What you're reading is exactly what Maurice and Yasser said at the match - thanks to the fine work of the stenographers at the scene. Please forgive them a typo or two and enjoy the commentary!!
Commentary by Maurice Ashley and Yasser Seirawan
ACM CHESS CHALLENGE
GARRY KASPAROV vs. IBM'S DEEP BLUE.
Day 2 - February 11, 1996
White Garry Kasparov
Black IBM's Deep Blue
Mr. Ashley Welcome to the ACM chess challenge. I'm Maurice Ashley. My partner is Yasser Seirwan. Garry Kasparov is playing against IBM's Deep Blue, and as most everybody here knows, he is down 1 tonothing already. So Kasparov needing to play well in order to come back. A big question for him is whether or not he can handle the psychological pressure of being down against the computer that, first of all, everybody thought he was going to beat, including himself and, second of all, he simply has no idea how strong it is because this version that they're using has never been tested and is clearly playing some excellent chess.
Yasser, yesterday's game was a model of computer cold-bloodedness.
Mr. Seirawan Precision.
Mr. Ashley It just did not care about Kasparov's attack and just ripped him off the board. It was unbelievable.
Mr. Seirawan It's terrible. I'm still recovering.
(Laughter.)
Mr. Seirawan Actually, prior to the match I had said, okay, it's great. This is wonderful. There's a lot of hype, the best computer the world versus the best human player in the world. Well, it's no contest. Garry is going to just win. And I would be shocked, shocked if the computer won any game. So naturally --
Mr. Ashley So you're in shock.
Mr. Seirawan I'm in shock.
So naturally Deep Blue won the first, and just as you were saying, Maurice, I can't fault any single move that the computer made.
We had dinner last night together with a group of ourselves, and we just kept going through the game at various stages, and we said, this is a very, very serious opponent for Garry. This is a very legitimate match, and of course now that Garry is down a point, he's got to prove himself. Yesterday I had spoken about the fact that in tennis -- and again I'm probably misattributing the quote. It was of Rod Laver, when he was going to sum up his opposition, he said, I only need to see 3 shots. I need to see the forehand, backhand and the serve, and then I will tell you how long or how many sets the match is going to last before I win.
And Garry said the same thing on Friday at the press conference. He said basically I need to see the computer on offense, on defense, and then the match is going to be mine. So he basically saw the first 2 games as just being his ability to sum up his opponent and then vanquish him in the latter half of the match.
Well, that may still work, but he's got his work cut out for him. He's made it more difficult for himself.
Mr. Ashley What about the psychological pressure on him? From what I've seen, I've seen Kasparov down in matches before. He was down against Anand in game 9. He was down against Kramnik in the Paris leg of the Gran Prix tournament and came back and won. In each case Kasparov seems to bounce back from matches. He is not just the kind of guy who goes down in chess games and falls apart.
Mr. Seirawan That's absolutely true. And not just the instances that you've just mentioned, but rather Garry throughout his entire career, this is his strength. In fact, it's the strength of all world champions, is their ability to recover from a loss. And like no other player in my history of playing chess, I became a professional in 1977, have I seen it in anybody to bounce back just the way that Garry does after a loss. He gets mean. And it provokes him and he usually turns it on and he manages to succeed. So I would say that right now Garry today would rather not be in a position that he finds himself; that is, he finds himself behind. But I think it's going do something to provoke him, Maurice. I think he's going to really get angry.
Mr. Ashley Angry?
Mr. Seirawan Angry.
Mr. Ashley Have you seen Garry Kasparov angry before?
Mr. Seirawan Definitely.
Mr. Ashley How well does he do when angry?
Mr. Seirawan He gets on what you might call his game face.
I'm seeing the monitor. Yes, there you have it there.
Mr. Ashley It seems that he has a game face on right now.
Mr. Seirawan He looks very serious.
While he was writing down his name, I think the table started shaking. I have played Garry 5 games in my career and those games have spanned from Yugolslavia, all over the place, and I've really seen Garry get himself up for the game. And I must say on Friday he looked really, really good. He was very, very sure of his victory in this match. In fact, we were having a joke about that. I know we're getting ready to get started, but before the match there was going to be half a million dollars in prizes that the ACM, sponsor of the competition, was providing. And so the ACM had said to Garry, okay five-eighths to the winner, three-eighths to the loser. Garry said no no no no no, all or nothing. I want the half million. I want the half million. So a compromise was reached and it became 400,000 to the winner, 100,000 to the loser.
And now Garry might be --
Mr. Ashley Regreting that decision. Well, they have indeed begun the game with knight of 3 and D5. You'll be able to see the moves on the monitor. Before we get to that, I would like to read something that was in the New York Times, and we'll keep that in mind for today. In the New York Times on Friday Kasparov said that the psychological aspect of the game will not be as important in this match and he says he'll give a simple example. When we humans play a game and try to organize an attack against the king and we bring 1 or 2 pieces near the king, these moves carry psychological pressure which doesn't work against the computer. Automatically we'll look at the opportunities presented by your pieces around the king, and if they are threats it will counter them. If there are no threats, it will play the other side of the board without another thought.
Mr. Seirawan Well...
Mr. Ashley Well, we saw that happen yesterday n this position. As you see here it looked as if Kasparov was organizing a famous Kasparov attack and instead of reacting to what looked like some mating ideas near its king, the computer cold-bloodedly played the move knight takes B7, which is a move we thought was insane. I mean why would you want to ignore the attack on the other side. And the computer responded with this move and, as you know, there was no attack because it already calculated the position down to a win for itself. So this is the kind of thing Kasparov is going to have to deal with and he's going to have to overcome this.
And I think this has to do a lot to his self-confidence because normally you're playing -- Kasparov playing against a human. Oh, my goodness. Don't let Garry start attacking you. You're going to die. And the computer ignores him and puts his knight over on B7.
Mr. Seirawan Exactly. When Deep Blue took the one on White's 29 move, if you look at the position -- and sometimes you just have to do that. You don't necessarily have to calculate, but you just count the number of attacking forces and you say in this position after White's 29 move, Black's rook is attacking the G2 pawn, the. Queen on F6 is hovering uncomfortably close to the White king. That pawn on F4 is an extra attacking unit and, in fact, the move knight takes B7 provoked the next move of Garry's, which was knight E5 I believe. Yes. And suddenly this whole attacking unit seems to be having a harmonic convergence, if you will right, over on White's king.
And as a human being, I'd be sweating because And as you said the computer very cold-bloodedly had calculated that the attack wouldn't work and in fact rebuffed the entire attack and won the game KASPAROV MOVES 1. NF3 D5..
Mr. Ashley This is what Kasparov has her to face, but let's get to the game in question. He began with a move. I think I won a bet here because I thought knight F3 would be move 1.
Mr. Seirawan I'm going to get another crystal ball. Not knight F6. D5.
Mr. Ashley I believe Fritz wants to get in on the act. It's making moves here.KASPAROV MOVES 1. D4 ..
And now the move D4. Well, we should mention that we are going over the internet right now, the worldwide web, and if you are unable to make it to the game, you can log in on the worldwide web and follow all the action and even here hear our commentary as we're speaking.
Mr. Seirawan Well, it seems the computer has paused in this position.
Mr. Seirawan On the second move.
Mr. Ashley On the second move, which seems a bit of a surprise. I mean if Kasparov had played the moved D4 D5 knight F3, I imagine it's completely ready for that, but somehow he's switched the move order around, and it stopped to think. I mean this position has to be in its opening book I would imagine.
Mr. Seirawan Most certainly, most certainly. It's interesting, I mean that's actually a very human feature. David Bronstein was famous for sitting down on his very first move spending ten, 20, 30 and sometimes even as long as 6 every 7, 60 minutes for his first move and we would say to him David what the heck you're coming to the game. You know you got White. You know your first move. Well, I was getting the vibration. I think the computer is getting the vibration of the game. It's trying to pick up what Kasparov might have in store for him.
I would also say why is he thinking on his second move that because chess, as we know, has such a huge variety of openings, it's not -- sometimes it's very awkward even for the human being to say well today I'll feel like playing this kind of a game. And Joel Benjamin, Joel Benjamin, who is the trainer, grandmaster Joel Benjamin who is the trainer for the DEEP BLUE team in terms of its opening is really known for his unusual openings and maybe he'd plan something.
DEEP BLUE MOVES 1. ... E6..
Mr. Ashley Garry is making many faces in thought.
Mr. Seirawan Lon Chaney, the man of a thousand faces.
Mr. Ashley No question. Kasparov is an extremely expressive chess player regardless of who he's playing. Many people believe he does it in order to intimidate the opponent or throw the opponent off, while DEEP BLUE will have a little problem being thrown off by Kasparov's expressions.
Well, now he stopped to think I just should recap the moves from the beginning of the game. Kasparov started out with 1 knight F3. DEEP BLUE responded with D5 and after Kasparov's D4 it went into a small think and then it produced the move E6 and Kasparov is on move. His position, what can you say about this position. Yesterday they whipped out the first 12 moves in 2 minutes. Now, you know, it's like they're trying to reinvent the wheel. What's the deal?
Mr. Seirawan Well, Black's second move was an invitation to the queen's gambit declined or even perhaps a Dutch defense with a move F7 F5 on move 3. But the move knight F3, by the way, instead of even E4 or D4 move on move 1, the move of Garry's opening move knight F3 is meant to be an extremely -- let's call it move. It's a cunning move. The idea is to induce the computer or the opponent, rather, to go into a variation where Garry hasn't yet committed himself.
And so these first 2 moves of DEEP BLUE's has basically cut out the effectiveness or the cunning, if you will, of knight F3 and so Garry is going to be perhaps facing the queen's gambit declined and I don't think he -- well he certainly doesn't look like he wanted to play that type of an opening. I think he wanted to play more of a ready or an English style opening that we've seen in this world championship matches.
Mr. Ashley Which I think would be better for him because those kind of situations are much more closed, much more maneuvering and the specialists Joel Benjamin for example you mentioned before and others who play against a computer know that the main weakness of the computer, if there's a real weakness, glaring weakness, is the fact that they don't play so well in closed positions, position that is demand a lot of maneuvering because they like the tactical ideas, and when they don't see them, they start just footsying around.
Mr. Seirawan Well, in fact that's become more and more clear to human opponents of computers. That is to say, it's very clear that the strengths of the computer is its extraordinary calculating speed. This morning I was looking and I was thinking to myself, we have DEEP BLUE. This is an engineering marvel. It's really hard to get your mind around some of these numbers. It's thinking at 200 million moves a second. Okay? 200 million moves a second. So I sat down with my pen and I began thinking to myself well how many seconds in an hour, in a day and so forth and so on. And I determined that it would take me somewhere in the neighborhood of 14 years to make a decision every single day to equal what DEEP BLUE is doing in 1 second. And yesterday's game was something like a 3 hour 50 minute game or a 3 hours and a half. I mean it's extraordinary. All of our lifetimes we won't make all the decisions that are made in a single ten minute period what DEEP BLUE is doing. It's extraordinary.
Mr. Ashley And yet Kasparov is still competitive in many -- the layman thinks when you ask the layman the question what do you think between Kasparov, the best player, and the best computer what do you think the chances are, and the layperson goes you mean the computer is not killing him already? And it's an intuitive response. You say yeah the computer should be destroying him. And yet Kasparov is able to, has been able, we'll see how it goes in this match, but he's been able to dominate computers and the top players have been able to dominate computers.
Mr. Seirawan Indeed. And before the match I thought I predicted a match score by the way of 4 to 2 in Kasparov's favor and again I didn't think he would lose a single game. And again last night of course at the dinner we were at a dinner of some heavy, heavy hitters and in the computer science world, some really great minds and they were all explaining to me how wonderful the computer was and it could do this and it could do that. And I said that's great, that's great, that's great. And they said, you know, nothing that Kasparov does the computer won't have seen. And I said that's absolutely true. I have never beaten a computer that hasn't anticipated my every move. But I'm always beating the computer. So it's not that just that the computer can anticipate the move. It's what does it mean. What's the valor, the strength? The computer doesn't yet still understand the underlying . Strength or value of this move. And by the way we do have a move on the board.KASPAROV MOVES 1. G3..
Mr. Ashley Yes. Kasparov has played G3 after much thought. And again DEEP BLUE in thought -- you know it's interesting maybe Kasparov is being foxy as you call it, trying to play some strange move orders that force the computer to think on its own as opposed to following any kind of preparation. How do you feel about that??
Mr. Seirawan Indeed, I think that in this particular case -- when I play a computer, I very often want to get the computer out of so-called book or opening theory. But in Garry's case, Garry is easily, easily the most formidable openings expert amongst humans in the world. I mean he has been coming up with novelties for his entire career indeed. Yesterday we saw that he was the first to introduce a new idea into the opening. And what he's done now with the move 3 G3 is of course he's anticipatesing fianchettoing on the king's side. That is to say he wants to put his bishop on the square G2 followed by castles. And in our way of thinking of strategy we call this building a house. The idea is to play G3 bishop B2 castles and the king is very well fortified on the king's side. So he's building a house. Also the move 3 G3 is invitation to a catalan. And this is what we were speaking about yesterday. What Garry really needs to do is get 1 of those advantages which is really kind of nebulous. I mean he just -- it's a long lasting advantage which he can do a lot of things with. He can go left, right, center, and so forth.
And G3 is actually the catalan opening is noted for giving White this kind of game, where he's slightly better for a long time and he -- the idea being to grind the computer down.DEEP BLUE MOVES 1. ... C5..KASPAROV MOVES 1. Bg2..
Mr. Ashley What is's interesting is that already you've come to this reorientation about the position. You're thinking of the direction it's going to go in and what type of structures will occur, the kind of exchanges that may be possible, even the type of game. You know of many of the middle game positions that occur, the end game positions that occur. The computer has played a very aggressive move. As I speak the move C5 in response to G3 and Kasparov playing bishop G2 instantly. But to finish my point, maybe you can elaborate on this.
Mr. Seirawan Sure.
Mr. Ashley You already even know some of the end games that are possible in this from your vast experience of chess.
Mr. Seirawan Right.
Mr. Ashley The computer doesn't. And will have to work out these subtlies, and think is what really gives the human that type of advantage.
Mr. Seirawan Indeed. What you see, of course, is if you imagine a graph, you have on the left side of the graph what we call the opening phase of the game, the opening phase. Then you have the middle game phase. And then you have your end game phase. And where computers are extraordinarily good is they have a perfect memory. They have a perfect memory. That's sometimes kind of hard for us to understand because you always have that -- something on the -- something on the edge of your tongue. You've forgot something. Well the computer never does. It's perfect. So the computer's openings, it knows from its memory perfctly. It can duplicate hundreds and hundreds and hundreds of thousands of games perfectly. And on its end game side it is its theory, it will play perfectly certain types of endings. However, this middle game, if you will, is a gap. It's a window where anything can happen from the computer's perspective.
But from our perspective as a human I say well if I play these particular openings I already jump into the end game and say I'll get those favorable end games that I love where I get a mass of palms on pawns on the queen's side or I'll get something going in the center. And this is what Garry has always done throughout his career is marry his openings to his end games, which makes him so formidable because his games look like just a smooth progression from 1 advantage to another.
Mr. Ashley And so he need not even calculate the way the computer will have to calculate every single move because he already knows what type of move is the best move looks like.
Mr. Seirawan Right.
Mr. Ashley And he knows where he's going. He already knows where he's going and the kind of situation he wants. It's going to be favorable -- he doesn't have to calculate too much. He just knows x y z and I'm there.
Mr. Seirawan Black move, C5, a very salient type of move. What you do of course in the opening is it's the center. You've got to control the center. A cardinal rule in the prince of opening play is play in the center. And the move C5 -- let me go over here -- basically what Black is doing is attacking White's only center pawn, the D4 pawn. And now what he's done, if you will, Maurice --DEEP BLUE MOVES 1. ... Nc6..
Mr. Ashley The computer has finally responded with the move knight to C6 and it's very interesting because yesterday DEEP BLUE spent no time in the opening and now just from this 1 little transposition it has already spent 11 minutes to play just these first 4 moves. Kasparov has used openly 6 minutes and now he has was he had.KASPAROV MOVES 1. O-O..
Mr. Seirawan So he's built and completed his house. And I was just going to say that what Black's strategy has been so far is to attack this pawn on D4 because indeed if he can wipe out the pawn on D4 then White won't have a toehold in the center. For those of you who are familiar with the game of chess, Black's setup is like a grunfeld defense. Do we all know what a Grunfeld defense is? Well, not so many hands. May I recommend Garry Kasparov's book on the Grunfeld for you? Garry is a world renowned experts with the Grunfeld defense. But that's when he's playing Black. So what he's done is played 1 of his favorite opening setups with an extra tempo. That is to say he's played the Grunfeld defense as White and this is 1 of his petition petition lines and has been for a couple of good decades old.
Mr. Ashley Again interesting, the fact that as a human you look at the position and say well this is like the setup as if I had the Black pieces, but with an extra move which is something that that's kind of pattern recognition clearly, is pattern recognition is which is what we humans compel at, and the computer doesn't care about. It's thinking right now about I'm happy about your Grunfeld defense.
Mr. Ashley So the position clearly one that I would expect quick moves, but already Kasparov's move order, just a switching of the move order has forced DEEP BLUE to think on its own. It's still coming up with very fine moves, standard moves, controlling the center, aggressive with pawns, bringing the knight toward the center of the board. But it is stopping to think somewhat.
Well, we have some illustrious company with us and we will point them out as the day goes along.
I see in the back computer specialist Hans Berliner, a real giant in the field. We managed to sit with him yesterday.
(Applause.)
And I see him wishing to make a comment.
(Audience commentary.)
Mr. Ashley Hans Berliner has made the point that the transposition should not affect the computer, and to him it must be some sort of bug that has thrown out opening theory for Black for the computer, which is pretty interesting to hear. The computer is going to have to figure everything out from the beginning.
(Audience commentary.) 5. ... Nf6. 6. c4. .
Mr. Ashley A very interesting comment, that everything, everything in the computer's memory is based on positions, not on how you get there. So that means that the idea of transpositional causing -- the transposition causing some kind of confusion in the computer should not really be the case, and maybe there's just a bug in the computer program that they didn't discover. Well, that's very interesting as an idea.
Before we discuss that any further, Black did play the move knight to F6 and Kasparov responded quickly with the move C4.
Mr. Seirawan Yes.
Mr. Ashley And Kasparov very intense looking right now with his head down, his hands together collapsed as if in prayer. We've seen that position before. But some dynamic stuff happening in the middle of this chess board. This is the kind of situation you would think Kasparov wanted to avoid, not having the center pawns with many exchanges possible and explosive tactical ideas at any moment and Kasparov just jumped right into the complications as if to say I am completely prepared here. I have no fear of what's going to happen.
Mr. Seirawan Indeed. This is 1 of Kasparov's favorite positions, is White. In 1983 I played with -- in 1983 I played with Garry in a tournament in Yugoslavia, in and what we have right now on the board today is 1 of the mainline variations of the catalan opening. If you will, if you follow me for a moment here, the normal variation for Black is to play the move d takes C4. And then comes the move queen A4. And then c captures D4, knight captures D4, queen captures D4. I know this is fast and I won't be explaining all the poses in the position. Bishop takes C6 check. Bishop to D7. Rook to D1. Queen captures D1 check. Queen captures D1. Bishop takes C6. Now, this is a very standard position for the catalan opening.
Garry has got a very nice look on his face.
So the position that we have in the middle of the board is a position that Garry has played many, many times and in Yugoslavia in 1983 he was playing Ove Andersen and gave a brilliant exposition of why the position is favorable for himself. So I'm sure he has some very good memories and of course the computer didn't go into that.6. ... dxc4.
Mr. Ashley Very interesting indeed. It has instead captured on D4 the move Black has, in fact, played is c captures D4, CD 4 and Kasparov kind of jumped up in his chair for a moment looking at that last move because it's not the standard move as you say.
Mr. Seirawan And indeed I was just about to say if Hans is right about the access or failure to access its opening library, in my view or my experience, I should rather say, the move c takes D4 this early is actually wrong.
(Laughter.)
Mr. Seirawan What happened?
Mr. Ashley The operator got up from the board, which he is allowed to do when it is Black's turn to move, and only allowed to do when it is Black's turn to go, but he is in fact calling the arbiter over and something happening at the board, Kasparov showing some kind of concern.
(Audience commentary.)
Mr. Ashley He may have made the wrong move on the chess board which Kasparov has seen before.
Mr. Seirawan The operator may have put in the wrong move because I wanted to say that the move C5 takes D4 in my experience is a bad move.OPERATOR MAKES WRONG MOVE.
Mr. Ashley No wonder Kasparov was surprised. And now because of human error, because of human error we have had to go back, we have had to go back 1 move and instead of the move c takes D4 for Black on move 6, instead of c takes D4 for Black on move 6 and on move 7 White playing 7 knight takes D4, the actual move the computer told the human being to do was d captures C4. So it was not in this case we human's feel a little embarrassed because the computer had played the correct move d captures C4. The standard move in this position. And the operator, the computer operator, had mistaken the letters and switched around and put the wrong move in. That could have a slight psychological effect on Garry. He must have been so happy to see c takes D4 thinking yeah I've got this computer now and then to hear that it was the wrong move and it's playing the right move has to have thrown him off just a bit. And maybe there was some ploy there. Okay, we won't put any negative --
Mr. Seirawan That was truly.
(Audience commentary.)
Mr. Seirawan Let me just repeat. The gentleman asked the question if the computer has lost its opening book and Mr. Berliner will address the answer, does that mean also the computer has lost perhaps its end game book? And Hans will say?
(Audience commentary.)
Mr. Ashley It seems to me that we are jumping the gun a bit saying that the computer has lost its opening library. That may be the case, certainly, and it may not be the case. But right now the computer is playing book, is finding all the standard moves, so something may just have happened momentarily.
Another comment made by Hans Berliner is that Joel Benjamin, who is excellent at playing against computers, he has won the Harvard Cup some 3 years running, the Harvard Cup being the competition between the top grandmasters, 6 top grandmasters and 6 top computer playing programs and Benjamin I believe has lost only 1 game out of 18 against computers. He's won with scores like 6 nothing 2 years in a row and they won and he finally lost 5-1. The DEEP BLUE team wanted the best to talk to about how to deal with the best, how to deal with world champion Kasparov. So they brought Joel Benjamin on to their team and asked him about openings, what kind of openings they think he should play, the computer should play against Kasparov, and so forth, all kinds of subtle questions like this. And Hans Berliner is making the point that maybe Benjamin felt let's get out of theory and let DEEP BLUE play on its own and see how strong it would be by itself. If that is indeed the case, it's doing very well.
Mr. Seirawan Yes.
Mr. Ashley I mean it's just finding the moves.
Mr. Seirawan I suspect though that we are back in its opening library because again these moves that we're seeing are very, very standard opening moves.KASPAROV MOVES 7. Ne5..DEEP BLUE PLAYS 7. ... Bd7..
Mr. Ashley And Kasparov played the move. Just to amplify your point, Kasparov has now responded with 7 knight to E5 and it instantly, with absolutely no thought to the matter, played the move bishop to D7 as if it was prepared for whatever he had in mind.
Mr. Seirawan Exactly.
Also, I just wanted to say that when the human operator had put in that last move of DEEP BLUE's, did you see how Garry just jumped? I mean he was so happy he made his response very, very quickly. And boy I tell you that would ruffle my feathers if I sat there and thought that my opponent had blundered and I pounded on my opponent's move going oh, yes, and the operator, I'm sorry, Yasser, I'm sorry. But we're going to have to make a better move for the opponent.
Mr. Ashley It's like the kid got the candy and then you said no, no, no, no. This is for next year.
Well, Kasparov seems to have settled into the game again and he doesn't seem disturbed anymore by it, and I think he's decided to come to this game much tougher. Yesterday's game he did not play as tough, he did not play a tough minded chess and stick with his strategy of making sure the position was not 1 that was comfortable for the computer where the computer's strengths would really reveal themselves and I think what you're seeing here is Garry Kasparov is extremely determined. He has his back to the wall. He cannot afford to lose another game that. Would practically put him out of the match already. And I think you're seeing almost somewhat pensive Kasparov in that he feels the burden on his shoulders. He feels that he's defending the honor and dignity of the human race and he said that. And he has said that.
And another thing that's interesting is yesterday Kasparov got up a lot during the game. You could see a nervous Kasparov. He got up on every single move. If he knew what your next move was or he felt a position was okay he made the move and got up. You're not seeing it now from Kasparov. He is at the board and I think he's determined not to lose concentration at all.
Mr. Seirawan Right. Indeed. He's dropped that a great deal. Back when Garry was a junior player and obviously we were competitors, I met Garry for the first time in 1980 in Malta, and I didn't -- I recognized the name, but I hadn't yet associated the player with the name. And I was watching the Soviet team and I was watching this guy literally, you know, going around the table like stomping around the table and with, you know, clenched fists and he was really moving around and I turned to Pal Benko, he, who is a grandmaster from Hungary and our team's trainer and I said pal, who's the caged tiger? And it was Garry Kasparov. So Garry at the board has always shown his energy, his explosiveness.
And I just will get back into the game for just a moment and say that this line of the catalan was developed by famous world champion Michael Tal. Has the audience heard of Michael Tal? Just a beautiful --
Mr. Ashley We hope you have, and if you haven't, definitely get a book on his games because you'll be in for a treat.KASPAROV MOVES 8. Na3 cxd4..9. Naxc4 Bc5..
Mr. Seirawan Glorious.
Mr. Ashley 1 of my favorite players of all times.
Mr. Seirawan And Michael had developed this theory and of course Garry Kasparov being from Russia and Azerbajian certainly followed the games of Michael Tal. And the intent of this opening is to sacrifice a pawn. And we see that DEEP BLUE is up a pawn and, normally speaking, we hate to give computers material. So Garry is pretty sure about what he's done, is this pawn on D4 of course is an extra pawn in the position. So DEEP BLUE is a pawn up. But what White has in return is to excellently posted nights, his knight on E5 and his knight on C4, really control a great number of squares and his bishop on G2 roars down this diagonal, putting a lot of pressure on Black's position. And finally also note that Black's king is still stuck in the center. So what White gains is time active pieces in return for the material of the pawn. The main lines in this variation continue with the move bishop C5 protecting the pawn on D4. White usually responds with moves like bishop F4 or bishop G5 but bishop G4, castle for Black and then a move like rook C1
Mr. Seirawan And it's a long material pawn sacrifice. White isn't a favorite to win the pawn back instantly, but when he does, and he usually does, he'll have a long-term advantage. So this was actually, I think, Maurice, a good idea, a good strategy being to sacrifice a pawn early in the game, knowing he can win it back and then in the long term have an edge that again we were talking about last knight grinding the computer down and playing a flawless technically game
Mr. Ashley A certainly well-analyzed position by Kasparov and now, as soon as I said it, Kasparov has decided to leave the chess table. A human operator getting him some water.
