't really an accurate indicator of when 'the day of the machines' is here (or not).
It could only look 12-20 moves into the future, and Kasparov played in a manner to limit what the computer could see by looking ahead. Since the computer had no strategy, but rather always took the best move he could see at the moment, Kasparov could keep him blind and cornered so it didn't see anything usefull to do in the short term, so it ended up flailing about somewhat (notice where it moved a peice and then moved it right back). Then all the meanwhile he was slowly playing out a much longer strategy.
It was a pretty big deal when Babbage built a machine that could do basic arithmetic. I'm sure people thought of his Difference Engine as being a "smart" machine, particularly since it could generate tables of numbers a good deal faster than a human. But if you looked at the machine, it was all cogs and shafts and springs and levers... I'm sure that once you got over the astonishment that a machine could do this seemingly difficult thing, you'd look at it and know that it really was still just a machine, and not truly a thinking thing.
We consider ourselves to do this mysterious thing we call "thinking," but we don't understand in a precise way what this means. It could be that our brains work in an algorithmic fashion, or at least that our brains can be simulated by a machine that works in an algorithmic fashion. The former seems unlikely to me, the latter very likely. Is there a difference between actual thinking and simulated thinking? It's hard to say.
When you look at these chess-playing computers, they're pretty amazing. They can certainly play one hell of a game of chess. But when you get right down to it, they're really solid-state versions of cogs and shafts and springs and levers. Are they thinking? (I want to say 'no', but I can't prove it.) Are they simulating thinking? (Maybe... it's hard to say since we don't know what thinking entails.) Is there a difference?
Computers can look at many more positions per second than a human but that is not as helpful as you would think because most of the positions examined are bad ones. While humans compare poorly in linear computational speed to machines, we are much better at pattern matching. A grandmaster has studied so much chess that he or she can effortlessly recognize whether a position is a poor one. A GM probably does not even consider moves that would lead that those poor positions any more than you'd consider racing a dune buggy across a mind field.
Yeah, thank goodness for my soul. I'd hate to see how badly I sucked at chess if somebody extracted my soul.
I hate this sort of thinking. If the question is, "What is it that allows humans to think abstractly and formulate efficient and creative strategies in the face of novel situations?" answering, "a soul" is just sleight of hand to avoid admitting that we don't know. Positing that every human being has a soul explains nothing, and tells us nothing that we didn't already know. Slapping a label on a phenomenon isn't the same as providing an explanation.
Now, regarding "near-infinite cycles," ask your math teacher about the logic inherent in the phrase.
You want the truthiness? You can't handle the truthiness!
I'm not a game programmer, but I imagine that some day such programmers will be able to add heuristics to chess engines so they can recognize such situations and handle them with some other algorithm.
What, if any, is the difference between "simulated intelligence" and "actual intelligence"? How do you know that our brains don't function as massivlely parallel search/inference engines?
Discussions about AI usually degenerate pretty quickly into arguments about whether or not we have some invisible, intangible, God-given "soul" or "spirit" or "spark." You're use of the phrase "*miracle of real intelligence*" would lead me to guess that you'd probably come down on the side that favors such a thing.
We humans are self-aware, yet we have not yet explained the mechanism of our self-awareness. Many of us assume that it therefore cannot be explained, that it is miraculous. I think that's a poor assumption. It may be, however, that we are incapable of understanding our own self-awareness, and incapable of understanding our own intelligence. Whether that's true or not, it does not follow that other animals and even machines cannot develop intelligence.
Why is it necessary to build a machine that plays chess "*as a human does*"? It's unlikely that any two humans play chess the same way, so which human would you have the machine emulate? Wouldn't it be better to build a machine that plays chess its own way?
Yeah, while humans are clearly helped by reading opening books, humans with a mathematical mind can invent them on the spot, to do at least a decent job (though an idea of "decent" varies a lot). It would definitely be a good show of AI for a computer to be able to do that, and they'd probably be closer to figuring out how to get computers to compete with humans in go.
One thing that has been pointed out by numerous posters is the belief that the final result of the match is the result of one bad move in one of the earlier games.
This is not necessarily meaningful. Either player could have played better or worse in any of the positions that came about in the ensuing games, making the actual match results the stuff of speculation about alternate universes.
Be that as it may, two things stand out about the match. The first is that the computer opponent is actually a commercial program running on commercially available hardware and not on specialized circuitry out of a lab somewhere. This alone is a very good indicator of how far computers have come as chess players. Not too long ago (at least in geological terms), there wasn't a chess program on earth that could win against the like of me. Nowadays, by contrast, commercial desktop hardware combined with shrinkwrapped software are giving a former world-champion a run for his money.
The second point of interest in the final game is Kasparov's choice of defenses.
Kapararov is one of the world's greatest experts on the opening--someone who prepares not just against continuations but against his most likely opponents--and yet, with the game and the match on the line he, chose to not play any of the 'milder' defenses to 1.d4 (for example, trying to reprise the line of the Gruenfeld he played against Karpov in one of their matches) but instead chose to play the black side of the Queen's gambit accepted.
When I was growing up and studying, the queen's gambit accepted was known to offer white good chances to develop a strong initiative based on black's disadvantages in central space and white's rapid development, and venturing the black side of Queen's gambit accepted was considered risky.
Apparently, Kasparov knows something I didn't when I was fifteen (duh).
Still, Kasparov's choice of opening certainly led to a difficult position requiring an accurate defense after white developed significant pressure on black's central pawn and on the queenside. However, the pressure rapidly dissipated following a set of exchanges that even gave black a short-lived counterattack on white's king position, leading to a position with even material and no real sources of play for either side, hence the draw.
It would have been interesting to see what would have happened in a longer match played under a different winning criterion like 'best-of-ten' or 'first to achieve a given score.'
To mail me, remove the 'mailno' from my email addy.
"Yeah. It smells, too..."
Attempts to make Turing type B (rulebased, heuristic) chess programs have failed so far, all strong playing programs today are of type A (brute force, with perhaps a little heuristics). You can't just make up simple rules for playing chess, those rules will not account for all possible positions on the board; what's good in one position can be instantly losing in another one.
For instance, even the little heuristics in a type A program such as Fritz can be uttlery wrong. Fritz lost the third game to Kasparov because there was a heuristic programmed into it not to push pawns that protect the king, while Kasparov had chosen an opening that resulted in a position where Fritz' only hope in the long term was to do exactly that. Fritz had not enough processor power to look that far into the game, and Kasparov knew it.
After that, what is the point?
The point is to create machines that play chess better than humans do now, so humans can learn from them and get new insights about the game. We know that it's impossible to play perfect chess (because there are more positions possible than there are atoms in the universe), but we also know that what humans (and computers) play today is still very far from perfect.
It won't be the first time, and it won't be a *thinking* machine - just another specialized machine.
Yeah, the classical AI bait 'n switch. First it's really intelligent to play a nice game of chess or do complicated mathematical equations, but as soon as computers can do it it's no longer thinking. The solution would be to sit down and define what thinking exactly is, except nobody has ever succeeded in doing just that.
It does not take a great programmer to write a program that cannot lose at tic-tac-toe against a human.
Yeah, but at the same time no one has written a chess program that cannot lose against a human. Humans still play better chess than computers do, even though the difference between them is getting smaller.