Mr. Seirawan Juicing up.
Mr. Ashley Certainly an interesting position and clearly 1 that Kasparov knows extremely well.
Mr. Seirawan Yes.
Mr. Ashley And not something he's going to be afraid of. And we'll just wait to see if the computer catches up basically with Kasparov's deep analysis with positions like these. He didn't even have to see the position before the game to refresh his memory because he knows these types so well because it's such a part of chess lore.
Mr. Seirawan Yes.
Also, I wanted to say that at 4 I have arranged for C.J. Tan, who is the leader of IBM's programming team, to come and speak to yourselves a bit of what I must say is a fascinating engineering marvel of having all of these specialized chips, all working together in what is known as parallel processing to create these calculations of just mind-boggling numbers. And you'll find him a delightful man. We won't ask him a lot about chess, but of course he's like a father, very proud of his creation and I think you'll find it fascinating insights, and he'll be with us at 4. I just say that as something you should all get yourselves to look forward to.
Mr. Ashley Most certainly.
Well, now we will check the times, and it seems Kasparov has a minute 46, is it?
Mr. Seirawan That's a tough 1. I can't tell.
Mr. Ashley There's a bit of a glare on the clock, making it difficult to see. And no chance at all on DEEP BLUE's. The glare is preventing us from seeing exactly what time is left.
Kasparov has 145 and somehow the display on the digital clock is very faint.
Mr. Seirawan I just wanted to say also we will be polling our audience and we do of course invite your questions. We have a question from the gentleman in the front row.
(Audience commentary.)
Mr. Seirawan Indeed. The question the gentleman asked, are we still in opening theory. Are we still in the theory of the catalan? Very much so. This is still opening theory. Virtually all of the possibilities have been mapped in hundreds of books. And that's one of the things that I'll just talk about very briefly, is the fantastic history of chess, the history of chess. We know that chess goes back to Queen Nefriteri. She played what is kind of an Egyptian chess and Aristotle is quoted as saying that when a stranger in a strange land plays a game of chess, you're sure to make a friend. These forms of chess eventually evolved into the sixth century a d game from India called chaturanga, which is a direct descendant of what we're playing today, western chess.
So western chess, if you will, or chess is 1600 years old, 1600 years old. It survived all that time. And not only has it survived, but throughout the generations the people who play chess love to share their brilliance with the future generations. So I have a very close friend. His name is Dr. John Nun from England. And last year he wrote a wonderful book about certain end games. And I opened up the book, printed in 1995. I open it up and the first chapter quotes an eighth century Arabic manuscript about a rook and knight ending and how it's won.
So the ashes from Arabs from the eighth century were analyzing positions that are still important today, classical, classical knowledge. So more chess books have been written about chess than I think any other sport in the world, literally hundreds and hundreds of thousands of chess books throughout the generations and countless languages have been written on chess. So when you say we are still in opening theory, for a long time to come, for a long time to come.KASPAROV MOVES 10. Qb3 O-O..
Mr. Ashley And we have seen it play the standard move bishop to C5 as you mention, but now Kasparov has responded with the move queen to B3. He did so instantaneously. So clearly still in Kasparov book at least.
Mr. Seirawan Exactly.
Mr. Ashley And when we say book, book is not just something written from on high and you must follow and say well everybody plays this, this, this, this, that. A book is a compilation of practical experience, practical experience of the grandmasters and analysis of the grandmasters, but not all grandmasters know the same book.
Mr. Seirawan That's right.
Mr. Ashley Kasparov has stuff in his library. Well, now in his computer library that we will not see for a few years to come and may never see the light of day, the kind of marvelous analysis that he has done over the years for all his world championship matches, the analysis that he's done and some grandmasters, I'm sure it's happened to you, keep a novelty that they're just waiting to play in their pocket for years and 5 years later some unsuspecting soul falls into the trap. So Kasparov definitely with his own book, we'd love to see it, but believe me you won't unless you play him a chess game and you're ready to lose.
Mr. Seirawan Well put. I was just going to say that this move is quite standard in the line. However, it's 1 of those moves that really, really annoys me when I play Garry. He's always forever find in that little nuance that most people miss. I had mentioned previously that his previous move that Garry played on his tenth move queen B3 that previously moves like bishop F4 and bishop G5 were a little more the standard and instead he kind of came up with, let's say, not the most common try. He played queen B3. And, of course, the intent of the move queen B3 was to attack this pawn on B7. If you notice that the queen -- yeah, okay. So the queen intent was to play queen captures B7. And if you'll look at that on the board, what you see now is a wonderful coordination of the queen with this bishop on this diagonal. The bishop on G2 and the queen on B7 are really strong to capture that rook in the corner. And I would have thought that DEEP BLUE would have thought twice. It was very, very quickly.
Mr. Ashley After going to B3, DEEP BLUE castled and now the possibility of playing queen takes B7 has presented itself, and if we just take a look over at Fritz 4 which is a computer program that's analyzing with us. It thinks that Kasparov will have a serious advantage, point 4, in its evaluation function, point 4 advantage, if the capture queen takes B7 is made.
So Kasparov, now knowing that he threatened queen takes B7, seeing DEEP BLUE ignore the threat of queen takes B7 must be coming as something as of a surprise to him wondering does DEEP BLUE sees something that no one else does. I am sure he would love to nibble on that b pawn, but is it going to be a case of the rat going for the cheese?
So Kasparov, stopping in his analysis now, head in hands, concentrating on this chess board. And I know he wants to take that pawn. And he would have just looked at me, laughed a bit and ripped that pawn off the board.
(Laughter.)
Mr. Ashley But again DEEP BLUE is a bit more formidable, so he's really giving it some thought as to whether or not he should capture this b pawn.
Mr. Seirawan He's definitely double-checking. And should Garry in the end grab that bit of cheese, the most likelihood is that DEEP BLUE will capture the knight, then we'll have a recapture.
Now, note that Black's rook is on that diagonal. So then would come the move rook B8, chasing White's queen all the way back to F3. But in this case, as we see, Garry has won back his pawn and he's got a very safe king. He's built this nice house for himself on the king side. And after a move like B3, well, in this particular end game, if you will, these 2 pawns, the A2 and B2 pawn give White a queen side majority, which, generally speaking, is an advantage. We might see a move, for example, like bishop B5. This is rather deep in the analysis. And we could see a move like knight C6, forcing the exchange of bishop for a knight, or even just a steady Freddie kind of move like rook to D1.
Is this what Fritz has been thinking about?
Mr. Ashley Well, it even suggested bishop G5, which is also possible. It really looks like a good position for Kasparov. And he needs this game. He really needs this game. He can't afford to lose it definitely. Draw would not be so bad, but, you know, a win would be so nice to go back and get the rest of the day and recuperate and just feel very solid and confident about himself.
So Kasparov now with a weighty decision, grab the b pawn or not, and here he's definitely thinking about it.
(Audience commentary.)
Mr. Ashley The question is how does the computer handle its time management? Maybe we can defer to Hans Berliner regarding this particular subject.
Mr. Seirawan Indeed, I'll invite Hans to answer that.
Just for your own edification, the match rules and the time control is similar for that of classical world championship match chess. That is to say, both players get 2 hours for their first 40 moves. And they can use that 2 hours in any way that they like. The normal human would just play his opening very quickly, come to very tough complicated middle game positions and perhaps spend as long as 40, 50 minutes on a single move. So the human -- and then at the end of the game, he gets into what we call Zeitnot, a German word for time pressure, where the human being has to move quickly.
And in this case does the programmer say to the computer hurry up buster make your move now or is it the computer that is making its own decisions? And Hans, maybe you can address that if you please.
(Audience commentary.)
Mr. Ashley So, in brief, Hans Berliner says it's all the computer's decision, and it's already been programmed into the computer to be aware of these significant factors. In complicated positions though it's interesting just on that point. Complicated to us may be basic arithmetic to the computer. You see what looks like strings of amazing variations. Wow, it boggles the mind. Let's start to figure out what move is the quick move here. And the computer is thinking come on, slow poke, I already know what the best move is. It's forced. Its ability is there. Its strength is there. So complicated, it's an interesting word we use. What's complicated for us obviously is not complicated for a machine that's looking at every moving and checking all the forcing variations and sometimes mate in 8 is seen instantly by the computer while we're sitting there going okay how does this work again if we try this and that the and the other thing and we're double-checking all the variations.
So we have to redefine the word. We can't just throw this word around as if we're speaking to fellow human beings about it because the computer redefines that. Complicated for the computer may be some boring, quiet position that requires a lot of analysis just to come up with the right move where we would say oh, this is easy. Just play this.
Mr. Seirawan Exactly.
Question in the back.
(Audience commentary.)
Mr. Ashley A question about the display in the lower right-hand corner of Fritz 4.
Mr. Seirawan What we're seeing here is the computer software program Fritz that we're running right here. And the idea of Fritz is not only to help us with the openings. That is to say, how deep Fritz's library so we can say, well, these moves have all been played. But down here we have Fritz's choices. He's looking at 49 alternate moves for Garry at this moment and Fritz has decided that in its judgment it would be best not to capture the pawn on B7 but rather to play the move bishop G5. And this is its evaluation function. This says that Fritz believes that White has point 0 6 of a pawn advantage.
Having a full pawn is like having a 1 point 00 advantage. So it thinks it has six one-hundredths of an advantage for White. That's Fritz's evaluation.
Mr. Ashley Which is next to nothing. Might as well not even speak on it.
Mr. Seirawan So Fritz thinks that although DEEP BLUE is a pawn ahead, it's not significant.
Mr. Ashley It's not a pawn it's going to keep.
Mr. Seirawan And, in fact, Fritz believes that Garry has a tiny edge even though he's behind a pawn at the moment.
Mr. Ashley Also, there's this depth function, depth function ten, which refers to the ploy, how deep the computer is going. And it's going 5 moves deep, as we understand it, 1 for each side. And also it shows a string of moves that it thinks is the most likely scenario in the next 5 moves. And as you said, bishop to G5 is its best choice with slight advantage for Kasparov, basically no advantage for Kasparov. Point 06 is too minuscule for words. But Kasparov, on the other hand, is still feeling the psychological pressure. And this we spoke about. The 1 thing that the computer DEEP BLUE will not ever have to face is psychological pressure. But here Kasparov feeling the heat, do I grab the B7 pawn or do I play bishop to G5. Indecision on his part regarding what the best move is and he's going to have to count on many human factors, 1 being his intuition, 2 being the state of the match, 3 certainly will be calculation. 11. Qxb7 Nxe5..
These are the kind of things that the computer is not bothered. I'm up 1 nothing. I'm down zero 1. It doesn't matter. And, finally, he has grabbed the B7 pawn, has Garry Kasparov. And we better hope, I should say Garry better hope, that this was not a poison pawn and he's not falling for a trap.
Mr. Seirawan A gentleman in the back has a question.
(Audience commentary.)
Mr. Ashley A good point made by an audience member. The operator is in fact allowed to make decisions in the case of the result of the game. For example, if DEEP BLUE were down a certain amount, let's say Kasparov managed to win its queen. Now, DEEP BLUE is ready to fight to the death. But it will also say I'm completely lost. It will show, you know, I'm down 9 point 00, completely lost. But, you know, let's play on a little bit. And not just play on a little bit. It will play on when it's a completely trivial position no matter the scenario. In that situation it's chess etiquette that we resign. We, as chess players, would give the respect to our opponent and say this is ridiculous. This is like being down 50 points in a basketball game and all the fans are walking out. So that is something as grandmasters, as top chess players we do. We say you've got the game. It's no point playing on. And you know let's go half have a beer or something. The computer will continue to play. The human operator makes the decision.
If DEEP BLUE falls behind a certain amount, it will resign. If the operator falls behind a certain amount, it will resign. So the computer -- the operator can make that decision.
Also in the case of draws. If there's a situation where it looks like there's a repetition of moves by force, like perpetual check, where there's -- you cannot vary from this, then the operator can say okay, draw and, you know, Kasparov could say no, let's check you some more, but obviously it would be a very trivial situation. Also if Kasparov were to offer a draw in what looks like a no-winning chance situation at all, the computer might be playing rook versus rook. For example, let's play 20 moves and see if you blunder. The computer will play those 20 moves. The operator knows it's a draw. The DEEP BLUE team will get together and say draw? Yeah, draw and offer it to Kasparov. So in those situations the operator can make a decision.
Yes, I remember this hand.
(Audience commentary.)
Mr. Ashley The question is, Yasser, it seems that since DEEP BLUE, and we know the DEEP BLUE team has put all of Kasparov's games into its memory, that Kasparov on the other hand -- this is the suggestion -- will not be able to do the same for DEEP BLUE. Does that not make it completely unfair on Kasparov?
Mr. Seirawan It is unfair in what you've said just now. The lady was pointing out that Garry has had a marvelous career, spanning several thousand games, and all of those games have been entered into the computer, whereas vice versa this particular version of the computer is absolutely brand-new, spanking brand-new, no history. Garry can't sit there and go through the games of the computer, and, therefore, cannot see what is a favorite for the computer. But it is unfair in the sense, from my perspective, that the computer should have no chance. I'm laughing because I was the 1 predicting that Garry would easely win this match and so forth. And so the computer of course won the first game. But what Garry also has is an extraordinarily creative mind. Okay? So he can literally play games with the computer.
What I mean by that is the computer can sit there and say okay, now I know what 1 of Garry's favorite positions and Garry can say I'm going to come to this match and not play any of them. And away he goes. He can find something absolutely unique that he's never played before, such as he played in game 1 and he had a good game, but then he became overly optimistic.
Mr. Ashley I mean that's the issue, isn't it? Flexibility. People want to believe that we're not a victim of our human programming and that we do learn from our mistakes and that we don't do the exact same thing over and over again. So in a given position Kasparov has always played the king's Indian so let me prepare against the king's Indian and Kasparov comes out and plays something totally different.
A good example is the match between Kasparov and Anand. Everybody knew Kasparov played the Najdorf. That's what he always played. The first 4 games Kasparov was always able to win none of those games. And suddenly Kasparov comes up with a drag on out of nowhere. No one could have anticipated the drag on would end up on the board and Garry Kasparov plays the drag on for the first time in his life and it stuns Anand and certainly if the computer had gone over games and anticipated all these moves, it would never have anticipated the drag on, ever because Kasparov was flexible enough and naturally he prepared himself and got ready in this game and played the drag on and he managed to win that first game against Anand. Kasparov can do that. The other thing is it's not so clear that the computer's moves are not anticipated by Kasparov. The fact is in chess there are only a certain number of good possibilities.
Hans Berliner was mentioning yesterday the move knight takes E5 has been played on the board, to stop that thought. Knight takes E5 by Black. Hans Berliner mentioned although the computer analyzes all these crazy possibilities, I mean 50 billion positions in 3 minutes, right, it seems unreal that we should be able to compete with such a force. But Hans Berliner made the point, he says because he himself programs these things. He said you will not believe the garbage that the computer looks at. He says you're looking at this position right now and in possible logical sequences.
And the computer is analyzing crazy things like right now the White king that's on G1 ending up in some variation on H4. It calculates that. It's looking at it. Does that position make sense to it? And it will look. It will look at completely random streams of moves because it's checking every possibility.
So although it's looking at all these positions, what a chess player is able to do is take your 50 billion positions and zero in on the 2, 3, 4 positions that matter or the ten positions that are really relevant and let's look at just these because everything else is just garbage. And a grandmaster can look at a move and go garbage, garbage. And the truth is it's harder to play a good move than it is to play bad ones. It's very easy to play bad moves, as I well know, very easy. Out of 40 moves on a chess board, maybe 3 are good. But the computer will analyze the other 37 and we, as humans, are able to get to the heart of it.
Mr. Seirawan Let me just get to the game. What I've done --
Mr. Ashley Have they actually gone to this position?
Mr. Seirawan No. What I've done is anticipate that the next few moves will be quite forcing. That is to say DEEP BLUE's rook on A8 is under attack by the White queen and White bishop. Therefore, I'm anticipating the rook moving, attacking White's queen, driving the White queen backwards. Then there's bishop on D7, if you notice, isn't very active. So I'm anticipating that do DEEP BLUE will bring it out to the B5 square and then White, using the same logic, this bishop on C1 isn't very active, will develop his bishop with the move bishop G5.
Mr. Ashley And in fact the rook has moved to B8, attacking the queen. And Kasparov has dropped it back to F3. And again, as you see just by a simple understanding, we're able to just get to the right position quickly. The computer got to this by analyzing about 600 million positions.
Mr. Seirawan Or more.
Mr. Ashley It looked at it and said that's under attack and move the queen and that's it.
Mr. Seirawan This is still all a part of the theory. I don't -- I think that there have been quite a number of games still played from this position. I just wanted to say that the position I currently have on the board is what I would start looking at if I was in my little laboratory at home, and I don't mean to say that. It's not more than a desk with lots of books and chess sets and lots of things spilled over
Mr. Seirawan I would start and say okay this is a position of the catalan. If I can't get an advantage from this position as White, then it means that I can't play the catalan for an advantage. And I'm certain that this position has appeared somewhere in Moscow on Garry's table and he's put a lot of analysis into this position. I note that Fritz is a big fan of White's position and that Fritz thinks that White's position is worth a whole half a pawn more. That is to say that White has an advantage of not a whole pawn, but half a pawn. That's pretty big, by the way. In international grandmaster chess, if I'm given a pawn advantage, I mean it's over. I'll wipe you out. I don't care who you are. Half a pawn is pretty good. I can work with that. So we're seeing that Garry believes that he has an advantage and the computer Fritz certainly concurs.
Mr. Ashley No. The question is does DEEP BLUE think that Kasparov has an advantage. And we will see, as the game continues, and it's not completely relevant whether or not you have an advantage. As you know, sometimes you have an advantage and mess it up. So there's still the problem of having to play the best moves, but you have to have confidence in Kasparov playing some good moves.
Mr. Ashley Yes, question here.
(Audience commentary.)
Mr. Seirawan I would start and say okay this is a position of the catalan. If I can't get an advantage from this position as White, then it means that I can't play the catalan for an advantage. And I'm certain that this position has appeared somewhere in Moscow on Garry's table and he's put a lot of analysis into this position. I note that Fritz is a big fan of White's position and that Fritz thinks that White's position is worth a whole half a pawn more. That is to say that White has an advantage of not a whole pawn, but half a pawn. That's pretty big, by the way. In international grandmaster chess, if I'm given a pawn advantage, I mean it's over. I'll wipe you out. I don't care who you are. Half a pawn is pretty good. I can work with that. So we're seeing that Garry believes that he has an advantage and the computer Fritz certainly concurs.
Mr. Ashley No. The question is does DEEP BLUE think that Kasparov has an advantage. And we will see, as the game continues, and it's not completely relevant whether or not you have an advantage. As you know, sometimes you have an advantage and mess it up. So there's still the problem of having to play the best moves, but you have to have confidence in Kasparov playing some good moves.
Mr. Ashley Yes, question here.
(Audience commentary.)
Mr. Seirawan The gentlemen has asked that now that the position is beginning to open up, isn't the position now becoming very favorable to DEEP BLUE because in open positions the computer plays well. The answer is both yes and no. What I mean by that is yes, it's good for the computer that the position is opening up because then it can put its raw calculating ability to better use, but it doesn't necessarily mean now the human is on a downhill. That is to say an open position, okay the computer plays better than a closed position, but in the open position the human player has a big advantage. Well, he's going to carry that advantage through. DEEP BLUE did not play the move that I anticipated. 11. Qf3 Bd6..
Mr. Ashley In fact, it played another very interesting possibility, bishop to D6 in response to Kasparov's last move queen F3 attacking the knight on E5.
Mr. Seirawan And that knight on E5 is a gorgeous piece. I mean, it was just attacking all of these lovely squares in Black's position. So that knight on E5 DEEP BLUE felt, well, it didn't want to --
Mr. Ashley It's got to go.
Mr. Seirawan It's got to go. That's nice.
Mr. Ashley But isn't there a really good response here for White in getting rid of the light squared bishop knight takes D7? Doesn't that help White in this position considering that he he'll have 2 bishops in an open position on a majorities on the left side of the board; that is the 2 on 1 at the left side of the board, pawns, wouldn't that be a nice situation for Whites to have?
Mr. Seirawan Yes. What Maurice is saying is that in chess you have what is known as the winning of the minor exchange. Bishops and knights materially have the same value. However, in the end games the bishops have shown to have a certain domination over knights. So it's often considered an advantage to have 2 bishops versus having a bishop and knight combination.
Mr. Ashley Not just in the end game. Here in this full middle game with all the open piece play, maybe the bishop will also have an advantage here.
Mr. Seirawan Indeed. That's possible too. I was going to say that knight takes D7 is a simple enough solution to secure an advantage that you're suggesting. Another idea, achieving the same plan, is to play knight C6, forking the rook and queen and provoking the move bishop takes C6.
Mr. Ashley I did something wrong.
Mr. Seirawan Okay. But I just want to go back and say that in this position I'm so enamored by that knight on E5. I mean I really like that knight on E5. Would I rather keep it there? And I'm just trying a tactical variation that probably isn't going to work because the computer will certainly calculate it. That's the move bishop F4. What did Fritz think of the move bishop F4?
Mr. Ashley Bishop F4 is certainly an interesting position. It looks as if you're blundering a pawn, but that in fact is not the case because of a response, it seems, like knight to C4 attacking the rook and the bishop.
Mr. Seirawan Attacking the rook with the knight and also notice that the bishop on D6 has now attacked twice the bishop on F4 and the knight on C4. So --
Mr. Ashley That is certainly a sacrificial possibility that is sound and Fritz 4 agrees that this would give serious advantage to White. So Kasparov, no doubt thinking about that as an option, double-checking the variation just to make sure he's not making a mistake. But he could play -- so he could play simply by taking on D7 getting the 2 bishops, or he could get the knight on -- he could capture on D7 with the knight. Which move would you like to see?
Mr. Seirawan I want to play bishop F4. But I don't want to see rook takes B2.
So I would say again that knight is a wonderful knight. Let's refortify it. Black could capture the knight, and again this would give White the 2 bishops, and he could capture that pawn with rook takes B2 and then we'd have a move like bishop takes D4 attacking the rook and attacking the A7 pawn. And after a move like bishop C6 -- I know this is a little bit deeper than normally I'd go for the audience -- bishop takes G2, queen takes B2, bishop takes F1, rook takes F1. Isn't this fun? This is what we do. This is what grandmasters do by the way. I mean when we're not reading or having fun, this is what we do. We just randomly go through hundreds of games. And what we have is this end game. But a very nice end game. And this is the kind of end game that computers aren't good at, and that is White has -- the material is completely even and by all rights a quite balanced game. 4 pawns versus 4 pawns on the king side, 1 versus 1 on the queen side. Equal material.
The computer will probably say okay it will make a draw, but this bishop is actually a very dominating piece and this would give White a serious advantage that he could use for a very long time. This position has already occurred to Garry and perhaps DEEP BLUE as well.
Mr. Ashley Definitely a nice move. If you can develop that piece, finally complete the development by connecting the rooks on the back, then why not in.
Mr. Ashley The rook is ready to jump into the game to D1 and C1, the d pawn on D4 that, Black pawn could become very weak. It's something that Black will also have to pay attention to. So at the moment Kasparov with the pleasant choice, which is always nice.
Mr. Seirawan Always nice.
Mr. Ashley Always nice. Certainly. And looking at the time, Kasparov has 118, I believe, left on his clock to DEEP BLUE's 121. So lots of time. Lots of time left for both sides. And in a position that's beginning to heat up.
You know, in chess games, chess is that kind of slow starting game. You know, you think not too much is going on in the beginning and you're waiting for the action to start and you really have to kind of be patient in the beginning because the tension just slowly, slowly builds and then the explosion happens. Something huge in the middle game. And we're still getting right at that stage. We're right on the cusp of it, where it's about to become extremely dramatic, and all sorts of complications will begin.
Mr. Seirawan A round of applause if you will for Maurice Ashley, who is going to be taking a break. Maurice will be back in about 20 minutes, and at this point it's a great honor, privilege and a pleasure to introduce the lead computer scientist programmer for the IBM's DEEP BLUE, Dr. C.J. Tan.
(Applause.)
DR. TAN This is probably as close as I will get to becoming a grandmaster.
Mr. Seirawan Well, it seems that your progeny, you must be very proud of game 1.
DR. TAN Game 1 was perfect. We'll see how game 2 works out.
Mr. Seirawan Indeed, it was.
Tell us a little bit because I am just fascinated with the engineering marvel that you've created. Please educate in layman's terms exactly what you've done and some of the problems that you've faced, please.
Mr. Seirawan Okay. The DEEP BLUE system or machine has gone through many transformations. The current version we're playing is in Yorktown Heights, IBM's research laboratory. And we're playing from here connected by telephone line to Yorktown Heights. So when the operator on the stage makes a move, the move then is transmitted to Yorktown. The machine will calculate and decide what next move will be, then will be reflected on the screen where he will then make the next move.
The system has what we call 32 processors or if you don't know what processor is, if you have a PC at home, a PC has 1 processor, 1 central processing unit. And this machine has 32 of them and all connected together to solve very large, complex problems. All right? And also what we have done with this machine is to add to each processor some special hardware that will allow you to do some search chess specific problems very fast.
So together with the 32 processors and the chess specific hardware, all the processors in this what we call parallel processing system are doing all these calculations, millions and billions together at the same time. All right? So now you may ask the question why can't I just get 32 Fritz together and solve this problem in the same way. Well, it's easy to put this 32 hardware together by some either telephone line or communications switch or whatever. The complexity comes when you try to take this same piece of software and make it run on 32 processing systems simultaneously together. Not only that you want to run 32 processors simultaneously to get 32 and improve performancr. So that's the key.
And if you can add more hardware, more software, more processors together, and as you add hardware your processing power increases, and you can add and keep on going on and add more processors. So that's the difficulty of constructing large super computers. And this is a project where we use chess as an example as a test case to learn how we build large systems, how we program complex problems such as chess so that as you build large problem, more processors, your performance can also go up.
And yesterday I think we found out that indeed it did pretty well. All right? So for the next few days, rest of the week we'll find out how really good it will be doing.
Mr. Seirawan In English of course you've got the expression too many cooks spoil the broth.
DR. TAN Exactly.
Mr. Seirawan And you're talking here about 32 cooks. So how come they're not spoiling the broth?
DR. TAN That's the key. That's why this computer, which we call SP-2, was announced by IBM in 1993, and it started out as a research project in research laboratory in Yorktown Heights. Many of the people in our group participated in that development, and in 1991 it was first announced as IBM product SP-1. 934 become SP-3. And today it's being used by many commercial companies, national laboratories, Air Force research lab, to solve many complex, civic and commercial problems.
And the reason why you don't have this phenomenon of 32 cooks to spoil the soup is precisely the key of processing, precisely the problem we want to learn more about by solving problems of chess in such a complex system. So the key is how do you coordinate all these processors?
For chess we examine millions and millions of chess positions to find the best move. Somebody said like you're trying to sift through tons and tons of sand to find a needle. So the key is how to synchronize our work and communicate very efficient communication for the processors, and that's the key.
(Audience commentary.)
DR. TAN The system, the SP-2 system is about 3 and a half feet wide, 4 and a half feet deep and about 6 and a half feet tall. And it weighs about 1,400 pounds for 1 unit. Each unit has 16 processors. And so we use 2 units. 2 units are connected together with the switch. The same switch that connects the 2 units together also connects all the other processors within the same unit so that each processor within the same unit talk to each other through this communication system, also talk to the processing units. So although they're 2 separate systems, logically they are performing this task as if they are 1 processing system.
(Audience commentary.)
DR. TAN If I hear your question correctly, is each processor working on the same -- no, we don't do that. If you do that, it willl be even more difficult to coordinate your task.
(Audience commentary.)
DR. TAN Well, opening time, it does opening libraries, and in the middle game, once it get out of the opening library, it does searches.
(Audience commentary.)
DR. TAN If the data base matches, if your prepared data base matches, what's going on, that's fine. Otherwise, you have to do searches.
(Audience commentary.)
DR. TAN That's a good question. What happened was something went wrong. It did not pick up the prepared opening library. So it's completely working on its own and it's not doing too badly at this point. We don't know exactly what happened, and it did happen.
(Audience commentary.)
DR. TAN Yes, we do that.
(Audience commentary.)
Mr. Seirawan This is getting scarier and scarier.
DR. TAN Am I talking too much?
Mr. Seirawan No, you're doing wonderful.
(Audience commentary.)
DR. TAN I don't know.
Mr. Seirawan How many lines of code?
DR. TAN I don't know. I don't know the exact number.
Mr. Seirawan But it was a huge number?
DR. TAN Probably.
(Audience commentary.)
DR. TAN The cost of the SP 2, if you buy the 32 processors we have, is probably 2 and a half million dollars.
Mr. Seirawan And a bargain.
DR. TAN We'll probably find out next Saturday.
Mr. Seirawan The question, by the way, was how many ply do you gets from the more processors. But now let me put you on the spot here, C.J.? How many ply in just say like a normal given position, middle game position like this, how many ply is DEEP BLUE seeing? By the way a ply is a half a move. So 2 ply means a move for White and for Black.
DR. TAN Right.
The previous version of DEEP BLUE, the prototype was looking at 11 or between 11 and 12 plys. I think this 1, because of much more power, is searching more than that. Exactly how many we don't know. Depend on the games. And also you have to understand, as you go deeper, 1 ply, the number grows astronomically. So it's a tremendous increase as you go deeper in your search tree.
(Audience commentary.)
DR. TAN How much power, electrical power? Gee, I didn't look that up.
Mr. Seirawan Whatever you can get, I imagine.
DR. TAN We did have to keep the power up in Yorktown.
(Audience commentary.)
Mr. Seirawan What about DEEP BLUE's Elo rating? Perhaps I can answer that. This version never had a rating because it never played a public game.
DR. TAN Right. So we will get probably an approximate performance estimate based on what happens here.
Mr. Seirawan The older version or the version just before this, the last generation, did it have a rating?
DR. TAN Well, I don't think computer has formal ratings. We estimate a performance rating of about 25, 50.
Mr. Seirawan So the previous model of a performance rating of about 2550.
I just had a gentleman come up to me after the game and said, well, for the first game you have an estimated rating of about 3200.
(Laughter.)
(Audience commentary.)
Mr. Seirawan Garry Kasparov rating is 2775.
(Audience commentary.)
Mr. Seirawan The question is IBM will keep their commitment to this project.
DR. TAN The goal of this project is really to learn how we examine complex tasks such as chess and so we learn from this and apply the technique to other application areas. So as we increase our computer strengths, architecture, add more systems, more processors to it, how do you paralyze a program, it will always be a good test case.
(Audience commentary.)
DR. TAN Computing all of them.
Mr. Seirawan The question is, is the special hardware systems used --
DR. TAN Used to accelerate the process of evaluation, search and move generation, chess specific tasks.
(Audience commentary.)
Mr. Seirawan Yes. Well, just to go back to the position at hand, I'm not going to let C.J. escape because I know you have a lot of questions. Garry indeed has taken some time. I don't know -- I guess he's in and around ten odd, 15 minutes. In my opinion, the opinion of Fritz, White is a little bit better, and right now I'm anticipating the move bishop F4, but Garry could be thinking about the move knight C4 attacking the bishop on D6. Now I'm stuck. Or just to go back here -- no. So he's got a couple choices.
(Audience commentary.)
Mr. Seirawan The question is, of course, Garry is behind in the match thanks to C.J. and his team, and Garry is under a lot of pressure. He not only has to win to tie the match, but he's going to have to win once or twice to take the lead and therefore he's putting a lot of pressure on himself. And that's absolutely true. The human being knows he's got to win. The computer is just playing the position. If it wins, a win is there. If it's not, go home and break bread. But there you have it. But I'd like to take advantage of C.J., and I know there's a little gentleman in the red shirt that's been dying to ask you a question.
(Audience commentary.)
DR. TAN Does the computer know if it's a draw or a win? The computer always keep going without stopping. And the operator has to decide whether to accept a draw or not.
Mr. Seirawan But C.J., we had that question for ourselves at dinner. The young man had said will the computer play for a draw or win. And you know as a human, if I take a lead, let's say I'm a boxer and I have a lead in rounds, well, I'll just dance around the ring and I'll refuse to engage my opponent and I'll win on the judge's score card. Now, you've got a victory under your belt. It must feel pretty good. But wouldn't you at this point like to draw the rest of the games and have you taken down maybe the contempt factor.
DR. TAN We don't want to draw, of course.
Mr. Seirawan Oh, going with the glory.
(Laughter.)
DR. TAN But those are tactical problems, concerns, strategy and so forth.
Mr. Seirawan But you as a programmer aren't -- it's not your strategy at this point to say to the computer go ahead and play some drawing variations. You're just going to let the computer --
DR. TAN That's right. You don't change the strategy once the game starts. You cannot change anything, cannot interfere with the program.
Mr. Seirawan Clear, but how about before the game?
DR. TAN By the way, we didn't play our opening library, so we don't know.
Mr. Seirawan Okay. Thank you. We'll take 1 more question. Make a good choice.
(Audience commentary.)
Mr. Seirawan Just a moment. The question was this is the 50 year --
DR. TAN Right.
Mr. Seirawan What's your view of the next 50 years.
DR. TAN What is the view of the next 50 years? Boy, I believe 50 years ago they will never have solved. I will be here playing Kasparov with the computer and it will be really hard to imagine what 50 years later will be. But 1 thing I am sure, humans will stay playing chess against humans. It's a creative game, involves ingenuity and all kinds of creativity that only human mind can possess, and human will continue to enjoy the chess game as it is today.
Mr. Seirawan On that optimistic note...
(Applause.)
Mr. Seirawan Well, that whole interview. As he was answering. I'm getting more and more worried about Garry's chances. But then he left it with us that only humans will have the ingenuity. So that was a very optimistic note. And I thank C.J. very much.
Maurice, welcome back.
Mr. Ashley Thanks, Yas.
Mr. Seirawan Did you get the scoop in the hallways? How we doing?
Mr. Ashley Actually I was just nibbling on something.
Mr. Seirawan I messed up on the computer. I was thinking that, going back to the position at hand, what White needs to do is he needs to develop the rook on A1 and the rook on F1 to bring both of his rooks in the game, but he's got a slight problem with his B2 pawn which is under attack by rook on B8. Therefore, I was thinking that there was the move knight D7 and the move bishops C4. Another move that I was just looking at was the move knight C4 and I was wondering what Fritz thought of this move. You attack the bishop on D6. You defend the pawn on B2, and then you prepare to bring the bishop out.
Mr. Ashley I think the problem here, a critical problem here for Kasparov is he has so many reasonable moves. Right at this point he could play all these moves you mentioned and it wouldn't be bad. The question for Kasparov, I mean he has several questions to deal with. One, he would like to know what the best move is. That would be very nice to move. But maybe there isn't a best move. Maybe it's the type of situation you end up in. A certain different kind of situation. For example, he could take the light squared bishop clearly. He could just play knight takes D7. Very simple. I have the 2 bishops. I have your 2 bishops, and what's the problem? I have a long term advantage. That's 1 position.
Another position he could play is knight C4 going after the dark squared bishop and controlling the dark squares, and he may not get the bishop immediately, but he'll be able to bring his pieces out without this change, this imbalance that's going on. And knight C6 is a different way because his queen ends up on the C6 square and that creates new tactical possibilities on both sides.
So Kasparov is at a big crossroads in the game, and it's not unusual for a strong chess player to stop in this position and think for quite a while.
We've seen Kasparov now thinking for 25 minutes on this 1 move.
Now, we talked about time management, and yesterday it was a really critical issue for Kasparov because he got into heavy time trouble near the end of the game, and he ended up just losing very quickly with the time pressure on him. The time pressure wasn't that significant at the point he made his significant mistake but still under pressure it's much more difficult to defend a difficult position. So time management is key here. It's square I because the computer will never face time pressure.
Mr. Seirawan That's right.
Mr. Ashley That's square I. That's part of chess strength? How would you feel if you're playing against an opponent you knew, a guy who never got tired who was always fully concentrated on the chessboard throughout the whole game, who managed time properly throughout the whole game.
Mr. Seirawan I couldn't insult.
Mr. Ashley You know, these 3 things are a part of chess strength, fitness part of chess strength, concentration. I mean concentration, the focus, not getting tired.
Mr. Seirawan Indeed. And I'll just add 1 other advantage, is that resourcefulness. I mean essentially when I play against a human being, I'm going to have that ebb and flow. I've got a big advantage. I feel good about myself. I kind of strut around feeling good, and I lose my edge, and suddenly I slip, and my opponent has the advantage, and then there's the struggle that goes back and forth.
So, normally speaking, even in the world of chess today, when people say well, what are some of the strengths of Yasser Seirawan, they'll say, well, he's a resourceful opponent. He's a tough guy to defeat just straightforward. But against the computer, the computer is not going to allow you that sudden opportunity to enter into the game and to be resourceful. So you're absolutely right.
And I'll just say another story. We heard earlier from Monty, Monty Newborn, the chairman of the ACM and of the chess project here. 25 years the ACM has been sponsoring computer chess competitions all over the United States, and I think it was 1976, I was a Seattle teenager. The ACM competition was in Seattle and I had just won the state championship so I was feeling really quite cocky, and these ten boxes, I'll easily defeat them. And the first game I played in front of a large audience the computer beat me. I felt terrible. And then I really, really had to work very hard, as we're seeing Garry work very hard today, to come back and beat the computer. 11. Nc6 Bxc6..
Mr. Ashley Well, he has finally moved. After much thought, Kasparov has decided on the move knight to C6. He used at least 20 minutes, I think more like 25 minutes on his clock --
Mr. Seirawan Absolutely.
Mr. Ashley -- to get there. It may not seem significant now because he has an hour and 15 minutes remaining, but it really was that he spent so much time in that single position.
Mr. Seirawan Right.
Mr. Ashley And that is the thing about a situation like that, is it consumes energy as a chess player. You know, you lose that energy as you're spending that time on that 1 position so concentrated, so focused; and you have to try to keep apportioning the proper amount of time for yourself on every single move. Otherwise, you could just end up in serious time pressure when a position gets even more complicated. The more the position goes along in that middle game that we're looking at now, the more possibilities start existing, arising and you have to think about all of these possibilities in the game. So Kasparov, with the time spent on that single move, has lost some energy, definitely some physical energy for the game and some of the time on his clock. The computer of course still has a lot of time, 121. They're both just about even. We can't speak of time pressure at all, but we will keep a watch on that clock throughout the rest of the game. It seems completely forced to move bishop captures C6.
A knight is a minor piece, not as good as either a queen or a rook. A bishop of course is of equal value to a knight in most positions. So we're expecting bishop captures C6 and queen captures C6, but of course the computer has to analyze its hundreds of millions of positions before coming around to what we know. More to come...stay tuned!
Mr. Ashley Yasser, you think that queen to E7 is a good idea?
Mr. Seirawan Well, indeed. There's something here. I hate to fault Garry here a bit, but my nose tells me that knight C6 wasn't the right move. And I do very, very much respect and appreciate the strength of Garry Kasparov. And like you said, he spent 20 odd minutes giving up the time advantage that he had gotten. The point here is that after this virtually forced swap of pieces, bishop takes C6, queen takes C6, and then I just suggested the move queen E7 for DEEP BLUE. If you count the number of pieces that Black has developed --15. Qxc6 e5.
Mr. Ashley And it has indeed captured on C6.
Mr. Seirawan Queen rook knight bishop rook.
Mr. Ashley Kasparov has captured queen takes C6.
Mr. Seirawan Note that White has a rook and bishop that still need to get into the game. Black's next move, for example, after a move like B3 preparing to develop the bishop will be rook C8. And Black's rooks are ideally placed. Plus he's going to be able to follow up by moving his pawn up in the center. And, in fact, Fritz now, well, just when I spoke, had gone to an absolute evenness. So I don't know about this move knight C6. And, again, it was certainly a move that Garry thought a great deal about.
Mr. Ashley So the question, long-term question, is the advantage of the 2 bishops? White has 2 bishops, a particularly potent light squared bishop on on C2. Is it going to be balanced, counterbalanced, by Black's lead in development after a move like queen E7 where he has more pieces in the game than does White. And then a move like rook to C8 attacking the White queen and therefore having even more tempo to play with. That's going to be a critical issue in this game, and we'll see whether or not DEEP BLUE will be able to make the most out of its lead in development while Kasparov tries to wet that down and bring his 2 bishops to bear on the position.
Mr. Seirawan Exactly.
Mr. Ashley But Kasparov looking very comfortable. After playing queen C6, he promptly got up from the board. So he must be feeling quite relaxed, or maybe he's really nervous and can't sit there anymore. But it seems to me that after that long decision he's content with where he is in the game and probably feeling like he made the right decision.
It seems, looking at the clock now, he has 105. The arbiter adjusting the clock in the room so it's easier for us to see. He's not doing a great job of it at the moment.
Mr. Seirawan Garry has used 55 minutes.
Mr. Ashley Garry with 105, not 115. Therefore, he thought even longer than we thought on that last move. 105. And DEEP BLUE with 118 left and now this very aggressive response by DEEP BLUE, the move E5 advancing in the center with its pawns has been played. E5 a very aggressive move. Those pawns could march down the board at the moment the e pawn's advance is curtailed by a bishop and a queen for White. But those are aggressive intentions, Yasser. The computer is thinking win win win.
Mr. Seirawan He's going big. 1 of the features of the Grunfeld defense, and we'll go back a bit because there's this pattern recognition between Garry's choice of opening today and his choice of defense when he's Black. He loves to play the Grunfeld.
And one of the things that often happens in a Grunfeld is that Black creates the center; that is, he allows his opponent to have a large mass of pawns in the center and he plays with his queen side majority and with this bishop on G2. So this move E5 was definitely within Garry's calculations. He probably didn't think so much of it because of this enormous Grunfeld experience. He probably is very happy when his opponent has the center.
If White was to play the move B3, if Garry was to choose B3 in this position, I don't like this move because that square on C3 becomes weak, and after a move like bishop B4, the bishop could plant itself, anchor itself, if you will, on the C3 square.
So going back for a moment, since White can't just develop his bishop because this pawn on B2, it's possible that Garry will just try the move rook B1 instead, anticipating that he will develop his bishop with bishop to D2 or bishop to G5 on his next turn. But I can't speak of Garry of having an advantage in this position. I definitely think that it will be a long term -- it will be a long game before White is able to prove anything. 16. Rb1 Rb6..
Mr. Ashley Yeah. It's a bit of a -- even in this position, for example, the 1 you're talking about, if Kasparov were to play rook to B1, to defend that b pawn, preparing to develop the bishop, maybe not a bad move is H6 for Black, limiting the bishop's possibilities and so forcing it to choose more passive squares like D2 for example, and then maybe Black can get something going. Well, certainly not looking like your wonderful Kasparov position is this 1. Kasparov certainly with enormous experience, as you said, in these types of situations so he will be looking to figure out what can be done, extract the most out of it, but not the kind of position you would think he really, really wanted to have against DEEP BLUE, particularly I think because it's a really open position. I mean truly open. He has, in fact, played the move rook to B1.
Mr. Seirawan Right. Garry played the move as anticipated. And it's positively time. It's positively time. We'd like to --
Mr. Ashley Yasser likes polls, people.
Mr. Seirawan I like polls.
Mr. Ashley I don't like polls so much, but he likes polls.
Mr. Seirawan I like to know what you guys are thinking. Hans Berliner told me yesterday the audience is always right. The audience is always right. So on that note, what I'd like to do is invite people to raise their hand for those who think that the position is completely even, completely even, that the game will probably be a draw.
Mr. Ashley Wow, a lot of people.
Mr. Seirawan And I'd like to get the opinion of those who trust in human beings and Garry, believe in Garry's strength.
Mr. Ashley A little less.
Mr. Seirawan I see a lot of people here who were influenced by C.J. He made a strong position.
All of those who think the computer is going to win?
You can all leave the hall. The door is open. Get out of here. The all star game is going to start soon.
Mr. Ashley I mean obviously if DEEP BLUE wins this game, Garry is in deep trouble.
Mr. Seirawan Thank you. Deep do-do.
Mr. Ashley We really have to hope. And just for the interest of the match that he's able to deal in this game and maybe win the game. He has long term possibilities. And they're twofold. It's not just the 2 bishops. Again, it's the 2 on 1 majority on the left side of the board, the a pawn and the b pawn versus the single a pawn. That gives White the opportunity in the very, very long term to create a past pawn by advancing those 2 pawns down the board exchanging 1 of them off for the a pawn and therefore having a passed pawn, and that pawn is going to tie up Black's pieces. That is extremely significant.
Thinking about this position, I remember watching a speed game played between [Ilya Goravich] and Walter Brown.
Mr. Ashley At the world championship.
Mr. Seirawan What has happened? They've got a camera glare. I think there's a light.
Mr. Ashley We can see Kasparov's time, but it's tougher to see the computer's time.
But my point was that in response to this last move, in response to Kasparov last move, rook B1, the computer has played rook to B6 attacking the White queen. Now, this game between [Ilya Goravich,] who is a specialist of the Grunfeld defense, a specialist, versus Walter Brown. I believe it was with Walter.
Mr. Seirawan Yes.
Mr. Ashley They played this game where Ilya whipped off the moves as Black. This is a reverse position. Remember, White is playing a Grunfeld in reverse. He's playing it with the White pieces. Normally Black would choose this setup. But Ilya won so easily. He just went bing, bing, bing, bing, bing. After the game I said Walter, your position was bad. It was very easy for me to do. I know this kind of position. And what the big point was was that 2 on 1 majority on the left side.
White was Black in that game, but with the reverse position it's White in this game was able to push the pawns down the board and create a passed pawn effortlessly. And the bishop on G2 would just support that passed pawn as it went down the board, and the rook is already behind the passed pawn. So that is the long-term advantage. The question is how is that counterbalance by that strength in the center of the board?
And a lot of us learned, Yasser, that it's better to have the center than to have the wing. If you can control the center, then you're doing really well. But in some situations the center is the target and you can't really push those pawns too quickly down the board. 17. Qa4 Qb8.
Mr. Seirawan I will get to your question in just a moment.
In chess you've seen a phenomenal evolution of style; that is to say, there are many schools of thought throughout the history of chess players. You had the Italian school of thought, the British school of thought and so on, and almost all of these have in common a classical style of playing chess, and that involves control the center. Along around the 1920's you saw the hypermoderns develop this idea that you could attack from the wings. You'd give up the center. You surrender control of the center, and you attack like a [pinzer] movement. You just saw DEEP BLUE blue play rook B6 attacking Garry's queen. Garry did bring his queen back to A4, as I expected, and now this pawn on A7 is under attack and DEEP BLUE will have to take care of that.
There's a gentleman who has a question.
(Audience commentary.)
Mr. Seirawan The question is it's a 6 game match and regardless of the score will all the games be played and, yes, they will. Yes. So we can have a wipeout.
Mr. Ashley So poor Garry, if he's down 0-5, he's got to go back for a sixth game.
(Audience commentary.)
Mr. Seirawan Indeed. The gentleman refers to the position at hand. And we'll go back just a bit. And we get back to this position that Garry had this deep, deep thought. And he ended up moving his knight to C6 -- oops, what am I doing? Okay. He moved his knight to C6, thereby winning or forcing the swap of knight versus bishop. And why didn't he just do that immediately? Okay. If we go back to this position where he had his deep think, Garry's queen is on F3. Okay. The current position he has his queen on A4. And in his mind it was worth the few tempo that he has lost to bring his queen into this active position. I like the queen on F3. I think I would have just taken the bishop. And that's a question of style More to come...stay tuned!
Mr. Seirawan But my nose tells me that Garry wasn't enthusiastic about the move knight C6. Normally when he's enthusiastic, the whole tables start shaking as he's writing down his moves. He seems a little subdued at the moment. 18. Bg5 Be7.
Mr. Ashley I don't know. Looking at this position, it's difficult to calculate the long-term advantage. It's very difficult for the computer to calculate a long-term advantage. And here White has 2 substantial advantages The bishops and the majority on that left side, on the queen side. That's not something that you calculate. The computer does not calculate pawn races very well. Because it just really adds so many possible moves in a given situation.
Now it has played queen to B8, defending the pawn and still preventing the bishop on C1 from developing easily, but Kasparov just flashing out his response after queen B8, he just played bishop G5 in a second, and he's getting up from the chessboard. Typical Kasparov when he thinks he's played a great move. He makes the move instantly and walks away and usually intimidates the human opponent. Not to intimidate the opponent here. In fact, Fritz 4, analyzing with us, thinks that Garry's last move was not at all wise and gave Black an advantage.
Mr. Seirawan For just a brief moment, Maurice.
Mr. Ashley But as it's continuing to calculate, it now believes that Kasparov's move may be okay.
(Laughter.)
Mr. Seirawan And actually --
Mr. Ashley These computers have serious contempt.
Mr. Seirawan Indeed. I was just going to say that Garry jumped on that move bishop G5 and he was very well prepared for the move rook B8. So I think that this is all still part of his charismatic way back when he had that big think knight C6. Let's just take a quick look at what I would expect to happen. If I had played the move bishop G5, knowing what I do about Garry, I'm sure he has prepared a variation like this. Looks like a very straightforward variation, rook takes B2, rook takes B2, queen takes -- rook takes B2, queen takes B2. Now comes a move like bishop takes F6. G F6. Note these 3 pawns are all on dark squares, which makes this bishop rather impotent. This bishop on G2 is going to be a very, very strong piece after the further move queen D7. The idea here is to attack this bishop on D6 after a move, for example, like bishop C5 we have this beautiful light square domination, queen G4 check, protecting the pawn on E2. A move like king H8 is forced and now bishop E4. And if you just take a look at the position, White is a pawn down, but on the other hand this F7 pawn is the extra pawn. And after further moves like queen F5 he would win the game.
Mr. Ashley So DEEP BLUE was wise enough not to allow its structure to be demolished on the kingside where its king is. It doesn't want to see its king opened up. If you remember yesterday's game, Kasparov willingly let his kingside get broken up to his detriment later in the game, but here, giving a lot of value on the safety of its king, DEEP BLUE has decided to play a very safe move, bishop to E7 protecting that knight and definitely for the moment adding pressure to that b pawn. Something is going to have to be done about that b pawn right now because it's definitely under attack, and a big question for Kasparov is how will he defend it.
You mentioned earlier that the move B2 B3 weakens the 3 c square, and if that square is weakened, maybe something like a bishop or even a knight potentially could land on that square or even a rook for that matter could create some possible penetration along the c line. So White's going to have to be very careful about its next moves.
The thing about pawn moves is every pawn move creates a weakness in your position because you can never bring pawns backwards. So once you've weakened a square, it's weak forever, forever. Because the pawn can never come back to defend it. And a move like B2 B3 weakens the C3 square, and that will be the case forever. Whether or not it's bad for you depends on other factors in the position. But that will be the case forever. And it's something you're going to have to accept in a given position. So Kasparov will be very careful about his next move, whether or not he'll move the pawn to B3.
(Audience commentary.)
Mr. Ashley Very interesting question.
Mr. Seirawan The question the gentleman asked is computers, or at least the ones that you and I, Maurice, have played have insatiable appetites for material and pawns. They just grab whatever isn't nailed down. And the fact of the matter is DEEP BLUE here registered a rather very human quality there to stop and say, hey, wait a minute, you're giving me something. What's behind this offer? This is a Greek offer. And I'm not going to take it. So yes, I think that's quite -- and, again, C.J. and his team have to be applauded because I think that's a big step forward. Yes, indeed. I would have expected the computer to grab that pawn just as Kasparov offered it and leap on it.
Mr. Ashley I don't know, Yas. The variation you showed before seemed to hold tremendous dangers, and I think it's still within the computer's ability to calculate that variation. The 1 I'm referring to is after rook takes B2.
Mr. Seirawan Right.
Mr. Ashley If you go back just a second before bishop to E7, rook takes B2.
Mr. Seirawan The previous move, yes.
Mr. Ashley Then rook takes, queen takes, and bishop takes F6, G F6. Queen D7 is with tempo and if the bishop were to, say, casually go to a square like C5, and I think a move like queen F4 is also -- F5, excuse me, is also very interesting, threatening a pawn and also threatening bishop to E4.
Mr. Seirawan Exactly.
Mr. Ashley I don't know if it's a bad move, but I think that moves like that the computer would have to be concerned about because the light square domination could become too much. So I think the computer probably calculated this variation, maybe made better moves than I'm making now, but probably calculated and said this is not what I'm going to go for because it's concrete. And then it just decided on the other move.
Mr. Seirawan Let's take a look at the position at hand. And I'm just wondering if I was -- Maurice has spoken rather eloquently about what we can do with our pawn majorities, and we all know that the only good of having an advantage is to utilize it. So to that extent the pawns have to be pushed. But if I push the pawns right away, well, after B4 rook takes B4, I'll end up just losing the pawn. 19. b4 Bxb4.
Mr. Ashley Well, Kasparov has agreed with you and played the move B4.
Mr. Seirawan But what I wanted to say is you take a look and the pawn gets captured. However there's a deeper view and that was after rook takes B4 queen takes B4. Queen takes A7. Well, I've done what I've set out to accomplish, which was to create a passed pawn on the queenside. I've just traded pawns.
Mr. Ashley This potentially could be a huge advantage for White, that a pawn rushing down the board to become a queen, potentially could be a large advantage.
Mr. Seirawan So again after B4, if rook takes B4, rook takes B4, the other alternative is bishop takes B4, still keeping that pawn on A7. However, I do have the move rook -- therefore, after the forced move A5, would I then play A3 and win a piece. Okay. So what I meant by all of this was that B4 is actually a reasonable move because the --
(Laughter.)
Mr. Ashley He's good.
Mr. Seirawan Because I'm doing what I'm trying to do, which is push my queenside pawns up the board.
Mr. Ashley Good variations, Yas.
Mr. Seirawan I thought the New York crowd was tough. These guys are really tough.
B4. If I can, the idea being that I would love to play B5 and A2, A4, just to get my queenside up the board. Keeping in mind that those central pawns would light up the board already, I'm playing catch-up here.
Mr. Ashley This reminds me so much of that Walter Brown/Ilya Goravich game, where the queenside pawns were just devastating. Just push the pawns down, and that was the game. It was very easy to do, and I think Kasparov is very proud of his last move, and if he can get his pawn to B5, he would be very happy.
Mr. Seirawan Oh. 20. Bxf6 gxf6.
Mr. Ashley But this was the other square I point that I think was even more ingenious by Kasparov, was the positional realization that on the right side of the board the computer would not calculate everything, would not calculate the bind, would not calculate that a light square bind would get set up and that in some future fantastic position Kasparov would be able to bring his pieces over to that right side of the board, and now he's ripped -- after bishop takes B4, he's played bishop takes B6 and, to be sure, bishop takes G6 will be played, and I think Kasparov has a wonderful position to be played. No doubt DEEP BLUE will refute the attack, and our position will look zilchy. But I think in our experience, if we have to commit ourselves to this position, okay, Yas doesn't want to commit himself.
Mr. Seirawan He's doing it.
Mr. Ashley So I'll commit myself. I'll be the successor. I think that White has great compensation. I don't see Black's extra pawn. I don't see the advantage that Black has. Because 1 thing -- and this may be just deluding myself, that the variation is good, if you look at this position after g captures F6, g F6, Black's pawns are doubled and double pawns tend toward weakness, and we almost don't even count them as an extra pawn unless they can start flying down the board. So I think it's material quality as far as White is concerned and he'll have a long-term bind on that right side of the board.
His question, only question for White, and I know for sure he'd have a good position, is if the White queen can get over to the right side of the board. That's the big question. After g captures F6, which is certain, then Kasparov will more than anything else want to bring his queen over to the right side of the board. Now this is a genuine positional sacrifice by Kasparov with the awareness that he's going to have a good game.
Mr. Seirawan I'm doing something wrong. I'm looking at bishop takes B4 as we've seen.
Mr. Ashley Just go to it. And computer has in fact played g captures F6, and a big decision by Kasparov. I think he has a couple of moves here, Yas, how to get -- make sure that bind exists forever on the right side. Bishop E4 certainly looks good. Queen D7 is another possibility, get the queen over there. Queen C2 is another possibility to get the queen over there. There are a number of possibilities to get a light square bind. We talk about color complexes. We have the opposite colored bishops and a lot of people -- Yas, maybe you can talk about this. Opposite color bishops for chess colors mean oh, it's a draw. But for grandmasters, oh, it is usually very different.
Mr. Seirawan Yes. The idea there was -- there's a wonderful book by David Bronstein, The Theory in Practice, Chess Theory in Practice, The Struggle, where he talks about playing on a complex of squares. So what has happened is that White has sacrificed a pawn in order to get the domination of these light squares, especially the D3 E4, F5 squares.
Black, conversely, if you notice a great deal of his pawns in the center and his queen rook bishop are all on dark squares. So it's almost like the 2 sides are playing checkers, but 1 guy gets the light squares; the other guy gets the dark squares.
In this particular case when you've played a light square versus dark square complex position the key as to who has the advantage is decided by the role of who is attacking. I've played Michael Tal in the candidates world championships in Mount Pilyear championships in 1985, and I had bishops of opposite color. But because I was on the attack I won the game and I won the game very easily. The realization of the victory was quite simple. Because I have a light squared bishop and my opponent had a dark squared bishop, my opponent couldn't offer a swap of pieces.
In other words, my opponent didn't have any counterpart to my piece and therefore I was playing as if I was a bishop ahead because I had all the attack on the light squares and he had all the attack on the dark squares. But I had the attack
Mr. Seirawan So the idea here is what White wants to do. What Garry wants to do is bring his king to the queenside and especially with the move queen F5 and bishop F4, what Maurice pointed out in a previous variation. And it may be that DEEP BLUE has to play a move like queen C8 to stop that. But if you note that after queen C8, this pawn on A7 will have been left on prix. That is a French term for end capture. And White will have won his pawn back. So after a move like rook B8, well, Garry would be very happy with his position. He's a pawn ahead and those light squares on Black's kingside are still weak.
Mr. Ashley And there's a past pawn. That a pawn is past actually on that left side of the board. So that is -- those are the critical questions for Kasparov, whether or not DEEP BLUE has already seen this position and feels that it can defend is in question.
Right now Fritz 4, Yas, says that it thinks that Black has a slight advantage, point 34. But let me tell you, in my opinion, that is an admission of the weakness in Black's position. Because there is no direct way for White to win the pawn back, at least not being shown by DEEP BLUE, some very -- by Fritz. Some very concrete way of winning the pawn back, which means in its mind it's saying okay, I'm up a pawn, but this guy has some serious compensation for it. In the value of point 66. Because it's saying that Black openly has a point 34 advantage. So instead of giving itself a 1 full point, I have a big advantage of a pawn, it's saying no, no, I may have a pawn, but, boy, there's some serious problems in my position so I can't say I have a full pawn advantage. I'm only going to give it point 34. That means full compensation in my opinion. So my instinct says this is good for White and there must be some way for White to at least create a balance, at least get a pawn back. But maybe even press something home in the long run, and we could see something very dangerous very quickly. So --
Mr. Seirawan And, there again, going back to the gentleman's reference, the computer is really happy to -- 21. Qd7 Qc8. 22. Qxa7 Rb8..
Mr. Ashley And boom Kasparov has played queen D7.
Mr. Seirawan Right.
Mr. Ashley And queen C8 instantaneously by DEEP BLUE. Clearly aware of the takers facing it on the other side of the board, and Kasparov, it seems, could win his pawn back. And he has ripped the pawn off the board. Queen takes A7 and Kasparov has regained his pawn. Now, Yas, there's a very exciting varying here. What if the move rook to A6 is played? Well, what does that mean for this position?
Mr. Seirawan I just wanted to say a couple of things. The first is you're absolutely right that Fritz wasn't giving Black full credit. It was saying hey, wait a minute. There is a lot of compensation here. But Fritz actually liked Black's position. Whereas myself, as a human international grandmaster, would sit there and say you can have Black's position. I'll take White's. I'll take Garry's position. And as we saw, when Garry played the move queen D7, the DEEP BLUE program said wait a minute, time out. I mean this is very serious, what is going to take place on the kingside. I'm going to give the pawn back. And now we have a position where -- well, Fritz thinks it's balanced, even though Garry just won his pawn back and Black still has these weaknesses on the kingside.
Your question was what about the move rook A6. Well, after the move rook A6 White has but 1 move, queen B7. Now, do keep in mind that I'm attacking the bishop. If your bishop were to move, for example, bishop down, I would trade queens and I'd have the old square.
Mr. Ashley Hello. Bishop B7 attacking both rooks. Not nice.
Mr. Seirawan So bishop C3 would be out of the question. And what Black might have to do is trade queens.
Mr. Seirawan Rook A4.
Mr. Ashley I don't want to play rook A4. Maybe I'd play rook B6 in that position, like so.
Mr. Seirawan Okay, rook B6. And then again what we've been seeing here of late is this nice little move bishop E4. And this is the end game that's very nice for white. He's blockaded this extra pawn and this pawn on the -- on A2 is actually the most crucial pawn in the position. So Garry would have a very large advantage, I think in this game.
Mr. Ashley I think that Kasparov has finally found a very nice way of creating the kind of position a computer really does not excel at. And it's the kind of long term strange imbalance in the position that a computer would have trouble calculating. For example, that a pawn. What the computer is looking at is billions of positions. It has to deal with -- it's going to ignore a lot of positions with the a pawn just pushing itself down the board. And Kasparov may not just push the a pawn down the board. He may not just go A4, A5, A6. But for us, we immediately kill the rest of the noise in the position. We're not looking at all those fancy moves. All we're looking at is the a pawn. How do I get the a pawn down the board? 1 problem to solve, how do I get the a pawn -- so instead of the computer has all these problems looking at trying to figure all these moves out, the human immediately gets into what the mate of the position is. And that mate from Garry's experience and our position is the a pawn. Not only that the bishop on G2 guards A8 where the pawn is going to score the touchdown. It's right on A8. So that bishop is ready to support that pawn when it goes to the final square. So Black is going to have to deal with that pawn on a dark square, slow it up before it gets that far and it's going to have to deal with it either before it gets to A5 or A7 because there's no way to deal with a pawn when it lands on A8. You have to give up a rook for it.
Mr. Seirawan Right.
Mr. Ashley So that's major plus for White. The other possibility that still remains is the question of the king's side, whether those Black pawns start going down the board and try to create some kind of havoc or will Kasparov be able to create a bind on that side? At the moment it doesn't seem as if he will be able to create a bind. It seems as if he can't organize a real tight bind on the right side.
Mr. Seirawan I think what Garry -- what the computer has had to do, the computer gave up that A7 pawn to prevent the bind. So let's get -- let's accept it as a given. And again this is what Garry did yesterday. He fell into a cardinal sin. What he did is he tried to force his attack home. Instead of nursing his positional advantage to fruition, he said calculate, calculate, calculate. I win. And I went after the attack. And in this position he should shift his mode of operation and he should be trying to push that pawn home, that a pawn home. And I was going to say the most direct method is just to do it. Is that a Nike commercial? You know, what's going on here?
Mr. Ashley From the sound of it.
Mr. Seirawan Push that pawn right up the board. If Black does nothing, I'm going to play A5, A6, just as Maurice said, and get the pawn. I'm not too worried about trying to establish a light square blockade, because now I'm going to put my money here with the a pawn. After A4 I like white's position. Also I just would say you have to admire what's happened in the last few moves. We went from a situation where Fritz was giving White an advantage. There was a pawn sacrifice. Fritz was giving Black the advantage and White's regained the advantage after 4 series of moves. So what Garry went through was a situation where he sacrificed his advantage only temporarily and he saw that he was going to win it back. And well he's got to be admired.
Mr. Ashley I think so far you have to be impressed by Kasparov's moves. He's got the game completely under control. He's -- you know, it really is kind of funny. For Kasparov to win, it seems to me that he has to somewhat play boring chess. He has to put us to sleep. Because if he lets it get too funky and too exciting the computer is going to figure everything out. We saw that happen yesterday. We all got excited. Wow, Kasparov is on the attack, this is going to be good. And the computer say uh-uh of and it worked everything out. And we're looking for mates and, oh, man it's not mate. It's nothing. And the computer was completely sure of the position and Kasparov just lost trivially. What Kasparov has to do is not make it get exciting and wild, as he would do against a human. He would love to do that against a human because that's his style normally. But he has to actually restrain himself. Yesterday he couldn't do it yesterday he was true to Kasparov. Went for a mating attack and it failed horribly. He knows before he came into this competition he has to restrain himself and play a kind of chess that yeah we might think this is not too exciting, Garry, you're beating it, but that's how you're going to beat it. Yeah, that's how he's going to beat a monster that can calculate everything. And that's what he's doing now. He's keeping the position completesly calm. He's making sure no real complications might occur. I might complete mini complications like the pawn sac, but he's not going to create wild complications where his strength will diminish relative to the computer.
Mr. Seirawan Can we just take a couple of questions.
Sir?
(Audience commentary.)
Mr. Seirawan The question is, the assumption the gentleman has made is that even the best chess players have coaches and trainers and does Garry have somebody that he goes over -- over yesterday's game with. And the answer is absolutely. Garry has some of the best chess trainers and coaches in the world and they oftentimes travel with him. I know that Garry brought from Germany Frederick Freidel, computer specialist, to help Garry in his computer chess understanding and preparations. And normally speaking Garry does travel with his trainers, although I don't know who he has brought with him specific little for this match.
Mr. Ashley Sometimes that's top-secret information.
Mr. Seirawan Other than Frederick. You get the insight openly months later. 23. Qa4 Bc3, .
Mr. Ashley Kasparov has played queen A4 and I must say I'm quite impressed by the move. Queen A4. Remember weep were that you caning about Kasparov keeping the game dull. No funky possibilities for the computer. Now, after A4 which looked like a very reasonable move there was the problem of queen to C2 that had to be looked at. Maybe it wasn't a good move, but it seems like that would have been a question in this position, Yas, a move like queen to C2. Now that attacks a pawn. The queen has penetrated into the heart of White's position and it threatens the e pawn.
Mr. Seirawan Right.
Mr. Ashley It guards some critical light squares.
Mr. Seirawan Right.
Mr. Ashley Looks like a decent move. May not be a move that was that great. Maybe Kasparov can actually handle a move like this. But in Kasparov's mind he's really thinking like his archrival Karpov, isn't he? He's saying we're going to prevent you from having any chances at all. I don't care if I don't get my thing going. You're not getting anything at all. Mine will last forever. He has this long-term advantage of his a pawn queening. But your little tricks, cheapo tricks, queen coming into position that I got to start thinking about, all these variations, I don't know what might happen, I'm not going to let any of that happen while I'm trying to execute my plan. So queen A4 has prevented that. That's a double-edged sword times. Sometimes you prevent somebody from doing something and -- and the computer has responded with bishop to C3, offering an exchange of pieces. 24. Rxb8 Qxb8. .
Mr. Seirawan Just to complete the thought, Maurice, I wouldn't be afraid of this type of counterplay because I think the a pawn would still be the key in the position. But I think that I'm a little bit worried about what Garry just did because I see that this -- oops. I see that he still has this idea of going after Black's king. The intent of the move queen A4 was to pull Black's king -- Black's queen, pardon me, away from the light squares. So he had queen A4, DEEP BLUE responded bishop C3. There was rook takes B8, forcing DEEP BLUE to play the move queen takes B8, which he will play shortly. And then I think that what Garry wants to do is again go after these light squares. He might come up with a move queen D7, which would run into queen C8 or he might try the move queen C2. And again he's kind of turning his attention towards DEEP BLUE's king. He likes -- you got to say this about Garry. He loves to go after his opponent's king.
Mr. Ashley But you know I'm actually very impressed by this because he's playing on both sides. He wants his a pawn to become a queen, but first he wants to tie Black down to the right side of the board and if all of Black's pieces are tied down to the right side where Garry is going after his king, he'll just do the old switcharoo on him and go back and push the a pawn. And you know you can't serve 2 masters. You can't deal with threats on both sides of the board. So this is seeming to me like great positional play by Kasparov, restraining any possibilities for the other side, keeping in mind the dual possibilities for himself and thinking in very long term strategic way, not concrete calculation on his part, just his great strategic thinking. I think this position is great for White.
Mr. Seirawan The operator has captured back the rook and we're now looking at the move queen C2, again with the overall aim of dominating these light squares, especially the squares from B1 to H7. So we're anticipating the move queen C2. And again I noted yesterday when grandmasters play in international competition, what they do is they love to put their pieces directly squarely in the middle of the square. I mean they really like it to be precise. And you just saw Dr. Feng from the IBM team just nonchalantly go queen takes B8, and Garry immediately reaching out and adjusting the queen. Okay. Here we go. 25. Be4 Qc7. .
Mr. Ashley Kasparov has instead in playing queen C2 -- let's recap those moves. The last couple of moves after Kasparov played queen A4 was bishop to C3, rook takes B8, queen takes B8. And instead of queen to C2 Kasparov has played this very interesting move, bishop to E4, actually quite thematic, getting that bind on the right side of the board. What this means is Black has an extra pawn on the right, 5 against 4, but this doubles. Those pawns want to come down the board. He wants to play on that side by pushing those pawns, play a move like F5 E4, bring his queen and rook behind everything and try to push his pawns down the board. With bishop to E4 Kasparov has made that an impossibility. He will now bring his queen to C2 or bring his queen via D7 to G4 or F5 and then he will have complete domination of the light squares and potentially a mating attack on the light squares. So now DEEP BLUE is going to have to figure out how to defend against this attack, especially considering that it has no light squared bishop to deal with the menace on E4. I think Fritz here is giving white a point 66 advantage. That is super significant. I think it's even more. I think that White has a really large advantage, maybe will win a pawn in a few moves and, if not, will show much more than just point 66. We'll see something quite dominating in a few moves.
Mr. Ashley You look at this black bishop. Who can it point to? What's its target? It's just sitting -- you know, it can roam free as much as it wants, but there's nothing to hit. It can't target anything. It's like I want to latch on to something, you know, but there's nothing to latch on to. Those pawns on the king side are perfectly safe and he doesn't even have access to them. On the other hand, look at white. White has target. White has the H7 pawn and he also has a king sitting over there if his queen ever joins in the battle. So for white it's serious. This advantage is serious. Kasparov knows it. What's also beautiful, I think, Yasser, this is easy for Garry to play. No losing chance, this kind of position. You just kind of coast along and see if you can find the best moves and Garry is -- you know, he likes it. You can tell.
Mr. Seirawan I saw that nice little smile.
Mr. Ashley He likes it. He knows he's not going to lose today, at least.
Mr. Ashley Unless he's dumb. But he's feeling like, how do I get this thing.
Mr. Seirawan I've been just looking at Kasparov's mannerisms and the eyes have been kind of sparkling a bit. I'm seeing the move queen C2, anticipating the move queen C2 attacking this H7 pawn. And what would be ideal is the queen and bishop make a lovely battery, but what they need to do is reverse positions. It's like the bishop's leading the queen as opposed to the queen leading the bishop. So after queen C2, we get a nice little tempo, a move like H6 and then we come up here with this nice little, a small move, queen D3 anticipating being able to swing over here to F3 and maybe to F5, reversing the battery.
Mr. Ashley I like that plan.
Mr. Seirawan So something like this is just ideal for white. He's keeping -- Gary's doing a very good job. He's keeping his a pawn. He's keeping the trump of his past pawn and he's getting some light squares on the king side. So queen C2 looks just like an ideal move that is going to fit in nicely with his strategy in the back --
Mr. Ashley Black almost has no counter play.
Yes?
(Audience commentary.)
Mr. Seirawan An excellent suggestion. The gentleman has suggested, what about since we know the intent is to get a battery the other way with the queen leading the charge, what about bishop B1 intending to bring the queen to C2 and coming down to H7 mate. I think that's an excellent move. Bravo. I really do. That was a good move.
(Audience commentary.)
Mr. Seirawan Also protects A2.
Mr. Ashley What about bishop B2, not to be a party pooper, or bishop A5 in that position? Maybe bishop A5 would be more reasonable.
Mr. Seirawan Maybe bishop A5. Well, again, this is -- all white really needs is to get on this lovely diagonal so he could whip his queen back anyway.
Mr. Ashley Queen A3 is a good -- Fritz is suggesting the move queen A3 which is very nice because not only does it have the queen D3 threat, it also has the threat bishop takes A7 check. And queen takes rook on F8. So that's a little sneaky idea that's possible in the position.
Hans Berliner?
(Audience commentary.)
Mr. Ashley Another interesting idea to be sure. We can say that Hans Berliner is not only a computer specialist but he played in several U.S. championships back in the good old days of Bobby Fischer. He was a regular on the U.S. championship circuit, so a very strong chess if player in his own right. All those plans, it's great for white. White can just figure out how to kill black. It's always nice when you're the 1 with all the plans and what can black do? Where is planning supposed to find the counterplay in this position?
Mr. Seirawan Exactly. DEEP BLUE is going to be reacting it. He's not -- he doesn't have the ability to make threats. And there again that's 1 of the threats. A lot of people say bishops of opposite colors draw. 1 of the strengths of having the bishops of opposite color is when you're on the attack you can put a lot of things together. And thank you, Hans. What Hans Berliner has suggested in a wonderful suggestion is just playing rook B1, the idea being this rook on F1 hasn't been a partygoer and he wants to get himself involved. And the idea is to bring this rook down to B7, attacking the queen, or putting pressure on that F7 pawn. So the idea of rook B1 coming down, playing on the light squares and letting the queen decide what she's going to do is perfect. This would almost force the move rook B8 and as Hans points out, after a virtual force, we got exactly what Garry has been aiming for. Remember when we saw the move queen C8 offering the A7 pawn as a nibble? Well, now that move queen C8 is no longer available because there's no rook to protect black's queen. So the queen would go to D7, slip back to F5.
Mr. Ashley Garry is smiling. Garry is actually smiling at the board. He's smiling. He's loving it.
Mr. Seirawan Well, it's nice to be on the sunny side of the torture, isn't it?
Mr. Ashley He has so many possibilities here. And you know, I didn't even notice that another target that can't move even better target, Yas, is that F7 pawn.
Mr. Seirawan That's a juicy 1.
Mr. Ashley That F7 pawn is nice. It's on a light square. It can be targeted. The h pawn maybe it will move up the board, but the F7 pawn is stuck. And that's another target for white. So white has all -- this is just beautiful position by Kasparov. If he pulls this move off, this is going to be 1 of his classic games. He's played it so purely, just getting his advantage.
Mr. Seirawan I'll just point out something else too. There's some very, very nice end games that can occur for White. If you imagine in this position you have a move like king F8 and then you go to a move like bishop takes A7 and you play a move like queen E8 -- and again I agree that White doesn't have to do this, but an exchange of queens produces a marvelous ending. What you have is White has 2 passed pawns. He's got this h pawn and he's got this a pawn. And all Black has is 1 blockader and 1 defender over here. So this would be a very, very simple win for Kasparov to realize. And I think that Hans is absolutely right and that we're going to see this move rook B1. That's just -- that's a devastatingly good move. And I'm watching Fritz in that analysis, that Fritz move -- that Fritz analysis I think was getting up to like 1 point 0.
Mr. Ashley Yeah. Fritz likes this for White. It's just that pure domination on the light squares. It's absolutely pure. Rook B1 looks devastating. Rook B1 is very, very strong move.
Mr. Seirawan And you're seeing Fritz just jump up beyond.
Mr. Ashley So he has a very strong position.
(Audience commentary.)
Mr. Ashley If Kasparov wins this game I don't think you'll have to coax him to speak.
(Applause.)
Mr. Seirawan Yesterday Garry was obviously in a sour move. It's a horrible thing to lose a game and certainly a match that Garry fully expected himself to win. So the answer to your question, sir, is that in case Garry does win, he will run out here.
Mr. Ashley I think we'll have to coax him to stay in his room, to hold him down, because he'll be out here.
Mr. Seirawan Garry, could you tell us about your win today? Absolutely. The win today was fine. Yesterday was an aberration.
Mr. Ashley But so far Kasparov is playing beautiful chess, beautiful chess. You know, you might fault a single move, but when you look at the strategy as it comes together as a whole, it's been absolutely wonderful and a completely marvelous exercise by Kasparov, positional bind. The assessment, the long term assessment of the differing pieces, I have a bishop, you have a bishop, but your bishop is useless. That's what Kasparov has on this chessboard right now, a very strong advantage. And I see him pressing it home. Again the same problem. Remember that position where Kasparov had like 3 or 4 moves? And he really and is a.
Sat and thought for a.
COUNCILMAN LONGSTRETH Time and time passed and time passed? He's in that setting again. And we have to see what he decides on. I'd like to thank grandmaster Yasser Seirawan who I hope will be back in about 20 minutes.
(Applause.)
Mr. Ashley And introduce Frederick Freidel, who is the programmer for Fritz 4. Welcome.
(Applause.).
Mr. Ashley An interesting question for me, Frederick, is is this precisely the kind of position computers play badly, I mean the way Kasparov changed the whole parameters where he has a light square bishop, the computer has a dark square bishop, but they're clearly not of equal value?
MR. FREIDEL I'm not sure. I thought that he was playing the wrong kind of position, which means that he now understands computers better than I do. Because I thought this was -- this is not good. And I think you guys also felt that he was getting into a position where Black had exactly what he wanted. But he seems to be right. He seems to be winning.
Mr. Ashley But why would you think that this was a good position for Black? Because we did like White for most of it because of what was starting to happen.
MR. FREIDEL Yes.
Mr. Ashley Was it because of the openness of the position?
MR. FREIDEL Yeah, of course. Tactics, pawns that can go through. Somebody said that it's very difficult for computers to calculate pawn runs and queenings of --
Mr. Ashley I'd be the guilty party.
MR. FREIDEL I'm sorry, but it's not true with this computer because they have enormous hash tables and they go down 15, 16 ply, 20 ply and find queens all the time. And I don't like it when you have 2 pawns in the middle and these guys -- you know, we played Garry in London recently and -- with Fritz. And Garry pushed 2 pawns down the middle and Fritz gave Garry 2 passed pawns in the middle and had to defend against this. And all the annotaters were saying this is -- you were there weren't you?
Mr. Ashley Yeah, I was.
MR. FREIDEL You saw the game. This is terrible, terrible.
Mr. Ashley It looked like Kasparov would find winning possibilities.
MR. FREIDEL What is the computer doing? The computer is drunk. It's completely out of its mind.
Mr. Ashley Had a late knight.
MR. FREIDEL Yes, exactly. What actually happened, we looked very carefully later was, Fritz looked at these 2 pawns and said he cannot do it. He cannot queen these pawns. I can always defend against it. How about me trying to queen my h pawn? So with a little h pawn it ran down the side and Garry was with all his pieces there until Fritz suddenly told him, no, you can't queen them. And he almost lost that game.
Mr. Ashley Garry had to come up with a fantastic tactical possibility just to save the game when we thought it was over, he was winning easily, but the computer came up with these miraculous moves.
MR. FREIDEL And that is the danger when Garry says these 2 pawns are going to win or these 2 pawns are helpless, hopeless, they can't win against me, and the computer uses its hash tables like Fritz was doing, to find out he cannot queen the pawns or I can queen the pawn and suddenly it's over and they can calculate. That is dangerous. So I didn't like these pawns, the open position and so on, but now it's good.
Mr. Ashley Now you see the imbalance of the bishops seem to be a serious flaw in the computer's thinking.
MR. FREIDEL And he understood this. He understands exactly what you guys were saying, the dark squares --
(Audience commentary.) 26. Qa6 Kg7. .
Mr. Ashley Instead of rook B1 he has instead, Kasparov has instead, played a tempo gaining move, queen to A6, which attacks the pawn on F6 and maybe prepares the move queen to D3 to attack the h pawn also. And then he can play queen to F3 and his queen is on the other side. So Kasparov has found a way to sneak his queen around to the other side by force. His queen will get, it seems, to the other side of the board by force. It may not have been the best move, but then again it looks like a good move too. Rook B1 can still be a possibility in position. The pawn on F6 needs defense. It seems to me that the only move is king to G7 and after queen to D3 the h pawn will be hanging and if Black -- Black defense with the rook, that would be terrible. Let's take a look at that variation, king G7, queen D3. And I think here a horrible move would be something like rook to H8 defending the move because then Kasparov can turn his attention back to the other side playing rook B1 and the rook is tied down. So here it would be necessary to play
Mr. Ashley In that position, after queen D3 I think it was necessary to play H6. But then queen F3, bringing the queen over, and then that would be a very strong position.
MR. FREIDEL Can you give me just 1 second with the program? I'll tidy up the game.
Mr. Ashley Okay. So Kasparov is not interested in just getting a little advantage. He's talking about total domination, board domination. Hold everything down. Hold him down on the right, get going on the left. You're totally tiled down and I'll just beat you up. Good strategy. And I think that Kasparov is showing classic chess here. He's just keeping every little bit of his advantage. And this position, to me, this is -- to me I'd just say this is winning. You know, I'd just say this is winning. The programmer has played -- DEEP BLUE has played king to G7 and hopping his queen back to D3 with alacrity was Kasparov. 27. Qd3 Rb8. . Queen to D3, attacking the pawn and now big question for DEEP BLUE what, to do with the H7 pawn under attack. Will he defend it? But then the next question will be what to do about the move queen to F3. Queen to F3, and finally the queen with its fascinating maneuver starting on A4, going to A6, going to D3, going then to F3 and then over to that other side of the board. Now, remember what I said about that great Karpov-Kasparov game where Karpov really showed -- I think it was the second -- I think it was the first or the second world championship match in Tarrish. You look it up. It was one of the games Karpov won with the White pieces and he literally made something like 27 consecutive moves with all his pieces landing on light squares, every single 1 of them. Well, look at Kasparov's last few moves. He's played the move bishop to E4. Of course his bishop can only travel on light squares, but... He played his queen from A4 to A6, a light square. He moved his queen back to A3, another light square. Queen to D3 attacking a light square A7. He now plans to play the move queen to F3 another light square where he can go to G4 or H5, another light square and 1 of the strong moves we're talking about in this position for him is rook from F1 to B1, another light square. And if the rook goes to B1 where does it plan to go? B7, another light square r.
MR. FREIDEL What about this?
Mr. Ashley Everything about this position, everything about this position is totally concentrated on the light squares. And Kasparov is playing a model of chess strategy here.
MR. FREIDEL Maurice, what happens --
Mr. Ashley The question being asked is what happens if the move rook to B8, a very interesting move, threatening to penetrate along the b line is the rook and ignoring the capture of the pawn at H7. Wow, that seems to me like interesting strategy. Clearly White can either take the pawn at H7, but that might cause rook B2 at some counterplay or even queen F3 would be interesting.
MR. FREIDEL That's what Fritz wants to play.
Mr. Ashley Well, queen F3 is not a bad move. And look at this. DEEP BLUE has ignored the pawn on H7. He's given that pawn away. Sacrificing now for the defense and played the aggressive rook B8, looking for a counterattack. Now the king will never be mated. That's very important to know that the Black king will never be checkmated because it can always run to the center of the board. It can always go to hide behind its pawns by going to F8 and then up to E7. So that was a very interesting variation by the computer, but I still think queen F3 is a great move. Kasparov need not use that and go all the way to H7. I think he has a very strong position and would be very aggressive. Is this typical, it Fred, let me understand, is this typical that the computer will sacrifice when it recognizes certain positional binds it might get into? I mean H6 seems almost similar, and move the rook away. Instead it's gone immediately on the counterattack. 28. Bxh7 Rb2. .
MR. FREIDEL I can't calculate whether it just sees that that's tactically covered, you know. I don't know. Fritz saw rook B8 very, very quickly. We can just check in this position, it takes 1 second to do something. When we play this you can see within 1 second it says no, no, no, no captures nothing, I'm going to B8. So I'm not good enough to be able to calculate why Fritz said there's no point. Fritz has found out there's no point in trying to defend that pawn.
Mr. Ashley Amazing. So now this is a big question. We talked about time. Kasparov has now fallen into the fourth quarter, if you will. He had 2 hours on his clock. He's used over an hour and a half, with 27 minutes left. Now, in this situation against a human I wouldn't call this critical. And Kasparov himself doesn't consider it even close to critical. He just ripped off the h pawn on the last move, ignoring the penetration. And rook to B2 by DEEP BLUE and it is attacking the a pawn, which was the plus in Kasparov's position. Kasparov not thinking for too long and giving up, it seems, his main trump on the left side of the board. Maybe he will find a way to retain this pawn. But certainly a bit of dangerous penetration by that rook. And Kasparov shaking his head a little bit. I wouldn't say he should be so concerned, but that last move bishop H7, I mean it looked unnecessary. It doesn't look bad. Clearly Kasparov is not losing at all. No question about it. He still has a wonderful bind on the right side of the board. We shouldn't be too concerned for him because he's not going to lose the game. 29. Be4 Rxa2. 30. h4 Qc8. . And he's dropped his bishop back to E4 and he sacrificed his a pawn. That was the trump in his position. Well, that's fantastic evaluation if it's good. If it's bad, it's garbage. And he's shaking his head. Kasparov is now giving up his a pawn for the sake of this other position where he's attacking on the right side of the board. Boy, this is a very surprising decision by the world champion, not having his trump of the a pawn and I would think that DEEP BLUE would be interested in ripping the pawn off immediately, but it doesn't -- it still has to worry about the bind that Kasparov has on other side. Remember, Kasparov has a passed pawn now on the other side. Seirawan had mentioned this before. The h pawn is past. And with that past h pawn there is some possibility against the Black king.
Yes, a question in the back.
(Audience commentary.)
Mr. Ashley Well, the truth is that there are pieces that can potentially block it. A bishop -- the question is, there's no way of blocking the h pawn. But yes, there is. If bishop from C3 goes to D2, it could then go to H6 and it could be very difficult to get rid of the bishop in that square. But you are right, you are right, that h pawn is very powerful. And clearly Kasparov has pluses, but it's a very unique situation. Usually, usually, and I'm sure Yasser will agree with me, when pawns are restricted to 1 half of the board, the defense finds it easier. It's harder when, you know, it's like tug of war. You're pulling me over here and I'm getting pulled over here. I can't be in both places at the same time. But if you force me to just defend 1 little space, then I can bring all the troops over and we'll just deal with this 1 little space. That a pawn was the diversion. If Kasparov now attacks on the only side he can attack on where there are targets, then DEEP BLUE potentially and, hopefully for it, will be able to bring pieces to defend that side. If that happens the game is a draw, no chance to win. Even with the h pawn. I would say bring a queen and a rook over to the right side of the board for Black, Black's okay. But that's far from happening in this position, quite far from happening.
MR. FREIDEL What I don't understand is that Fritz, after rook takes H2, says it's point 7 pawns up.
Mr. Ashley I think that the positional strength -- again what we said, what Kasparov did is either genius or stupidity. I mean there's really no in between because he's giving away -- and I think you would agree with this, Yas, he's giving away a huge trump, the a pawn. And now he's giving it away.
Mr. Seirawan Exactly. Well, we saw this yesterday too, where Garry really enjoys playing for that knockout on the king. And like you said he did give away this very important trump. And I'm a bit disappointed because he didn't choose Hans Berliner's idea there of rook B1. And we'll probably see it now, the move rook B1 trying to come down to B7. But in this case DEEP BLUE should force the trade of rooks with -- if we can just take a look at here, Fred, rook B1, rook A1, for example, rook takes A1, bishop takes A1, queen F3. And now we have a situation where what you're saying, all the pawns are on 1 side of the board and that makes it -- the winning chances a lot more problematic.
Mr. Ashley Kasparov has instead begun to advance his other trump. Now he has trumps on that 1 side. We talked about the light square weakness. That still is in position. The other trump is the past h pawn. But the window is very small. The window of that -- those trumps are very small because they all exist on 1 side of the board. Basically what we could say is that the A, B and C files if I may be this dramatic, the A, B and C files are almost -- we could almost cut them off the board, almost just like slice the chessboard in half -- well, not exactly -- now, for just those 3 files and take those 3 files off the board. They're not important to this game, not so important. The critical files begin from the d file over, the E, F, G and H. That's where the action is going to take place on that side of the board. All those pieces you see Black's -- the A, B and C files, you see how they're over there hanging out? They don't want to be over there. There's nothing over there. There's nothing for them to deal with. They want to come back to the party. And the party is light squares on the right side and the h pawn that is thinking it's going to be a queen some day. So that's where they're going to have to get back to. Now, Kasparov has very hopefully, very hopefully for him, very delicately evaluated that this is great. I got a past h pawn. Your king is wide open. Okay. Maybe your king will go to E7, but at least it will be stepping out the way and the h pawn will be running down the board. Maybe he's already evaluated all this and realizes he has a tremendous advantage. And if that's the case, he's really played his level best as he is world champion.
Yes?
(Audience commentary.)
Mr. Ashley We have a member of the audience who thinks that Black can resign and who thinks the game will be over very soon. Quite possible. We will see how the game goes.
Mr. Seirawan 1 of the things that I don't like about Garry's play is, if you realize, just as you've spoken that it becomes play on 1 goal. That goal is on the king side. That means to say all the play takes place on that goal. I would say that what Black should try to do is just what White should try to do, and that is bring his queen to the king side. Therefore a move like queen C8 taking some of the light squares, if you could put that on the board, Frederick, I think that DEEP BLUE in this position should play the light -- queen C8.
MR. FREIDEL That's what Fritz wants to play.
Mr. Seirawan And then after a move like queen F3, I'm thinking of the move queen H8.
(Audience commentary.)
Mr. Ashley Yeah, but I can always exchange off rooks, rook A1. All exchange off rooks. 31. Qf3 Ra1. .
Mr. Seirawan And we've seen again queen C8. And if Black is able to play a move like queen H6, he will put up a very, very firm blockade. It will be very difficult for Garry to win the position. I still think that Garry has a very big advantage. We've seen queen F3 now.
Mr. Ashley And the game has progressed as such Queen F3 by Kasparov. And I must say, Kasparov -- the points we're making, Kasparov is very aware of. He's very aware of. He knows, look, I'm playing on 1 side, I am, but I strongly believe that it's good, that it's right, that it is the correct strategy and that I have enough to win. No question there's some thought in the back of his mind saying, yeah, I strongly believe it. I know, though, that a lot of times these positions are very defensible because they simply are. And he knows it better than we do. So whether or not he's going to be able to generate enough action to win this position is a big question. And Kasparov has taken a huge risk giving up his a pawn to try to win this position.
Mr. Seirawan I was just going to say 1 of the things that -- because every time White tries the plan of rook B1, he faces rook A1, there's 1 other possibility, of getting the rook on F1 and the game, and that is to play the move F2 F4. So if we'd just see 1 variation that is attractive at the moment, a move like queen H8 for DEEP BLUE, should DEEP BLUE play queen H8, then we have a move like queen G4 check. Key decision. What Black's going to do here. There's king F8 and king H6. Let's assume king F8. And now there's the possibility of F2 F4. And this is what I was trying to say, is that White might be able to bring his rook into the game by trying to open up the F file with the followup move F4 takes E5.
Mr. Ashley Pretty unorthodox of bringing the rook into the game.
Mr. Seirawan Okay. This is what DEEP has --
Mr. Ashley Rook A1. Now it's getting harder and harder. You know, you have to be very concerned. And look at Garry Kasparov shaking his head. He knows that he wanted that rook. He really wanted that rook. If he had his a pawn, exchange those rooks, you're dead.
Mr. Seirawan That's right.
Mr. Ashley Just dead because he would attack on the right side. He would attack the king. The queen would run over to defend and he would just start pushing that a pawn home. But his a pawn no longer exists. So he does not want to exchange off rooks. He really doesn't want to exchange off rooks. A very interesting suggestion made by an audience member is the move bishop to B1.
Mr. Seirawan We can take a look at that. I would support that move because again I feel that if you are just going to get into a queen and bishop versus queen and bishop position, it gets problematic and perhaps impossible to score a win. And therefore the move bishop E1, just preserving rooks, would be a good solution.
Mr. Ashley Even here, though, that's problematic. You know you're tying down a couple of pieces.
Mr. Seirawan I know. 32. Rxa1 Bxa1. .
Mr. Ashley Maybe it will still work out. It's not a bad-looking move. But Kasparov has instead ripped off the rook immediately. Rook takes rook on A1.
Mr. Seirawan I think Garry has spotted a very forcing line. And I've been kind of staring a bit at it myself. The variation goes Bishop takes A1, the natural recapture, queen to H5, intending queen to H7, check. Now a move like queen H8, just to make the defense, queen G4, check. And again this is kind of the moment of truth. If you play king H6, White can adjust its target a bit and go after the F7 pawn with a move like bishop to D5.
This is the forcing line I have in mind. It's a little bit long winded. Queen D7, king G6, bishop takes F7 check -- bishop takes F7 check, queen takes F7, H5, check, king G7, H6, check. But it doesn't work.
Mr. Ashley King back to G8, and it doesn't quite work.
Mr. Seirawan Okay. Well, there you have it. See how close I was?
Mr. Ashley Kasparov probably had that in mind.
Mr. Seirawan That's right.
Mr. Ashley And now he's in trouble because he didn't notice that possibility the way you did.
Mr. Seirawan The other problem -- there's another problem. After queen H5, I think a good move would be the move queen C1 check instead of the move queen H8, queen C1 check. And again the intention is that to bring the Black queen to the king side, king G2. And again we have this move queen H6, just getting the queen into shape here. This is going to be a very tough 1 for White to win. He still stands very nicely because this H4 pawn is in big trouble. Queen G4 check.
Mr. Ashley But you know, you should remember that. When defending, a good idea is to get rid of everything on 1 side and then focus your efforts on defending on the other side of the board. And here that has happened for DEEP BLUE. It was in a very difficult position. It may very well still be in a very difficult position. But it's putting up maximum defense, maximum defense and resistance. This kind of defense is a kind of plan I think Gata Kamsky, who is America's number 1 player at the moment, is famous for. I've seen him do it not just in slow games, but even in speed chess games. His instincts, even when at worse, are just uncanny. He starts maneuvering on the 1 side where he's worse, exchanging off the pawns and quickly going to the other side and hunkering down and defending. And this is what you're seeing here. That Black bishop is still useless. It's bad, really bad.
Mr. Seirawan We have a new function on our middle board. Frederick is about to wow us and explain to us exactly what that graph means.
MR. FREIDEL While you guys were analyzing, Fritz was storing the evaluation. And we can easily make it display that as a bar graph so you can see where the game -- how the game progressed, that White was a little better according to Fritz. There was a little glitch there. Fritz thought at least in the analysis at this position that Black was a little better here, went back again, swung in favor of White. And now we are here in favor of White. In fact, we're here. Another little item -- I've been watching this -- Fritz makes commentary and annotates itself. We had a variation here. Now, where was the variation? This was the variation, I think.
Mr. Seirawan Yes.
MR. FREIDEL And in this position Fritz brought up the commentary, you can see on the left-hand side, not many choices are there, DEEP, for the king. The reason is that somebody typed in the name of the owner as DEEP BLUE. So Fritz will address the opponent as MR. BLUE or DEEP.
Mr. Ashley So Fritz talks to you, huh?
MR. FREIDEL Yep. So where are we? 33. Qh5 Qh8. .
Mr. Seirawan We've seen the move queen H5. Kasparov's last move was as we said, that he would play queen H5. Now if Garry was left to his own devices, what he intends to do is go queen H7 check, forcing Black's king back to the F8 square. And then you notice that the h pawn can trundle down the board unimpeded, and then he would follow up with H4, H5, H6 and so on. So that was why we thought the move queen C1 check was necessary in order to be able to blockade that pawn on the H6 square after queen C1 check -- do you want to just put that on for a moment there, Fred, queen C1 check, king G2, queen comes back to H6. It makes a very effective blockade. As Maurice was pointing out, Black's bishop on A1 still has nothing to say or do in the game, but if Black manages to swing his bishop from A1 back to the H6 square, then I don't believe that Garry could force the issue. So this is going to be very critical now.
Mr. Ashley The next few tempi are going to be enormous if that bishop gets over there. 1 thing we should point out is the exchange of queens is a draw, just a draw. No winning chances.
Mr. Seirawan That's all.
Mr. Ashley Forget it. White cannot exchange queens, period. The h pawn will never go home if that happens.
Mr. Seirawan Right.
Mr. Ashley Because the dark squared bishop will just simply prevent it from advancing and he won't be able to crash on that color. So White will avoid the exchange of queens at all cost. Barring the win of a piece, of course. He will simply not exchange queens. So the advantage for Kasparov is clear, but may only be theoretical in the position we're in. And it's going to be very difficult. It's going to demand the utmost technique from Kasparov to win this game and it may not even be possible. So a very tough, tough decision made by Kasparov to play like this. And I -- you know, I mean, I saw this game and it looked wonderful to me the way Kasparov was playing. Every move looked poetic.
Mr. Seirawan I agree with you, Maurice.
Mr. Ashley But this 1 without giving himself -- that A pawn was the ace in the hole. I liked that A pawn. I liked having it on the board because in a situation like this the A pawn would be a ridiculous distraction. If you could start running the A pawn down the board, and exchange queens it would be a ridiculous distraction. But without the A pawn on the board he's limited to just 1 side. And it's always hard to win a position like that. So we will see if Kasparov will be able to do what is necessary in this position, but DEEP BLUE has played the move queen to H8, which was a bit unexpected. I mean we saw this variation, but I thought that the other 1, with the queen on H6 was better, but apparently DEEP BLUE thinks differently. 34. Qg4+ Kf8. . And Kasparov, as we said, refusing to exchange queens. Queen G4 check played immediately. And now the big question of where does that king go. Does it go to H6 or to F8? Well, F8 takes it away from the passed H pawn. The H pawn is the big trump in the position. So we may see king H6. Those are the only 2 choices in this given situation. I've got to say though that bishop on A1 is really a stupid piece.
Mr. Seirawan It truly is. But on the other hand it takes only 3 moves.
Mr. Ashley And then it's a great piece.
Mr. Seirawan And we can sign score sheets and play in 2 day's time.
Mr. Ashley Even that's tricky. Because if the king goes to C8, Yas, then the bishop can never go to C1 because, if C8 checked, bishop takes queen on C1. So I know there are going to be some strange things. Kasparov's king is very safe compared to DEEP BLUE's king in this position.
Mr. Seirawan Exactly. But he's going to have to clarify his advantage in the next 2 or 3 moves because I really think that after bishop C3 bishop to D2 --
Mr. Ashley You'll never get anything.
Mr. Seirawan You're not going to break Black's position. So here is the key question| if we see the move king H6 or not. And I've just asked Frederick what choices does Fritz see in the moment.
MR. FREIDEL Well, king F8. But if I ask Fritz to show me the best 2 moves, you can see that the value is almost identical. It doesn't care too much.
Mr. Ashley And it has indeed played king to F8, leaving the H pawn alone, leaving the queen to handle that H pawn, that passed H pawn that White has in the position. So technically now it's a bit -- it's a bit unstable. I think this position for Black a little, a little -- but still a huge positional threat. That is bishop to C3, bishop to D2, and bishop to H6, after which White's advantage will have just dissipated. And I think Garry will be frustrated by that kind of result. So that's the threat now. Can White come up with anything to deal with this positional plan of bishop to C3 to D2 to H6, which would just simply equalize the position and yield him no winning chances whatsoever.
Mr. Seirawan That was why I tried to focus my attention on that wonderful line where I was sacrificing a bishop to the F7 pawn. Let's give it a shot. Let's try queen to D7 and see if Garry can break through with this move. The idea is to create -- oops. I was going to say, the idea of queen to D7 was to focus on the F7 pawn and see if we couldn't keep Black tied up with the defense.
Mr. Ashley I think this last move is just a professional move on Kasparov's part. If you look at the number of move we're on, we're on move 35. He has to get to move 40 to make the first time control and get an additional hour on his clock. He's now down to 18 minutes and 17 seconds on his clock. Not a critical time pressure. But often you'll find a grandmaster when playing in a tournament, when he realizes he's almost to move 40, he'll gain a couple of free moves if he can repeat the move once or twice. The third time, then the opponent can claim a draw. But once or twice, no problem. 35. Qc8+ Kg7 .
So here Kasparov is quite willing to play the free move queen C8, knowing that Black has absolutely no option but to play queen G7, and then bring his queen back to G4 with check, and he got a couple of moves for nothing. And then when the queen goes back to F8, he could play queen D7 and he's gained 2 free moves and he's not going to have to worry about time pressure. So that's what I think is motivating this last move queen C8 check. I don't see any other reason for him to play it at the moment.
Mr. Seirawan Exactly, I agree with you, Maurice. 36. Qg4+ Kf8 .
Mr. Ashley And quickly he jumped right back, queen G4 check. Very professional. You see this in thousand of grandmaster games, when a situation like this arises near time control, repeat the moves a little bit, gain some time on the clock, so you don't have to worry about it, get to move 40, and sit and think about what you're going to do. So very professional by Kasparov.
Mr. Seirawan I don't think it will be a draw.
MR. FREIDEL The interesting thing here is that DEEP BLUE is recalculating the position here and trying to figure out whether this time it's better to move to H7 -- to H6 or to F8.
Mr. Ashley This is an interesting issue. If DEEP BLUE were to get to this position again for that third time I believe it was mentioned that the good programs, I think Hans Berliner said, if they see a move that's slightly interior or basically the same as the third -- the 3, 4 repetition, it will opt for the other move. So maybe a glitch in the program gaps could try to exploit, repeat and go to H6.
MR. FREIDEL It will go for the draw. Now it's A pawn down and it will go for a draw. See, Fritz is already showing a draw because it says great, if White is going to repeat moves, I'm going to go for a draw. Thank heavens.
Mr. Seirawan A nice point here, when I play chess in international competitions and I sit there, boy, am I the optimist. I love to look at all the ways I can win. And they just seem to spring to mind. And I'm looking obviously from Garry's perspective because I'm really rooting for him in this match and I'm seeing after the move king F8 queen to D7, not to C8 as we saw on the last turn, queen to G7, then I have the wonderful and DEEP and brilliant mate in 1. Queen to D8 mate.
MR. FREIDEL And Fritz saw it.
Mr. Seirawan Fritz saw it. So I'm sure that Garry -- I'm sure he's mused dollars it over and thought to himself, boy, I could mate the computer in 2 moves. I wonder if it will allow me. Anyway, the idea of queen D7, and it's now on the middle board here, is to refocus the energy. We know we've got 1 trump, the H pawn, but that oftentimes isn't enough, as Nimzowitsch told us in his many works, is to win a game, you need 2 advantages. And in this particular case what White could try to do is to play -- to follow up the move queen D7 with the move bishop to D5 in order to focus the attention on the F7 pawn, and then we have 2 advantages. We're attacking the F7 pawn and we have our H pawn. 37. Bd5 Ke7 .
Mr. Ashley Kasparov has played a slightly different move. After king to F8, he's instead played bishop to D5.
Mr. Seirawan Immediately.
Mr. Ashley Yes. Bishop to D5, which harmonizes with your plan of putting pressure on the F7 pawn, and in fact has introduced a small threat of queen C8 check. But even that position looks like a draw.
Mr. Seirawan Yes. Even at the cost or at the opportunity to win the F7 pawn, I don't think that Garry is anxious to exchange queens. But this is basically what his only chance is, is if DEEP BLUE has enough time to reroute his bishop back to the H6 square, he'll draw the game. Therefore Garry has to hit and he's got to punch now. So the idea of redirecting his energy to that F7 pawn has got to be the right thing to do, redirecting his forces, pardon me, with the move bishop D5 followed by queen D7.
Mr. Ashley Let's take a look at this variation, bishop to C3, which looks logical. How does 1 follow up?
Mr. Seirawan Let's just go and do the direct thing, queen D7. Not check, but simply queen D7. Now, the idea being that queen G7 ends the game in 1. So virtually queen G7 and queen G8 are forced. Those are the 2 choices. Right?
MR. FREIDEL Yep.
Mr. Ashley Queen H5 is also possible. Queen H7, how does White play?
MR. FREIDEL We have a lot of -- okay.
Mr. Seirawan We're doing -- here in the commentary hall we're having a computer showing us the way to beat the other computer. This is a good 1. So Fritz decides on queen H7 being the best?
MR. FREIDEL Yeah.
Mr. Seirawan Okay. Now let's see. We can't -- it seems to me we'd like to throw out the move H5, but since you just take it off that wouldn't be very clever. We have to do something else. How can we combine our threats? The only thing I really see in the position is that White could bring his king up to the H3 square in order to be able to play king G2, king H3, G4 and H5. But then again this is what allows that bishop to -- Black would play bishop to D2 and bishop to H6, and I think Black would be satisfied.
Mr. Ashley This board is so small. I mean, he's limited. Remember I said about the A, B and C files not being important. After bishop to D2, after bishop to D2 --
(Laughter.)
Mr. Seirawan What happened?
Mr. Ashley After bishop to D2 and bishop to H6, the A, B and C file become irrelevant to the game, and literally you may never see another piece land on that square again, on 1 of those 3 files again. The board is very small. That means it's more easily defensible. But Kasparov is going to have to come up with an ingenious way of making progress here.
Mr. Seirawan I have another idea. Let's give queen --
Mr. Ashley Whoa. King E7.
MR. FREIDEL That's what Fritz was looking at.
Mr. Ashley King E7, yeah. But --
Mr. Seirawan That's why maybe queen D7 was more precise, to prevent this as a defensive move. What do you see?
The idea had been being able to play bishop D5 and queen D7. Kasparov chose to play bishop D5 on his first move. If we could just stay with this analysis for a moment. Queen D7 --
Mr. Ashley After queen H7, you mean?
Mr. Seirawan Yes. If we could see that once more what we were just looking at. Bishop D3, queen H7, queen H7. What Black might try to do is snuggle up a bit. Try the move queen D8 check. The king would have to go to G7, and then we play a move like queen to E7. And I'm just wondering if we can't squeeze Black like a game of bridge, for his hand. No, he would still have queen G6, I think, and we couldn't break the fortress.
Okay, sorry. Another attempt failed. Okay.
Back to the position at hand, king E7. It looks like a very good move because it keeps White's queen a little bit out of the game. White could now try to move queen F3 with playing queen A3 check, maybe bishop C4, and try to sneak in on B7.
Mr. Seirawan All right. Let's give that a shot. We're looking at the move now for Garry queen F3, bishop to C3, stopping the move queen 8 for check. And now Maurice had the idea of bishop back to C4, opening up the F3 diagonal so we could go queen to B7 check. This could work. More to come...stay tuned!
Mr. Ashley Boy.
Mr. Seirawan How does Fritz like this? It is giving quite a good advantage. Maybe bishop A2 is even stronger actually, so queen C-8 is not playable.
Mr. Ashley Then you might be able to combine and play on both sides of the board on this position.
Mr. Seirawan The idea of bishop A2, and again the intent is to be able to play B7 check and snap that little F7 pawn off the board.
Still it looks like Black can defend this position. The board is just so small. I mean, I keep getting back to that. It has become limited by the fact that there is nothing on the A, B and C lines, and so you don't target anything over there. Right now Kasparov would like it if you could say take your stupid A pawn back, put it on A7 and give me my A2 pawn back and now I just crush you, just an easy win.
(Audience commentary.)
Mr. Ashley Bishop to B1 Kasparov certainly interesting, to keep the A pawn -- well, too keep rooks on the board. That only kept rooks on the board. It still bothers me that it happened.
(Audience commentary.)
Mr. Ashley Right. He took the -- well, yeah, the H pawn. When he took the H pawn, I have to say that that comes as a major surprise to me, and the fact that Kasparov did it so quickly, it seemed to him that he Kasparov going to get this quick bind in the H pawn, and that Kasparov going to be enough, and he allowed his A pawn to go off the board. But it Kasparov such a major trump. It Kasparov an ace in the position to give it up. Now you see even a world champion is finding this position not trivial where he can make easy wins.
The thing that we had mentioned before, the opposite colored bishop, that means that White is powerful on one color and Black on the other. Kasparov is seeing this dark bishop making the maneuver. And bishop A6, he will never in his life get the bishop off of A6. That bishop will sit there forever. If he lets that happen, I think we will be seeing a draw. I can stop doing commentary up here. I can just leave. I don't think I would be allowed to, but I would be ready to. This position is so even'ish. 38. Bc6 Kf8. .
Mr. Seirawan Let's see if Gary can make the break that he needs. He is trying bishop to C6. He has changed plans. We were going to look at queen F3, but his intent is very clear. He is moving bishop C6 and attacks the -- well, he will set up what he has before.
If he can force Black to play the move queen D8, then there is an opportunity with that H pawn because after queen D8 we would just go H5 followed by H6 and queen G7, H7, H8
(Audience commentary.)
Mr. Ashley The question was has DEEP BLUE seen everything now since the position must be much less complicated because of all the reduction of material and the pawns limited to one side.
MR. FREIDEL Definitely not. Because there are more possibilities now in this game than there are elementary particles in the universe, atoms. You know, the number of atoms in the universe is just a tiny little number compared to the positions that derive from this.
Deep Blue is probably looking 15 my. It is not going very, very deep because there are queens on the board. Queens are a nuisance for a chess program. They move all over the board. When the queens are off, it goes three, four, five plys deeper, maybe even more. But with the queens, it is looking maybe 16 and a half moves a head.
Mr. Ashley That scares me.
Mr. Seirawan Deep Blue has not considered every single possibility from this point onward, but, on the other hand, it is still frightening in itself that it might be able to see as many -- 39. Bd5 Ke7. .
Mr. Ashley Not to interrupt you guys, but after bishop to C6, king F8, Kasparov has repeated moves again and played bishop to D5, and that is his 39th move. So he makes one more move. He is at time control and he can relax again. He can say, okay, I have another full hour on the clock, and I can try to figure out how I'm going to beat this thing, so very professionally done by Kasparov to make sure his time does not run out, but still we are wondering what the great strategy is. If he has a beautiful winning idea, why does he not go for it? He would probably be thinking in terms of that winning idea if he found it yet.
I have seen this happen on many occasions where you just repeat the moves to make sure you make move 40 and then you triple-check your winning idea and you go right for it.
Mr. Seirawan I know that you are working with Garry, Fred, but it had to be quite clear that Garry Kasparov was more than upset by yesterday's loss. What did he feel he had to do today to equalize the match? What kind of frame of mind was he in? And did he, in fact, achieve the position that he wanted to?
MR. FREIDEL I can't answer the second question, but, well, just come back fighting and it is a complex strategy. You have to make sure you don't lose but also gain winning chances. Let's talk about something else. This is a bit difficult.
Mr. Seirawan We don't want to get into any trade secrets.
Going back to the point here, if the computer realizes it is behind, you can almost predict --
MR. FREIDEL Absolutely. Fritz says king E7 draw. You can see the 0 point 00, king E7. It is not going to look at anything more. It closes its eyes. I'm going to repeat. If he plays something else, I start working on a new continuation.
Mr. Seirawan Let's look at something a little more enjoyable. After the move king 7 --
(Audience commentary.)
Mr. Ashley Why is DEEP BLUE thinking of a position that is already thought about is the question. It has seen this position before and has managed to calculate this exact position before. Why does it not just play this same one again?
MR. FREIDEL I can't seen DEEP BLUE's time, but if DEEP BLUE has a lot of time, then there are algorithms that say when we have a lot of time, just calculate again one ply deeper. See if there is anything else. Maybe there is a win somewhere and can dig this out. It cannot tell that there is no win here like these two can. They can say Black cannot win, don't look for anything. But DEEP BLUE has to consider it might find some spectacular move which wins. 40. Qf3 Bc3. 41. Bc4 Qc8 .
Mr. Ashley Or repeat, as it has done. Now a critical move 40 has presented itself to Kasparov, and he has played the move queen to F3 with the not so subtlety queen A3 check, winning a bishop for nothing. I think this is not a very good move because it forces Black to play bishop C3.
Mr. Seirawan But let's take a look at that because I wanted to work on an idea if I can. The idea was bishop C3, which we anticipate to be a forced response on the part of DEEP BLUE, then bishop A2, Maurice, which I think is quite a nice one. I want to threaten the move queen B7 check, and I'm just wondering what Black is going to do to counter that. I'm anticipating queen H7? Is that what --
MR. FREIDEL That's what this computer wants.
Mr. Seirawan So we will give queen H7 a look, and now, how I can finagle you here?
Mr. Ashley How about queen A8? I want to keep you there.
Mr. Seirawan Let's take a look, queen A8. And now we think that we have this nice little threat. What I'm going to try to do is I want to play the move queen A3 check, so that I force you not to be able to go over to the kingside. So, for example, if you play a move like bishop B2, which may be a stupid move, I don't know, and let's make a move like bishop D5 just to get our bishop back into the swing of things, and bishop H6, so we allow Black to set up his best setup, but his king mate, king A3 check.
And, now, if you play a move like king E8, I can get mine A4. Is that what Fritz says?
Mr. Ashley King D7, and basically what I have tried to do by this is to pull Black's king away from its sheltering pawns.
Mr. Ashley Maybe if you had put the bishop on C4 you could play bishop C5 check in this position. So a lot of possibilities if you keep the king from going back over to the other side.
Kasparov now leaving the board having comfortably making the time control with is 11 minutes to spare. He is going into the grind mode where he will try to win this game by great grandmaster technique, if it is indeed winnable at all.
I was about to say he is going to make the computer suffer.
(Laughter.)
Mr. Ashley If I had Black, I was saying I have to play another hundred moves against this guy, and it will be terrible. Bishop to C3 is the next move and it will seek to find the best defensive ideas. And both sides, even the operator, is getting up on move 40, leaving only DEEP BLUE on stage to -- the only one who is really working right now is that computer. That is something about computers. While we go and need a rest and stretch our backs, they just keep going and keep going. Kasparov now back at the chessboard. This is the basic problem, that the computer is not tired; Garry is.
Mr. Ashley Kasparov has already looked at queen D5.
Mr. Seirawan I was going to make a couple of comments. The first, you really have to say this. This is really an historic and marvelous event where we are seeing for the first time the absolute world's best player versus the world's best computer program in a world championship tile.
Mr. Ashley Let me just interrupt you to say that a new operator has been sent in, the relief pitcher, so to speak, Murray Campbell of the DEEP BLUE team. It is not really like a one-on-one thing here. That is not fair. They are doing that old wrestling tag team kind of thing. I am a little tired and now we send somebody with fresh fingers in and they are feeling good and maybe they know something that you don't.
Mr. Seirawan To continue the point, a very historical precedent-setting match, and really a lot of kudos are in order. We have to thank ACM, who for 25 years have been sponsoring chess events all over the United States, and I know that Monty Newborn, the chairman of ACM, has done a great job. Would you join with me in applause? So they really have our respect.
Also, even more troublesome is the fact of the matter is I don't want to play a computer in professional competition because for me, I like that pathos of nervousness. I like to have that energy from my opponent to see that he is under the attack or he likes his position and Gary has never shrank from this type of a challenge
(Applause.)
Mr. Seirawan Very good. 42. Qd5 Qe6. .
Going back now for a moment back to the position, because actually this was your original idea, Maurice, and I would like to support you on it. The tempo defense queen C-8 may not be all it's cracked up to be. We see queen C-8. There is the move queen to D5, defending the bishop and attacking the pawn on F7. There is the defensive maneuver queen E6, and now I have the move queen C-5 check, attacking your queen and king.
Mr. Ashley Now, what I tried to do in this is be a little bit tricky because if you notice, I pulled your queen away from that blockading A6 square, so after a move like queen D6, I will play the move queen A7 check. And again you have to defend with the move queen D7. I will play the move queen A3 check, preventing the queen from running backwards. Now I will try queen B3, presenting a new problem.
First of all, what I have done is drawn your queen away from the blockading A6 pawn, and now I have mixed your pieces up a bit, and it makes it a little awkward to defend. It enhances what you said. Maybe even queen A2 was possible in that position.
So that bishop D2 will never happen, you know, and a defense like queen B4 definitely won't happen, and that maneuver you just pulled off is excellent. That was beautiful, bring all the pieces back to the left side of the board, and the H pawn is now thinking I'm free.
Mr. Seirawan This type of thinking for yourself. This is parallel processing. Maurice and I --
Mr. Ashley We should get that 400,000. What's up with that? We could handle this process.
MR. FREIDEL I hate to tell you this, but every single move was instantaneous by Fritz. You can watch it. Sorry. I'm including queen A2, no, no, not A3; A2 at the end here and B3.
Mr. Seirawan That is one of the beauties that I do like against playing against computers, is that they do, in fact, anticipate our every move, but it doesn't prevent us from beating them. So I'm glad that Fritz could --
Mr. Ashley What's up with this? Murray Campbell is on the opponent. It's like the hot line at the White House or something. What's going on.
Mr. Seirawan We have some sort of a problem. You have heard C.J. where he said that he was going to be making sure with Con-Edison that the electricity did not go off, so they are having a problem. Frederick is going to go on his way to see what is the problem.
What the spectator just said is that when Murray Campbell, a new operator, came in, he was a bit taken aback and immediately he started typing away. I have no idea why.
Mr. Ashley It doesn't matter. They could always put this position on the board and have the computer play from this position if it is not a real technical problem on the board.
There was a comment that Fritz could draw this position.
Remember, we have BABY BLUE in back. They did have another backup system, so if this one were not to play, they could put the position in and they would play.
(Audience commentary.)
Mr. Ashley It may very well be. We have seen that many times before.
Mr. Seirawan Let me just bring the rest of the audience on up with what Hans Berliner said.
A lot of the public doesn't know that one of the advantages of the computer is it has a great calculating speed, but it can actually end up squashing the computer. That is to say, in this particular position, DEEP BLUE, as Hans has pointed out, may have calculated a win for Garry, and so it makes the assumption that since it has seen the win, so has its opponent. So since DEEP BLUE knows that by playing normal moves, normal looking moves, they will lead to a loss, it may try some absurd looking defense which is even worse, which accelerates the defeat.
Mr. Ashley It may not accelerate the defeat, but it is absurd to us because we know instantly that that is losing, but we don't know that the other variation is losing, so, for example, if it is preventing a mate, a force mate, we don't see it, right, we have no clue. We might go down to variation and mess it all up. But it goes oh, no, I have to stop force mate in 20 and here is a piece and we think, hey, thanks for the piece.
So yeah, that is a very profound point. And how do you deal with that? How do you prevent it from doing that?
(Audience commentary.)
Mr. Seirawan What Hans was just saying was that as the computer is thinking, his program, what he tries to do, is include a little solution as the computer is reiterating. The intent is to let the computer play one of the moves it previously considers, and even though it may also lose, at least it is not as obvious a loss.
(Audience commentary.)
Mr. Seirawan The question that the gentleman poses is in tournament regulation play, if a player touches a piece, it is a touch move rule. But we are not talking about a touch move as much as a touch thought. You had that thought. You have to play it.
(Audience commentary.)
Mr. Seirawan It is not the same. DEEP BLUE has not made a move at this moment. It is thinking.
Mr. Ashley If it puts the move up on the screen and it is telling the operator play, stupid. I see more than you do, then it sort of has made the move.
(Audience commentary.)
Mr. Seirawan We are talking about an in-built program that the programmer would put in well before the game even begins. We are not talking about the programmer on site. It is telling the computer don't give away a rook.
Mr. Ashley So the computer will check itself.
Mr. Seirawan I just wanted to say that we, Maurice and I, using Frederick's program, have tried to determine if Garry can improve the position, and I just wanted to point out that if Garry can achieve this position, Fritz has given Garry a much bigger advantage, one point 44 pawn, than he had when he went into it because a moment ago when they were making those repetitions, Fritz had given the position absolutely --
Mr. Ashley I thought the maneuver you just pulled off was absolutely poetic and maybe Garry's gut feeling still said to him, I have real chances here and I could not lose this type of position. We have to be impressed by Kasparov being able to do that.
In the meantime, they are working on the computer. The programmers are working on the computer. It seems to have some kind of difficulty. The arbiter is over at the computer screen. A few of the programmers. Team DEEP BLUE is still right by the computer, so they have not resolved the problem yet. Kasparov has yet to return to the stage, so apparently some issues regarding the computer. We don't know what that is, and I believe Fred Freidel went to check and will report to us shortly. More to come...stay tuned! (Audience commentary.)
Mr. Seirawan Just to go back for a moment, this is the maneuver, and in this position you are suggesting the move and that could be a good move. Before I answer that, I just wanted to say that throughout this whole thing, DEEP BLUE's chess clock has been ticking. It's very hard to get a read, but it has been ticking down for sometime.
Is this DEEP BLUE's 40th move? 41st.
Mr. Ashley Kasparov has returned to the board, and he is not understanding what is going on. It seems that he is not sure what is happening at the board. This is one flaw whenever a computer is involved.
Mr. Seirawan I just wanted to note -- how many of you were with us yesterday? For the sharp eyes amongst yourselves, you will notice they are playing on a different chessboard. Yesterday when Garry came in, he immediately had a problem with the chessboard and apparently the wood craft hadn't had the pieces perfectly, and I mean to within a thousandths of an inch lined up, and he thought that some of the squares kind of melded a bit, and he insisted that on today's game they use a different chessboard.
Mr. Ashley I think there is still a problem with the chessboard, and they are going to get yet another one. Chess players are particular about their equipment.
Mr. Seirawan It didn't have a made in Russia stamp on it.
Mr. Ashley Kasparov is writing something down on his score sheet. What I do not know, a lot of strange things going on now.
Mr. Seirawan We do see seven minutes on DEEP BLUE's clock by the way. DEEP BLUE has used an hour and 50 minutes to this moment.
Mr. Ashley I think they are going to get back to playing some chess moves. Let's get going.
(Audience commentary.)
Mr. Seirawan The question that the gentleman asked is if you took two computer programs, the very, very same software and you carried them against each other and you had them play random games, would they start repeating, and the answer to that question is --
(Audience commentary.)
Mr. Seirawan Right. That is exactly right. They have a self-learning. I know that most of the top programs, and I'm sure that this includes DEEP BLUE, it's not like a Pac Man where you will get into this loop where it will keep repeating and repeating, but rather it will store the game and say I have lost that already a dozen times. I'm not going to lose it a 13th.
Hans, can you buttress my point?
(Audience commentary.)
Mr. Seirawan What Hans says is that there is a lot of self-correction opportunities within the computer so that he believes that no, you are not going to see that constant repeat of the same thing.
Mr. Ashley I would like to ask Hans one question. This computer here, is it possible that it already has these end games in its data base, that the bishop with the H pawn, that bishop and H pawn is actually a draw and it would avoid all the variations thereby force -- like if Kasparov wins the F pawn, but it trades queens off, big deal. It is a draw, but would the computer say wait a minute. He can win my F pawn. I don't want to play that position.
(Audience commentary.)
Mr. Ashley Would it know though that kind of basic thing with all of these pawns on the board?
(Audience commentary.)
Mr. Ashley Wouldn't that be a problem, Yas. From Hans Berliner, this basic position where we know that bishop and H pawn -- bishop with the wrong colored H pawn, that is the H pawn, the final square for the H pawn is a dark square. It would promote on a dark square, White to H pawn. Since White has a light squared bishop, it means that white will not be able to assist that pawn in becoming a queen, so although it is past, it doesn't have a future if the queens are off the board. It doesn't have a future. So what we are looking for is a situation where if black can exchange queens off, even if you gave black that back F pawn, the one on F7, the position is still a draw. If it was just the bishop and the pawn itself, we know it is a draw, but to extrapolate, even with extra pawns on the board, it still can be a draw.
But the computer might be thinking, I don't want to give you a pawn, so I'm not going to calculate any variation where you win my F pawn. And that is a serious point because maybe it will play a much inferior move because of that possibility.
Hans, is that correct?
(Audience commentary.)
Mr. Seirawan You are raising some serious issues there.
MR. BERLINER The question is what caused it. We will find out later on, but, you know, there was clearly something going on that was upsetting people, and the question is what is there that was upsetting people, and I expect it was that it made some ridiculous move. If it would just stop, then they would be legitimate in restarting something. But if it made a move or something like that, then the resetting question is one that is really one that has to be looked at.
Mr. Seirawan We will try to ask somebody from the DEEP BLUE team what did take place.
Let me bring you up to date. I beg your pardon. DEEP BLUE did play the move we anticipated after bishop C4. It played the move queen C-8 attacking the bishop on C4 and we have predicted the move queen D5 which got us into that position, this nice little move. And then we arrived at the position that Fritz has given Kasparov a very large advantage.
Mr. Ashley And possibly Deep Blue has said I don't want to do this. Let's go for something else.
Mr. Seirawan The other gentleman walking down had a question regarding the stock market. Basically he was saying if you could do this with a game of chess that involves all of these huge calculations, you could do it for the stock market, and before I ask Hans to comment on that, we had a conversation very much like this last night where basically everything becomes a question of calculation and if it's calculable, it is programmable, so therefore...
(Audience commentary.)
Mr. Seirawan But the point that I wanted to add to that is today, of course, programs are used in the stock market, and in fact people have written their own programs for trading, so, yes, computers are going to be aiding us in our stock investments more and more.
Mr. Ashley It is difficult because the stock market is also a function of human reaction to that market, to certain trends happening. You know, somebody hears something happen with the treasury and they say hey, that sounds like something good and a bunch of us go out and lose our money. That is a very complex factor to have to build in also, just how human beings will respond. I think the variables in the stock market would be much worse than the variables in chess because the goals are not as clearly defined. In chess it is one simple goal. The stock market, forget it. There is a large, large goal within that definition.
Mr. Seirawan That was what Hans was saying. He kind of raised this issue. First of all, I don't know what happened, but the conjecture, let us suggest just for argument's sake, that the computer went down, a systems failure. It is absolutely permitted by the rules, then, prior to the match that the operators can reset the computer.
(Audience commentary.)
Mr. Seirawan If the computer just went down, just went down, and there was a wire or a default or something happened, fine, then they reset it, no problem. Nobody is going to get angry, but the issue there is let's say it saw it was losing and made a horrible move that the programmer said wait a minute. That is going to lose. We are not going to play that move.
(Audience commentary.)
Mr. Ashley I have been informed by Joel Benjamin, who works with the DEEP BLUE team that what in fact happened was the phone line went down and that interfered with the play, and they had to get that back together. We are waiting for the phone line to be reconnected. That answers the question.
Kasparov has played the big time move, queen D5.
Mr. Seirawan And Kasparov played move queen D5.
Mr. Ashley And DEEP BLUE has played queen to E6.
Mr. Seirawan Capturing that pawn on the check, DEEP BLUE immediately, immediately responded queen E6, blocking the threat and now the two players are right into our main line. We have been predicting the move queen C-5 check for sometime, forcing the move queen E6 because both queens and kings are under attack.
Now, I'm threatening the Black king as well as threatening to pick up that pawn on F7. We anticipate the move queen D7 blocking the check threat. Then we come up with the move queen E3 check and maybe we can do something different here because otherwise DEEP BLUE might lose right away. We had tried queen B6 in the past and then we came up with queen A2 and stopped and said Gary is doing marvelous things.
Let's go back to just this moment. If DEEP BLUE can deviate, it would be here. If he was to try to deviate with the move king D8, he could -- oops. Then we could -- wait. Then we can play queen down to F8 check in order to pick up this F7 pawn. If there was a block of the move with queen E8 then, we would come back with queen D6 check, queen D7 and then we would win our pawn. So everything would be nice. So this deviation -- 43. Qb5 Qd7. 44. Qc5+ Qd6.45. Qa7+ Qd7..
Mr. Ashley Kasparov has played a bit more interesting move. Queen C-5 looked very obvious, but the world champion has opted for queen to B5 creating only one threat of attacking the Black queen. I am somewhat surprised because we had felt that the queen on A2 would be better than the queen on B3, but it looks like Kasparov plans on putting the queen on B3 instead of the Black queen moves out of the way. But this is a very interesting situation, a very interesting situation because queen B7 could be an attractive alternative. Let's take a look at that move. Queen D7 for Black and White could just win a pawn straight with queen take queen and then bishop takes F7. But I think, Yas, in your estimation this would be a draw.
Mr. Seirawan Absolutely. We have spoken about this for sometime. In a four versus 3 situation where bishop is off color and all pawns on the same side, it is a dead draw without any consideration whatsoever.
Mr. Ashley But will the computer play a move like this knowing for sure that it is going to lose a pawn? Because we know it's a draw, but it might not know it's a draw because there is no experience, first of all, to draw on and it takes a lot of moves to prove that this position is a draw.
Mr. Seirawan Outside of the move queen B7 DEEP BLUE could play the move queen D6 and perhaps transpose into that previous line, but I must say, and I don't mean to be too critical of Garry, queen C-5 check looked very forcing. Queen D6 would have been force response and then that whole line that we got into. So he must certainly have seen that.
The first thing you should always analyze, by the way, when you play a game of chess is check. You have the opponent -- if you have an opportunity to give your opponent a check, certainly, certainly consider it. So Garry Kasparov, the world champion, had considered that.
Mr. Ashley Wow, the answer has been made clear. It is play the move queen to D7 willingly, willingly offering the exchange of queens for a position that would be a theoretical draw, and I have to say I'm impressed, very impressed that they could program that, the loss of a pawn.
(Audience commentary.)
Mr. Seirawan The gentleman points out that in the bulletins regarding DEEP BLUE's strengths it has been programmed with a hundred million end game positions and that sounds even --
(Audience commentary.)
Mr. Seirawan I think you right. I think that is true, and the data base of DEEP BLUE is one of the most researched in the world, so it's conceivable, conceivable, that it saw something it could draw from its data base, but Kasparov is playing it smart. I think we are going to be transposing right back into that line we looked at before, queen C-5 check a moment ago, but it was a nice test. BLACK RESIGNS
Mr. Seirawan Understand that obviously chess is a fight. It's a battle. It's a war, but it is also about testing your opponent, simply testing, putting problems in front of your opponent and saying solve this. Solve this. So Garry saw an opportunity to ask its opponent to solve a test, and it did so.
Mr. Ashley That's very important. That just shows the cunning of Garry Kasparov because that last variation, he could have played queen C-5 check right and queen D6 which is actually forced. I don't see any other move than queen D6.
Like you said, he put this minitest to the computer. Now he knows that the computer can solve those kinds of problems. When he goes back with his trainers or whoever he is working with, he can say when I played queen B five that thing actually played queen D7 and offered me a pawn. It was not so materialistic as I thought. That was very, very smart of Garry. I'm very impressed that he did that. I'm impressed that Garry did that professional type of test, but also by the computer for the move it did.
Mr. Seirawan I must say he has done that throughout this game with those repetitions and so forth. I just got a chance to speak to Monty Newborn of ACM and asked him we got the news from Joe that the line went down and was there anything more to it than that. Was there some fault line didn't work or a safety line didn't trigger, and he said no, the lines went down. We were making all kinds of conjectures and conspiracy theories, but we couldn't prove them.
Mr. Ashley Queen D6 is played with the same understanding. If the queens are exchanged, no advantage for White. And Kasparov has gone for the other variation with queen to A7 check. We mentioned this variation before.
Mr. Seirawan This seems to be a direct transposition of the line we spoke about, and all that happened is Garry took two extra moves to get there and he said, I'm taking these moves to give a test to the opponent and it's wonderful. What I said before, and what Garry said at the start, all he wanted was to start this match seeing how DEEP BLUE plays on the offense, and seeing how DEEP BLUE plays on the defense and he is getting in his money's worth today.
Mr. Ashley I have to say I think the main weakness that many players have felt about the computers is its play in end games. Not its specific knowledge about exact positions, because, of course, it just knows those positions. It doesn't have to think about them. But general plays, bringing the king into the game, the importance of passed pawns, which ones, the significance of passed pawns. Always been a question.
I remember when Anan played CHESS GENIUS -- was it London? Yes, London when Kasparov had already lost to CHESS GENIUS in a match. Anan played the second game against CHESS GENIUS. It made one of the most horrible pawn moves you have ever seen, really horrible. And after that pawn move he said, I'm going to win this game and just killed it in the end game, and it could not foresee everything in that position.
But it sometimes will play end games very weekly, and I think that Kasparov feels that he has some chances here. The program is not going to know all the right things to do that he would know to do from experience, and he is really tickled in this position.
We are anticipating a move like queen A3 check, which looks very strong, but instead queen to -- 46. Qa8 Qc7. 47. Qa3+ Qd6. .
Mr. Seirawan A8.
Mr. Ashley A8?
Mr. Seirawan An interesting move.
I was going to remark on what Maurice was saying about the end games. It has really been true. If you go back 25 years and you start looking how computers were playing end games, it was great fun. We could all join in and have a good, good belly laugh, and then the laughs stopped, and they have been stopping ever since. To date we are saying that okay the end games still have been a bit of a problem for the computers, but they continually get better and better and better. We are seeing it today. This move queen D7 was very impressive.
I know a few people had their hands up.
(Audience commentary.)
Mr. Seirawan Well, the gentleman's question began with the idea that well, Kasparov gave up the A2 pawn because his experience would suggest that even though he was giving up a trump he would still have enough in hand to win the game, and I would say that that is a bit of a stretch by Garry. I don't think he got here after he gave up the A pawn, but certainly his experience told him even if I don't win the resulting position, I'm going a lot closer than losing or anything like that. He has no losing chances. And I just want to talk about his last move because that has taken myself by surprise.
We have spoken about Black's king, and again what Garry has managed to do is very, very carefully draw -- remember Black's queen was way over here. He has managed to draw Black's queen away. He has also managed to draw Black's queen away from this H pawn.
And this last move, queen A8, was designed to stop the king from coming back. What DEEP BLUE has now played is queen C7. So apparently DEEP BLUE could do nothing in this position. He has no plan other than just to prevent White from --
Mr. Ashley Looking a little closer to, Yas, I notice that Kasparov had a tremendous threat with his last move, queen A8 and what was to play the move queen to G8, the move queen to G8, let's get back to that position. If the threat queen G8 would have gotten behind black, attacked the F pawn, the F pawn just falls.
Mr. Seirawan But what we are saying is that if this was the threat and that was why DEEP BLUE could not make a casual move like bishop to D2 so in this position DEEP BLUE has made the move queen to C7 in response to that move. The light squared bishop is under attack. Kasparov would like to be able to move that bishop backwards to A2, but maybe it wouldn't be a big mistake in this position. He is going for this other position which looks quite favorable for him.
Mr. Seirawan Let's talk for a moment because I absolutely think you are right, Maurice. That move queen G8 would have been devastating. So Garry said to himself, let's give the computer an opportunity to make that mistake with bishop D2 because the best move for the computer was queen C7 which the computer found, and now Garry said I will go back to the main line, and you are going to have to play queen D6. So all along the line he is keeping his options open.
Mr. Seirawan . He is making those little tests.
Mr. Ashley Look at Garry now. I'm a bit taken by this scrunched up face.
Mr. Seirawan That's his game face.
Mr. Ashley I'm not sure, Yas. I have watched this guy over several chess games. It is not a happy face.
Mr. Seirawan This is a face of suffering as he is fighting his way through. When he is losing, you can really see it.
Mr. Ashley Yes. His hands are like this. He is covering his eyes. He can't believe it. He is looking to the sky, muttering to himself. It is very, very clear when he is losing.
But this last move shows that in every chess position there are possibilities to win. Every chess position you can squeeze and squeeze, and if DEEP BLUE were human, he would be suffering or she would be suffering, definitely. He would be just suffering. I can never win this stupid position and this person is playing a hundred moves, making you make all the accurate moves.
Mr. Seirawan I just point out a very nice little mate here that in this position if DEEP BLUE were to try to step backwards, there is this nice little mating combination.
Mr. Ashley King A would be continued by bishop five check.
Mr. Seirawan Notice how the bishops, the White bishop and White queen have been working beautifully in tandem with one another. And by transposition they have reached exactly the position that we reached about some 40 odd minutes ago. Along the way Garry has gained some vital information about the computer. He has learned that this is a very strong computer. 48. Qa2 f5. .
Mr. Ashley All of these moves have been played, these minutes have passed, and we came to this position ourselves pretty quickly, and that's okay, the computer, it is again a real illustration of how, yes, you can look at 50 billion positions in three minutes, but we get to the heart right away, just right away. Sometimes we will overlook some of those unique type positions but generally human says that is what you want. I can find it right away. Or within a much less of a time span. So here they have arrived at this position and now we liked Kasparov's position after the move queen to A2. We really liked Kasparov's position after queen to A2 crashing in on the A7 pawn, winning it not with queens off the board but with queens on the board, which means that the light square attack will regain its aggressiveness. He will be back in earnest looking for that black king. F7 will be weak. There will be a lot of potential dangers, and the beauty of this long-winded maneuver, not only eyeing the F7 pawn, is that the bishop on C3 can go to D2. I love it. I love it. H5, H6. It is going to be great.
I think it is now desperation time for the computer. I don't know. I see that Hans has joined us again. But I think it is desperation time for the computer in the sense that the F7 pawn is gone. If you allow me to simply chalk it, come back, reset, I then will have a position of queen and bishop versus queen and bishop where that F7 pawn has given some vital protection, and you will not have any of the defenses or fortresses that we have been seeing. I think it is time for Black to go for the gusto. And I think he is going to have to play.
If we do see this move, queen A2, by Kasparov, I'm sure he is calculating it now, the gusto being that you have to try to play E4 and use the trump of the central pawns. So now we might see an end game like this. If you notice, this is a totally, totally different end game from the one we have been talking about. Black has this passed pawn.
Mr. Ashley He is calculating the moves with E4 and D3 getting a passed D pawn.
Mr. Seirawan Pardon me. If Kasparov was to in this position play queen A2, I'm anticipating that Fritz -- that DEEP BLUE has to play E4, and then comes the move bishop F7. Now comes the move D3. Kasparov would play E takes E3, DEEP BLUE E takes D3 and we have a totally different end game structure. Now Black has a passed pawn on D3, but White has A3 to 1 majority on the kingside. It means that he might be able to make two passed pawns, and after a further move like maybe bishop, E3, we will see a move like maybe D2 and now we have a move like queen A7 check. I think that this is going to be very strong for White because the White squared bishop stops the pawn, D2 pawn, from going further. But this is what Kasparov is looking at. More to come...stay tuned for comments from Kasparov!
Mr. Seirawan So after queen A2 -- no. Queen B3, the gentleman suggests as an alternative, that way kind of snuggling up to that bishop on 3 and making sure that that pawn does not get frisky. That really was my original idea.
Mr. Ashley We were worried about queen B4 in this position.
Mr. Seirawan That's what it was. We were worried about queen B4.
Mr. Ashley And now an offer of the exchange of queens, and now, if you go back in that position.
Mr. Seirawan Then he goes back to queen A5.
And you are not making progress.
Mr. Ashley You knew there was a good reason that my move would have been rejected.
Mr. Ashley So queen B3 is queen B4.
Mr. Seirawan Poll time.
Mr. Ashley No, not another poll time.
Mr. Seirawan All of those that think Garry is going to win it?
Oh, yes. Oh, yes.
We are not going to ask if the computer is going to win it, but who thinks that the computer is going to save the game?
Mr. Ashley A lot less.
Mr. Seirawan Here we have it. We see a different --
Mr. Ashley Kasparov has indeed played queen A2.
Mr. Seirawan And we have seen a different form of desperation. I predicted the desperate move E4. Instead Fritz came up with --
Mr. Ashley DEEP BLUE played F5 in a heartbeat.
Mr. Seirawan Very quickly.
Mr. Ashley Well, and now after all of the beautiful poetic maneuvers, the computer has decided to attack. 49. Bxf7 e4. .
Mr. Seirawan Kind of like James Bond. I'm losing.
Mr. Ashley Now bishop F7 and Kasparov with his hands hovering over the F pawn, and now he has captured it. And now the major question is clear. Kasparov has won the critical F pawn, the last pawn to guard light squares in the heart of Black's position, so it is possible that Kasparov will penetrate along the light squares, but he has done so at cost, and that is lifting the blockade off of Black's central pawns and now they are beginning to exert force on the board. Now, they can possibly march F4 in different variations. Kasparov just did something that you may laugh at. But Danny King when they are doing commentary for the Grand Prix events, we love when this has happened. He has taken off his wristwatch. It took the whole game for him to do that. I was wondering when that was going to happen. Every single time you watch Kasparov play a game of chess, if he comes off with the wristwatch, he is ready to go.
When he thinks I really have to look at this position and he thinks he is worse, the jacket is automatically off. That means this game is heating up for him.
Eagerly marching forward is DEEP BLUE with this mass of central pawns that it has now played the move E4 and that looks a bit intimidating to me. If it weren't down a pawn, I would be liking Black's prospects a lot.
Mr. Seirawan But the point there is, this tells if you will, poker players that are familiar with poker, understand that when they play a point oftentimes the opponent reveals their thoughts by scratching themselves or doing something of nervousness and because I read Kasparov so well, I have always been wanting to get into a poker session with him because I know that wristwatch comes off he is bluffing.
So we have exactly what we predicted. DEEP BLUE has realized its predicament and is getting very serious including just sit there and wait for Kasparov to build back the position. So it has now played E4 and I'm expecting that after the move bishop C4 just preventing D3 and preparing to bring my queen down here, A7 attacking the black King, finally Garry can do what he does love to do so very well, which is going after his opponent's king.
We did have a question in the back.
(Audience commentary.)
Mr. Seirawan Indeed a very good question.
The question is basically in a position like this where black has pawns on D4 and E4 there is the -- there is a very nice trick which involves playing D3. After a capture, to then play a move like E3 trying to break up, trying to break up the pawn structure around White's king and I think that is something that Kasparov will have to look out for.
Mr. Ashley I hate to be the pessimist, but I think that in this position, I think that this is the kind of position that DEEP BLUE has been waiting for, and the reason I say that is how many advances does the computer have? It could play with D3, looking to create a passed pawn. It could play with E3 looking to break up the kingside position and it could play with F4 with the exact same intent. I'm afraid of F4 because of what it could do to the White king position, so now we have a situation where many, many complicated possibilities present themselves, not a single one of them very clear what is going to happen, and this is precisely the kind of situation DEEP BLUE excels in being in. I still feel that White ought to be able to control the game, should be able to, but it looks as if it is going into that realm of chaos, a bit less control on the chessboard and that brute force calculation, the ability that the computer has to do that so well, will serve it much more so than it would serve us to have a little feel for the position.
I have to admit that now I realize that if I had this position as White now I would think now it is work. It's really work. It is not just we can rosily say let's play that little maneuver we were thinking about. This is dangerous.
Mr. Seirawan I will take the role of the devil's advocate. That is to say that a while ago I was worried about fortress IBM that is the pawn on F7 and F6 and E was really hard to break and now I have to be happy because Garry has made some very, very serious progress.
The F7 pawn has been snipped and although I agree with you, Mauer read that these pawns on F5 and E4 represent forms of counterplay, do keep in mind that they are on light squares. Before my bishop could not attack them because they are all on dark squares, so now they have been made more vulernable.
Do keep in mind that I could mate Black very easy, for example, after a move like F4, just to put it on the board, what did I do? I go like this. No. I go F4. I can make black and just a couple of moves and I will just show you one quick one right away. Queen F7, king F4. And I thank you for your aggression. I love pointing out those mates. They are lovely to muse over. So if black is going to throw himself at me, I will be very happy in beating him.
(Audience commentary.)
Mr. Seirawan The young gentleman has a very good question. 50. Bh5 Qf6.51. Qa3+ Kd7. 52. Qa7+ Kd8.53. Qb8+ Kd7.54. Be8+ Ke7. .
Mr. Ashley If White were to decide on playing bishop to C4, which Kasparov did not decide on, the young man had asked what if E3. When we get to Kasparov move bishop to H5 in a movement.
Mr. Seirawan I think this is a reasonable idea because what you want to try to do with the move E3 is break up the King side. But after a move like F4, the reality is that White's king is still quite protected. I'm about to go down with my queen to A7 check, as I predicted before, and also this pawn on F5 is weak and these two guys are blockaded. I could play bishop C3. Let's take a look at what Garry did instead. Garry tried the other approach, bishop H --
Mr. Ashley That's a bit of a strange move to me.
Mr. Seirawan Surprise.
Mr. Ashley A bit. I mean not a bad move, but a bit of a surprise because the bishop seems to be off side and it doesn't have any potential squares itself. It has no safe moves except to go back to F7.
He is planning now to play the powerful move queen to F7 check and rip-off that F5 pawn to which he will have two extra pawns. Not a bad looking move. You know, it was a surprise at first but looking at it even more you see the point to the world champion's idea. Now the question is what can black do? Maybe a move like queen E6.
Mr. Seirawan Fritz thinks of the 31 moves at his disposal, queen E6 is the best among the group.
I point out this move bishop H5 is very much in the style of Garry Kasparov. Not only does the move introduce the threat of queen F7 check, but there are also variations where the E8 square is now protected too so that you can imagine let's say the bishop going back to F6. I will give you a queen E8 mate. So he likes eyeing this queen diagonal.
Mr. Ashley Fritz 4 is enjoying this position with White. It thinks that White has a substantial advantage, not just the extra point, which is evident, but it gives White a decisive winning advantage, so it thinks that there is some real chances here.
In response to bishop H5 has been played by DEEP BLUE and that is stopping the pen trace on the F7 square but when you think about it, it is quite for White that black's pawns are not inching any closer. As long as he can keep Black from inching those pawns closer, he can carry out his business from getting his pawns.
What is going to be White's plan? I'm a bit puzzled here, yeah. Maybe you can help out. What is going to be White's plan here, how is he going to press home his trump in this position.
Mr. Seirawan I think it is pretty straightforward.
Mr. Ashley This move G4 is going to play an essential part in the game. I think the great thing about Kasparov in this game is the way he has kept the computer out of its element for so long. Here it seems as if the game has lost a little bit of its control and he still kept it where it is not so complex and he is the one that is always striving for the initiative in every single position. He is back to that chess that he should have been at in the first game where the computer's strengths are much less, planning is much more important. You know, the positions like where we found a position, immediately found this planning setup where we said queen on A2 and bishop on C4 and we figured out how to get to it. That is the kind of situation where calculation is irrelevant. Not totally but less so. So he is doing that in an excellent way, just beautiful
Mr. Seirawan Garry has played great chess. What I think Garry would really like to do is once again draw the black king as far over to the queen side that he can.
Mr. Ashley What about that move you mentioned queen E8 check?
Mr. Seirawan That's fine. I'm just wondering if he has maybe even a stronger variation and a move like queen B7 check after say queen D8, queen D4 check -- queen D5 check. I beg your pardon. And now queen D7. And now to try to introduce the bishop into the game. Maybe with a move like bishop to E8, keeping in mind I'm eying these pawns so that they can't move down the board.
Mr. Ashley And that queen is so beautiful on D5.
Mr. Seirawan And I'm preparing to play queen D7 or in some cases just very simply bishop D7 and take that fellow. So I think we're getting to a position now where Garry is a favorite to win a second pawn.
Mr. Ashley Which would be a win win kind of position.
Mr. Seirawan You bet. And by the way we do see Fritz agreeing with me somewhat that, okay, Garry is a pawn up but he sees a third of a pawn is on its way.
Mr. Ashley Still, though, there's a lot left in this game. There's really a lot left in this game. It's not just something that's over and done. Kasparov's advantage will not be checkmate because he's not going to checkmate the king. He will try to win some material and then queen a pawn. So queening the pawn there are still some obstacles in the way, in the shape of a black queen, potentially the dark squared bishop. The king of course is being pushed away systematically from that side, but still a bishop and a queen can put up a really formidable defense. So we're still now -- where were we now? Move 54. I would say another 20 moves still before the game really becomes clear.
Mr. Seirawan And by the way this is very consistent with the strength of the computer. I have noted that as the computers have gotten stronger and stronger that my ability to defeat them takes me longer and longer. Before I might be able to checkmate a computer in 23 moves. Well, now it suddenly becomes 40 and sometimes 50 and sometimes 60 moves. So I've got to put in a lot of energy to beat the computer. And I'm sure Garry is looking forward to a free day tomorrow because okay the first 1 really rattled and the second 1 maybe he takes a deep breath. But for Garry to beat the computer he won't be -- they won't be miniatures.
Mr. Ashley No.
Mr. Seirawan It will take a lot of time.
Mr. Seirawan Also I'd like to talk about something else and that is the worldwide web. Are we all familiar with the worldwide web and the internet? All right. Well basically what is what has happened is although we love chess and certainly we're fascinated by this match artificial intelligence, we cannot obviously convince the broadcasters of ABC to cover 6 hours live of a chess match. But we can convince the people -- we can convince ourselves that we should do it on the worldwide web and the internet. So there are many sites including IBM site and the USA today site and there are other chess sites. I know there's a GMT site, the ICC site and all around the world a lot of sites are featuring this game. Now, what happened yesterday is that the IBM site was hit so often that it absolutely gummed things up and thousands and thousands of people -- I think Terry was telling me that they had 6,000 hits per minute, 6,000 hits per minute globally all over the world. And I know the other sites were also being inundated by people who wanted to follow the match live. So you had mentioned the USA today site, the IBM site and I think that they now are making more and more hardware, team 1 lines available just to accommodate millions of people.
(Audience commentary.)
Mr. Ashley A question about time management. Kasparov is in great shape. He made the first 40 moves in the allotted 2 hours. He got an extra hour on his clock. He now has 33 minutes remaining to make 7 moves. That's plenty of time because he's now on move 54. He's about to try to make his 54 move. So he has plenty of time on his clock to work this out. But the next time control would be sudden death in 30 minutes. So he doesn't want it to be too complicated in that time control because then he's got to make decisions very quickly. And if his time runs out he simply loses.
Mr. Seirawan Let's speak about that for a moment. There are 3 time controls in this match. The first --
Mr. Ashley Wait a second. He just swung that bishop down to E8. Not straight on the diagonal. He swooped it around in a circle. I was thinking what is he doing? Has he to forgotten how to win a bishop or what?
Mr. Seirawan Bishop E8 check, introducing the bishop into the game. I was a bit surprised because I was kind of happy about the bishop being there maybe it going to G6. So Garry has a forced variation in mind obviously. In fact Deep Blue's only illegal move is -- oh, no he could actually play E6, but we anticipate king E7. But there are 3 time controls. The first is 40 moves in 2 hours. So that's pretty straightforward. We get plenty of time to think. The second is 20 moves in 1 hour. And indeed DEEP BLUE has played the predictable king E7. Garry is speeding up here. I think he heard your suggestion of time management.
Mr. Ashley Time management and sudden death.
Mr. Seirawan So he swung his bishop around, but now he's kind of given DEEP BLUE an opportunity to move his king over to -- no, I can only chase you out. That's good.
Mr. Ashley What were you saying? What if the move king to F7 is played?
Mr. Seirawan I'm thinking if king to F7 -- 55. Bb5 Bd2. .
Mr. Ashley After that last move bishop to B5.
Mr. Seirawan Then Garry would play bishop C4 check. And after king to G7, I could make a move like queen to G5 check, king to H6, and I got to have something good here, please.
Mr. Ashley It's not clear. I would hate to see the black king end up on H6 again after all that hard work to get it over on the queen side. Now suddenly it's standing on H6, blocking my pawn like Akeem Alajawan. I'm not really appreciative of that.
Mr. Seirawan This kind of goes back to what the gentleman had said to me before, that it was a forced win. And I thought that what black should have been doing was parking his king on H6. And I thought it was a mistake by DEEP BLUE to -- and it was very crafty of Garry to force the king away from the H6 square. So now giving black the opportunity with his last 2 moves of bringing his king over to the king side is a little bit disconcerting. I don't think I would have given DEEP BLUE that opportunity.
If I may just to finish the thought about the time control, the third time control is after 60 moves, each side will get 30 minutes for the rest of the game. So even if they play another 20 moves, he doesn't get anymore time. And that will be very, very difficult for Garry because the computer will be playing in nanoseconds, and Garry will be taking minutes at a time. So Garry certainly doesn't want to see his 70, 90, hundred move game. He just wants to see a nice sedate 60, 65 move game.
Mr. Ashley This game is not going to end in ten moves, though.
Mr. Seirawan I don't think see either. I think that the best move for DEEP BLUE now would try to be able to play king F7 and move his king as quickly as he can back to the king side.
Mr. Ashley Yeah. It looks like a logical plan to me. Why not?
Mr. Seirawan Why not indeed.
Any questions?
Boy, we're doing such a good job that all your questions are being answered.
Mr. Ashley We're glad to hear that.
We have a question over here.
(Audience commentary.)
Mr. Seirawan Very good. A very salient question. When you're on the defensive, it's really a good idea to reduce the material as much as possible. And especially to reduce the pawns. So the problem, however, for black is it's very hard for him to reduce his pawns. If he tries to push them, they'll get captured. If he tries to push his E3 -- pawn to E3, it can be bypassed with a move -- let's go back for just a moment. If you try to play the move E3, it could get bypassed with a move F4. So it's very hard for black to force an exchange of pawns. DEEP BLUE has surprised me.
Mr. Ashley He surprised everybody, bishop to D2 has been played.
Mr. Seirawan Bishop to D2, mm-hmm. So he didn't take advantage -- no, DEEP BLUE didn't take advantage of the opportunity to bring his kickback to the king side. This may haunt him. Because now through a 4 series of checks we can definitely ensure that the black king is really far away. Garry has this opportunity of forcing the king to the C5 square. Maybe now with a further move, for example, like bishop to A6, I could threaten in some cases to bring my bishop back down --
Mr. Ashley You have to be careful. Bishop to A6 might hang. Even then you could rip a couple of pawns off but --
Mr. Seirawan But check.
Mr. Ashley Yeah, but not necessary.
(Audience commentary.)
Mr. Ashley Win what pawn on A4?
A VOICE He means D4.
Mr. Ashley D4 is defended.
Mr. Seirawan The gentleman suggested that queen A7 does attack this pawn. Right now the pawn is defended. Let's go back. Okay.
(Audience commentary.)
Mr. Seirawan Well, we've been speaking about that and seeing that for several hours. Provided that White keeps the queens on the board, the bishops of opposite color, it's good for White to have bishops of opposite colors. But as soon as he trades queens, Black's happy he has the bishops of opposite color.
Mr. Ashley The basic rule is if you can simplify it as such, if the bishops of opposite color as you see here, dark squared bishop versus light squared bishop and both sides have queens and rooks or any such combinations, a single rook -- single rook is less so. 2 rooks -- I would say pieces more like a queen and a rook or even sometimes a case of single queen, as we see here, the side that's on the offensive has really good chances of winning, really good chances because of the constant harassment on the other color, constant. You can never defend the other color because the bishops just are crossing each other, passing each other like ships in the night.
So this situation is still full of life, full of fight because Kasparov has a queen on the board and then there are extra factors like he has an extra pawn, the exposure of the Black king. All these lend themselves -- lend these ideas to Kasparov maybe having a serious if not completely winning advantage in the position. But as soon as the queens go, poof, that's all changed. Then it's drawn.
Mr. Seirawan I notice that Fritz has suggested just playing the move H5 and that were White to do so, then Fritz anticipates that White has a 1 point 78 advantage. And I'd like Hans to answer this question, if we talked a bit about professional etiquette and okay, my program is losing, boo-hoo-hoo, I want to resign the game. What is that threshold where you, as a programmer, have to, you know, pick up your chips and go home?
(Audience commentary.)
Mr. Seirawan So to repeat, your answer is that until you get 5 point 0 down, you ain't giving up nothing. 2 pawns isn't enough.
(Audience commentary.) 56. Qc7+ Kf8. .
Mr. Seirawan Very good. All right. So the answer there is that I mean Kasparov has continued to build on his advantage and we've seen -- whoops, Fritz didn't like that last move, queen C7 checked. But I was wondering if a 2 point 0, a 2 pawn advantage was enough. Hans says not for him. He wants to see the 5.
Mr. Ashley That's a lot, man.
Mr. Seirawan That's a lot.
Mr. Ashley You can forget it.
Mr. Seirawan Okay. Again, Garry is not taking advantage of the opportunity to drive Black's king towards the center and the queen side.
Mr. Ashley Isn't that training? He's letting the king go back where it can help the pawn out. It seems like poor strategy.
Mr. Seirawan On the contrary. He's actually forcing Black to go back. I think the move king E7 might just --
Mr. Ashley King E6 might be deadly.
Mr. Seirawan King E6.
Mr. Ashley Bishop C4 with mate, hello.
Mr. Seirawan Did you guys all pick up on that 1? That was a good 1. Okay. So we're going to see the move king F8. King F8 is going to be forced. And again, I don't quite understand why I should be forcing your king to a side of the board where you can blockade my pawn. So I'm not happy about that decision.
(Audience commentary.)
Mr. Seirawan Indeed. Time management question. Here you go again Hans. Here we have a perfect situation. An absolutely forced move. There are 2 alternatives.
Mr. Ashley King E6 is mate.
Mr. Seirawan It's mate 1.
(Audience commentary.)
Mr. Ashley But you don't even have to go to depth in it except 1 half ply.
(Audience commentary.)
Mr. Ashley Okay. It's been thinking now --
(Audience commentary.)
Mr. Ashley But here is mate in 1.
(Audience commentary.)
Mr. Ashley That seems so strange to me. I mean it's -- okay. How many -- we've had at least ten seconds passed so that means it's already analyzed 2 billion positions. Right? So -- well, you know, king E6 is mate. That's 1 position. What the heck else are you looking at? This there's only 1 more. There's 2. It seems really paradoxical. It's down right weird.
(Audience commentary.)
Mr. Ashley The question is, maybe White wants to put A pawn on H6 -- A pawn on H5 and then put A pawn on G6 and exchange queens and that way promote the possibility of the g pawn becoming a queen because now it's on a color the same -- the final square with queen on it is the same color as the light square bishop.
Mr. Seirawan I can certainly address that. 1 of the beautiful areas of chess that I really have enjoyed is the area of problem chess. It often happens that in problem chess they set up positions where White has 3 pawns more than his opponent, than bishops of opposite color position, and it's not enough. So again White must avoid at all costs a trade of queens because I think it would reduce his end game chances almost to zero even if the pawn had become a G pawn. H pawn or G pawn, it's not a big enough difference. And in fact if Garry was to win A pawn and just have a 2 pawn advantage, it may not be enough to win the game. 57. Bc4 Bc3. .
Mr. Ashley Well, this is very interesting. Kasparov's last move was bishop to C4 preventing the king's march to the G8 and H8 square up to H6. And this has already created a pretty interesting situation because Black is really running out of great moves. Black's queen is rooted to the F6 square or limited in the range of where -- in fact, I think it is rooted to the F6 square.
Mr. Seirawan Exactly.
Mr. Ashley Maybe G6 is another square for the queen. But this problem of queen F7 mate presents itself. If the Black queen were to leave, of course it wouldn't be done now, but if the Black queen were to leave. And also there is another problem of H8 check and taking the pawn if the queen were to leave. And queen G6 is not a viable square because queen G7 and queen D4 check. So Black's queen is rooted to the square it's on. That leaves Black with very few moves. Black's king has 1 move, king E8, and the bishop maybe can move to H6. That is an alternative. But it's still puts Black in this closer and closer to supervision. Maybe you can elucidate on that, Yas.
Mr. Seirawan In fact, this is actually an interesting area that for some reason the computer has a little bit of trouble with.
Vooksvung is a German word which basically means compulsion to move. You're compelled to move. There are positions where you set up the perfect fortress. Every little piece, every little pawn, everything is in perfect working order. You've reached your maximum setup. And now your opponent says okay I'll move my piece, I'll make an innocuous move. I'll pass. Please, your turn. And you say but I don't want to move. I've got everything set up the way I have to. But then you're state in a state of vooksvung. You're compelled to make a move that weakens your position. Computers actually don't understand vooksvung very well. Oftentimes the best way of winning a position is to force your opponent or compel your opponent to weaken his own game. So that is to say, don't go for the direct attack. Rather, pass. Literally making a move that says okay this move does nothing. It doesn't injure my position, but now I'm going to force you to injure yours. The computer doesn't -- well, I don't know DEEP BLUE on this 1, but the computer doesn't understand very well the point of vooksvung. The point of vooksvung being to force the opponent to weaken his position, the computer will usually try to force a direct win and therefore not take full advantage of its position.
Mr. Ashley And there's some beautiful vooksvung situations where you just make any old move. You just move the king, move the pawn and in some situations you wouldn't believe it is vooksvung. You have to look at the board a few times to believe the fact that it's your turn to move makes you move. And you'd prefer not to move because you have such an ideal setup.
Well, in response to Kasparov's bishop C4, DEEP BLUE has changed direction and gone back to the square C3. And that is -- that is certainly a bit more in Kasparov's favor because the bishop was planning to get back on that long diagonal on H6.
Mr. Seirawan Help me out here.
Mr. Ashley But certainly -- I would say that's something that Kasparov wanted to have happen, Yas, this setup that we see there.
Mr. Seirawan Exactly which we're about to get to because I hit a wrong key. Yes. And here we have it.
What I was going to say, bishop C4 was played, bishop C3 was the answer that -- and now Garry has said, isn't it lovely. All you're doing is playing bishop D2, bishop C3. I've got you perfectly where I want you, you big bad computer you. Your queen can't move. Your pawns can't move and your king can't move. Therefore, the only piece that can move is your bishop. So as you're going bishop D2, bishop C3, what I'm going to do is I'm going to just improve the position around my king side. So king G2 getting off a dark square, avoiding any potentially annoying checks.
And the other thing that Garry can very nicely set up is he can try to set up the position -- oops, excuse me. Let's assume that DEEP BLUE plays bishop D2. He can try to play the move F3. If after F3 he forces the move E takes F3, check, E takes F3, what happens is these 2 pawns, the pawn on D4, the pawn on F5 are now splits and weak. And therefore after further moves like bishop D3 and queen C8, I'll go after it and win that pawn. If you don't play -- if you allow me to, I'll just capture this pawn and through a series of checks I'll pick up on the pawn on E4. So Garry will soon be working on a direct win. In the meantime he says I'm just going to make all these tiny little moves.
Mr. Ashley Wonderful planning.
Mr. Seirawan And I'm going to win the position with little moves. That's a very nice thought.
Mr. Ashley And Kasparov leaving the board again and the challenge here on DEEP BLUE with nothing to do, it must simply sit and defend the position.
Mr. Seirawan Another thing that actually looks very attractive, I see it now, I don't know why Garry didn't do that as well, is let's assume that we see a move like bishop D2 now, not F3 anymore, but we can now also look at the idea queen D7. The idea being to simply follow it up with the move bishop E6 to take that pawn and then to take that pawn. So a couple of direct ideas here and in the meantime no counterplay for DEEP BLUE.
Mr. Ashley Yeah. This is a painful position. It's painful for a human to play for sure.
Mr. Seirawan Right.
Mr. Ashley When a guy has 2, 3 different possibilities and you have to be on the lookout for every single 16 them. Kasparov is really playing it professionally and it's taking him quite some time, but he's not going to do what he did yesterday, which is go directly after the computer and have these forcing variations. But, clearly, we have a situation where it's still going to take him some time to execute the win, and he's --
Mr. Seirawan And the time again, he's got 26 minutes.
Mr. Ashley 26 minutes to make only, what, 3 moves?
Mr. Seirawan That's fine.
Mr. Ashley That's easy. So after he makes his next 2 or 3 moves Garry will then have a half hour. Whatever time he's left with at that point -- well, that's all the time he'll have for the rest of the game. So again I mean if the computer can last and last and last and draw out the game as long as it can, that seems to me to be the computer's chance, which is just to play as many moves as possible and hope that Garry physically won't be able to make all the moves he has left.
Mr. Ashley There are so many factors. And people sometime say chess players, all they think about is chess. They're not interested in anything else.
Mr. Seirawan That's not true.
Mr. Ashley We know some chess players are like that. But we know that in order to play chess well, 1 great quality is physical fitness.
Mr. Seirawan Indeed.
Mr. Ashley A great quality. You need to exercise. Anand, when he had to play Kasparov a match, was biking 15 kilometers a day every day to get fit for the match. Kasparov is a work horse. He goes to spas, get massages, does a lot of swimming to stay healthy because imagine sitting down there with no breaks. This is not like a job, you know, where you get your 900 to 500, but you work from 9 to 12 and think I had enough of this job and go to lunch, you break for an hour and come back and finish up and do your 7 hour day.
This guy has been working from the beginning, and he has to work all the way to the end. No breaks. The guy makes a move, he's got to come back and play chess. And if he's not fit to do that, he's not going to be able to last. Fatigue is going to set in, and he's going to start making errors creeping into his calculations, and then he'll make mistakes and lose the game. So a guy like Kasparov has a serious regime before every tournament where he definitely gets himself in tip-top physical condition so that he can fight these chess games without feeling the fatigue. I mean look at him. He's working hard at the board, head in hands, working hard, a lot of energy. And he'll feel it. After a good chess game, you know what it feels like. You go home, oh, man that was a serious fight. Your head is pounding, your body is feeling achy. Just want some rest.
Mr. Seirawan You're drained. It's quite true that physical labor is a beautiful labor, and you really feel it in your muscles and you can take a nice shower and you still go out and have a good dinner on the town. No problem. After a good hard physical workout. But a good hard mental workout doesn't allow you anything like that. All you want to do is crawl back to your hotel room, get into bed, read a good book and just go to sleep and turn out the lights. And it's absolutely true that after a very, very hard grueling chess match, chess players lose a lot of weight. I mean Kasparov is quite lien. I think it's's around 165, 170. But I'm sure after some of his matches with Anand and Karpov he's lost quite a number of kilo.
Mr. Ashley Well, Karpov loses quite a bit. I remember Karpov when he played Kasparov the first match, the very first match the play lasted, what was it, 47 or 48 games, over 5 a 5 month period. People who saw Karpov said he looked as if he just came from the grave because his face was so drawn and pale and he was taking stuff -- drinking coffee all the time trying to keep his energy up, but the strain, the physical, mental agony of having to play an arduous match is like nothing you know. And I mean we've played matches and I know it's -- it never leaves your mind. When Kasparov, we're walking around after this game, we think good game, interesting, we come back on Tuesday and no problem. Every one of Kasparov's waking moments somewhere in his mind is this stupid computer. He's going to have to go play. Somewhere in his mind it's the strain of this match. He could be eating, just casually watching TV, but in the back of his mind it's like when you have something really weighty on your mind. It never leaves you. And it's the strain of the match and you see him there exhibiting some of that energy he has to use in the game. Just the strain of the match makes it so difficult. It's rough. You're glad when a match is over. Air ecstatic. Just finish the stupid match because it's so hard to maintain that intensity over a weak, 2 weak, 3 weak period.
We have a question here.
(Audience commentary.)
Mr. Ashley He said isn't the computer programmed to look at checks first and if those are decisive wouldn't it just stop.
(Audience commentary.)
Mr. Ashley So what you're saying is that mate is not -- the computer didn't find mate to be as important -- so important that it will stop analyzing a variation that immediately leads to mate and do something else?
(Audience commentary.)
Mr. Ashley Well, that's a weird thing for a computer to to do. 59. Kf1 Bc3. .
Okay. DEEP BLUE played the move bishop to E1 and I'm not sure exactly what that did, but Kasparov has decided to remove the bishop from his position by playing the move king2 F1. And that is his 59th move. He needs to make only 1 more move for this time control. He has 21 minutes on the clock. So he need not worry about that at all.
Mr. Seirawan I was saying earlier that 1 of the things that Garry is certainly considering is F3, in order to break up this team, Black central pawn team. But after the move F3 this bishop on E1, you know, well, it's annoying. It's simply annoying. So what Garry has done is he's decided that before he does anything, let's just get rid of that bishop. So he's played king F1. It in no way juries his position and he's just saying get out of here, move that bishop away t I was going to say that also instead of the move king F1, okay, let us assume that we see a move like bishop C3 from DEEP BLUE. Then this move queen D7 I've been looking at that for several minutes now, not calculating 200 million moves a second, mind you, but I've been looking at it and it really looks like a really very, very powerful move. This is almost winning A pawn by force. I don't see how the computer is going to handle the move bishop E6 and bishop takes pawn.
Mr. Ashley Well, let me understand your threats. Let's say for example bishop D2 is played in this position.
Mr. Seirawan Okay. Let's try bishop coming back to B2.
Mr. Ashley And if you play bishop E6, play F4.
Mr. Seirawan Bishop to E6, F4. Now I have several things that I can do. I can simply win the E4 pawn. I can give you a series of checks beginning something like check, king here, check and take that E4 pawn.
Mr. Ashley Wait. Let's improve on that defense because maybe king E7 is possible after check.
Mr. Seirawan Okay. Check, king to E7. Okay. So you're coming automatic me from a different direction. Uh-huh, that's true. And after queen D7 check I would only be repeating the position. So I could do something maybe better with a move like queen check -- no, I can't get what I want.
The other suggestion is that after bishop E6 forcing the move F4, a straightforward move might be like G4, the idea being to simply play G5, never forgetting that I have this F7 square mating you.
Mr. Ashley Yeah, that's tough.
Mr. Seirawan So G5 becomes a very serious threat. Just to continue the line a little bit further, you can play F3 and I can play a move like E takes F3.
Mr. Ashley DEEP BLUE has in fact moved his bishop back to C3. It has retreated its bishop to C3 with nothing really better to do. And Kasparov with his options here. And funny thing Kasparov just did. It's a bit hilarious because you know he's playing DEEP BLUE. He did something else. And you see him doing it again. He's looking up at the computer operator. Kasparov does this a lot when he's playing against someone just at the moment when he thinks he's winning. He looks at you like are you stupid? I mean, really. And he does it every time he thinks he's winning. I've seen him more than 40 times playing the game. And just at the moment where he's thinking you're dead, you know that, right, I'm just going to kill you, he starts looking at you like you know it's over. Are you ready to give up? And he looked up at the computer operator as if to expect that response from him. But I love the way they're so stoic. They're just sitting there like you know nothing. We're going to act like the computer and just be stone faced. We may see -- we may look -- no matter what. And I think they may have discussed this. No matter what they see on the screen, they could see mate in 6, you know, and somebody sees mate in 6 you know, they go oh, we're going to lose. But they're going to be stoic about it like it might be mate in 6 but you're not going to see it on our faces. And I think Kasparov looked up looking for clues, maybe this human across from him will reveal something to let him know that he's on the right track and Murray Campbell is sitting just like I'm a rock. Forget it. If it's a win you're going to have to find it yourself.
(Audience commentary.)
Mr. Seirawan Yes. We're currently -- Garry has --
(Audience commentary.)
Mr. Seirawan Oh, I'm sorry. The clock they're using is a FIDAY standard clock. FIDAY is a French acronym. FIDAY, or the International Federation of Chess, is the third largest sporting body in the world and the standard clock they're use something a FIDAY clock which is a digital clock so that the players know exactly how many seconds they have. And I can tell you that Kasparov has 18 minutes and 131 seconds to make his 60 move of the time control.
By the way just to speak on that for a moment, a lot of you probably aren't aware of how very closely chess parallels the most popular sport in the world. The most popular sport in the world is soccer or football. And the largest sporting body is the international Olympic league, Olympics to you and myself, the next 1 being Atlanta this year. The next largest sporting body is FIFA, the Federation of International Football Association. It has about 185 member countries. And the third largest sporting company in the world is FIDAY, which has about 170 nations where throughout the entire world people are playing the game of chess using the same rules they are here. And if you think about it for just a moment, that's pretty impressive because even the game of basketball which I see dearly love is played differently whether you're playing it on a high school level, a college level or the professional leagues. So chess --
Mr. Ashley Or in different countries, mind you. There are is some European basketball and U.S. basketball.
Mr. Seirawan Right. So whether I'm with Frederick in Germany or I'm in the Great Wall of China or the beaches of Rio, I'm playing the same game. And it's a very popular game and it parallels soccer mightily in the sense that soccer is the -- is so popular in the rest of the world but if you want lose $100 million let's start a professional soccer league in North America. It's a great way of losing a lot of money. And chess is really played overseas where all the great big international competitions are. That's why we're so lucky to have this competition here in Philadelphia.
Mr. Ashley It's happening more and more, though, isn't it, that chess is finding a foothold in the United States, all these big tournaments being held. Naturally the world championship was held at the World Trade Center in New York City. The New York Grand Prix which has been held there 2 years in a row with a prize of $160,000. This event, a number of other tournaments, 1 big tournament happening in March which I'm going to have the pleasure of playing in, in fact, against some of the great chess players of the world. So it's really happening that chess is taking off. Well, maybe not taking off but certainly gaining a foothold here in the United States. Well, Kasparov has played -- 60. f4 exf3. .
Mr. Seirawan A major move.
Mr. Ashley Something that -- I mean this changes so much. And it's amazing that he's decided on such a concrete way of carrying out his plan. But it seems as if he has his sights set on that F pawn, the pawn on F5, and he does not want that pawn to go anywhere. And so he's putting on the pawn on F4 first. There's a famous principle that when you want to attack something make sure it can't move out your way. But F4, very dramatic. That seems to bypass the 2 Black pawns, but maybe Kasparov has estimated that they're not going anywhere.
Mr. Seirawan Indeed. A moment ago we just looked at the variation, if we can go back -- oh, you've overwritten it. Okay. Go ahead and play F4. We had been looking at the variation where we had played queen D7 setting up that bishop E6 and we saw Black as a response play the move F4 himself. So Kasparov with this move really wants to prevent that defense at all from happening. But just as you say it's a very committal move. I had been looking more at the continuation with F3. But Garry, as Maurice has said, is setting his sight on a fixed target and that target is the F5 pawn.
Mr. Ashley I think that Kasparov has actually carried out a specific way of getting Black to take that pawn on F3 because in that lineup that you talk about, queen F7 bishop F5, the queen pawn is dead.
(Audience commentary.)
Mr. Seirawan Has Fritz anticipated that E F3 is the best move?
MR. FREIDEL Yes.
Mr. Seirawan Okay. What Fritz has suggested for DEEP BLUE is that it cannot allow this pawn on F4 to remain. And in fact DEEP BLUE has captured the pawn. Before we capture the pawn let me just go back for a second and say if you take a look at the position, we know that Black's queen is trapped, is literally -- must play guard to the F7 square. And the move F4 stop the bishop, Black's bishop, from defending the H6 pawn. So that H4 pawn would have waltzed up the board if he hadn't played en passant. So we have seen the move en passant. 61. exf3 Bd2. .
Mr. Ashley And Garry captured back on F3 immediately. So the landscape in this position has changed dramatically. I mean if you look at it now, what was once a healthy chain of pawns for Black has now become really scraggly pairs separated with this last exchange would seem very forced because of the weakness of the F5 pawn and now Kasparov can again set his sights on that pawn, in addition to the fact that he may have a pawn duo racing down the board. When he attacks the pawn on F5 it will have to move and then Kasparov can play the move G4 and that's going to be very nice for him.
Now DEEP BLUE playing the move bishop to D2 and for some reason Kasparov looking at Murray Campbell there. I don't know why. But now some interesting things. Kasparov has to be very careful about his own king in some strange situations, really not right now. His king is a little bit more exposed. Not as exposed as Black, certainly, but a little. And now it seems as if Kasparov carries out his plan of queen D7 and bishop to E6, the F pawn will have to move and then he can play the move G4 if and start booing I go with the pawns down the board.
Mr. Seirawan Exactly.
Mr. Ashley I like that plan.
Mr. Seirawan This is what makes human chess players so much superior to computer players. I mean we have been looking at the move F3 and I was trying to explain the anticipation as a human that, oh, well I've just got to adore Kasparov's position because like you were saying not only have I broken up the team nicely, but that pawn on D4 and that pawn on F5 are so weak.
I do like the move queen D7, but what I even like more is maybe just playing the move F4, just to play the move F4. That way I cut out that bishop on D2. I will hide my king very nicely, by the way. I will play moves like king G2 and king H3 having a nice little shelter. But even more than that, what are you going to do about that H4 pawn. I'm going to play H5, H6, H7, maybe a queen and then give you mate on H7. 62. f4 Ke8.
Mr. Ashley Looks like an excellent plan. And now Kasparov, trying to make a decision on that score. And F4 would certainly do the same thing we talked about too freeze the pawn on F5. It could not move. It would be a target for attack, also in addition to this other plan that you're talking about, and then that would just win a pawn. And look at this. Fritz 4 has jumped off the scale now to give Kasparov a 2 point 19 advantage, and indeed the world champion has played the move F4, a strong move. DEEP BLUE responding immediately with king over to E8. This looks a bit desperate.
It may have had to do this, actually not a bad plan. It seems not a bad plan, the idea to stop the mating ideas, stop the mating ideas, just in case the h pawn goes down the board, queen F7, that move is no longer checkmate because the king can run away. So king has moved in addition to which the move queen D7 for White has also been stopped for the moment, queen D7, because the king would simply capture it. So a good try for the moment, but it still looks as if the computer is in big trouble. That f pawn looks dead. I mean I have to think that that f pawn is dead in some variation.
Mr. Seirawan Well, let's just do it. It looks very direct to me. Queen C8 check.
Mr. Ashley King E7 is --
Mr. Seirawan Is boxed. Queen C5 check. And again you can't pose your queen because the f pawn would be hanging. So you'd have to go back with the king. And now I'd play bishop D3. And now just simply bishop D3.
Mr. Ashley And that's scrumptious. The pawn is gone. Kasparov with amazing technique, beautiful technique, has found this position.
Mr. Seirawan If we round up and capture that F5 pawn, what White will have on the kingside is simply 3 versus zero, 3 connected passed pawns. 63. Qc8+ Ke7.
Mr. Ashley But now, finally White could potentially exchange those queens off the board, and then 3 connected passed pawns just march. So if Kasparov -- well, I was about to say if Kasparov sees this position, of course he actually does. But if he wants, he can get this position, and maybe something else. But Kasparov has really worked hard and has played a fine game, a superb game. And as much as we may have second-guessed him on a couple of turns, he shows why he is world champion, and why he understands the game better than anyone else on the planet. And it looks as if he's really pressing his advantage home. But we will not call the game as over just yet. The computer may be a better opponent than anyone else on the planet because it knows quickly when it's losing the game, and it will try and put the best defense up. But I don't think it's going to be able to solve these problems. And Kasparov has indeed played queen to C8 check the first move of this combination, queen C8 check and now we are anticipating a queen move, that queen blocks and the queen takes F5 immediately. So king E7 looks like the only move. We've seen DEEP BLUE think on occasion on only moves before, so don't be surprised if it thinks now. But what a great game. Kasparov, those light squares. This is like a model game.
You could study this game over and over and over again and get new truths from the game. I think the computer, you're right, though, did not bring about the best defensive ideas. The position as a king 8 H6 seemed much tougher. But that's interesting because we said the same thing about Kasparov yesterday.
Mr. Seirawan Yes.
Mr. Ashley You know, you can't win a chess game unless the other guy makes a mistake. So, you know, the thing made a mistake. There's no argument about that. It's not perfect, thank goodness.
And we have a question in the front.
(Audience commentary.)
Mr. Ashley I think this spectator with a very, very strong point. And that's, I think Fred you can make a point about this, the idea of not giving counterplay to the computer, to any computer because that's always problematic.
MR. FREIDEL Yesterday evening Garry said to me very late at night what we showed you was how you shouldn't play against the computer and today he's going to show you how you should play against the computer and apparently he's done it.
Mr. Ashley Well, there you have it. The word is prophylaxis, isn't it? It's a very common word in chess circles. I must admit that my vocabulary expanded as a result of playing the game of chess. All these great terms, prophylaxis and all this stuff. 64. Qc5+ Kd8.
Mr. Seirawan Prophylaxis is simply the art of strengthening your position in a subtle way and yet at the same time robbing your opponent of his activity and his activity pieces. And that's exactly what Garry has done and it's exactly what the gentleman has pointed out, is that what Garry has done in a very nice, benign way, is simply leave the computer without any clear source of counterplay. And Garry's last move, queen C5 check, is going exactly in the line that we think. And I just see that while I'm looking at Fritz's graph, it seems to be growing and growing and growing, kind of what happened last time -- yesterday, we see Fritz now believing that Garry has a greater than 2 pawn advantage. I certainly think he has a big advantage. So I don't -- I very -- when I sit there and play a game of chess, by the way, I don't evaluate a game and go boy isn't this wonderful, I have a point 67 advantage. No, I say boy, isn't this wonderful, I have a clear advantage or I really like my position and I think in those terms as opposed to doing crisp scientific way of thinking.
Mr. Ashley Well, DEEP BLUE has responded with the move king to D8 and now it seems a simple shot, bishop to D3 will win A pawn by force and Fritz -- 65. Bd3 Be3 66. Qxf5 Qc6. and he has played it indeed. Kasparov with his now bishop to D3 about to win yet another pawn and go 2 pawns up in this end game and that should just be an academic win. And this really looks like a world champion game to me. It looks like a game of the very highest class of players, 2 of the top players in the world playing. You would not be surprised to think that this was -- and I had the same feeling yesterday, you know, that White -- when White played knight B7 the computer, when DEEP BLUE yesterday, the last move after bishop onto D3 was bishop to E3, and allowed Kasparov to attack, that right there you might think is a cold-blooded computer move. But we also said that's a Kortchnoi move. Victor Kortchnoi, famous, 1 of the best grandmasters of all time. He's famous for taking your pawn if he doesn't see what you're going to do to him. He doesn't clearly see what you're going to do to him. So we started even giving the computer styles of the great players. There was 1 famous game between Capablanca and Marshall where Marshall first initiated the royal attack with Ruy Lopez and Capablanca beat back the pawn and won the game. And the computer has done this. It really looks like a human almost in some of its moves. You really couldn't tell the computer was playing the game on many occasions. Kasparov has now played queen takes F5 and is quite happy to exchange queens off in this position. 3 connected passed pawns will win the game without effort.
Mr. Seirawan I was just going to say that I couldn't help but smile. Garry when he absolutely full -- with full knowledge knows he's going to win, he'll take his piece, he'll snap off and capture the opponent's piece. But while doing the snapping he is he'll kind of grind his move into the board as just a reminder that there is it is, buddy, that's the winter.
Mr. Ashley That is that's typical Garry.
Mr. Seirawan That is Garry. And what we've been talking about all along is the bishops of opposite color. It's something that Garry should absolutely avoid, was trading queens, but he has managed to bring about a situation where he has 3 connected passed pawns on the king side, so that an exchange of queens would have been perfectly okay. Garry could have won the ending very easily.
Mr. Ashley Well look at this, his last move, queen to C6 by DEEP BLUE, look to go penetrate into the H1 square. In fact I think this is a mate threat. Finally. The computer hoping that Kasparov will play some lazy move. Let me be really dumb and play H5. And queen's H1 check, king E2, and after queen G2 check, Kasparov is being mated. 67. Qf8+ Kc7.
Mr. Ashley 2 moves. King E2, that was not nice. So the computer seeing its chance right here with the move queen to C6. But there are many moves for White in a position like this, many moves. I mean we can check if we want to. We could also play bishop E4. I think also -- I think queen C1 check is not dangerous at all because after king to G2 we can go to H3 and up to G4 and the White king is safe in front of its pawns. So a safe move is bishop E4, definitely.
Mr. Seirawan What I would like to do if I was Garry because what he's already proposed to do is to exchange queens. So if I was in Garry's shoes what I'd like to do is do a series of queen checks that will result in an exchange of queens. 68. Qe7+ Kc8. .
Mr. Ashley And he has begun with the move queen to F8 check, forcing some kind of king move, either king to D7 or -- well, king D7 would allow bishop F5 check. So DEEP BLUE has played king to C7. And now Kasparov has played queen E7 check.
Mr. Seirawan Which Garry has done. And the idea is if we can achieve a position like king B7, queen B4, check, king C7, I got what I want, which is an exchange of queens. So this is what DEEP BLUE is now having to avoid. 1 idea for DEEP BLUE might be to play after king B6, queen B4 check, is to drop back to the a file. And let's see if we can --
Mr. Ashley But now queen A5 check and queen B5 does the same thing.
Mr. Seirawan Exactly. Maurice points out quite economically that I get what I want, an exchange of queens. So we've got to go back a bit.
Mr. Ashley How does Black avoid the exchange of queens and then have to face this rush of pawns? It doesn't look possible.
Mr. Seirawan Well, let's see.
Mr. Ashley What about king C8? Is that possible?
Mr. Seirawan Virtually the only other possibility is king C8. Now we can look at -- well --
Mr. Ashley And it has in fact tried its best to avoid the exchange of queens because that's going to result in a losing ending. Maybe bishop E5 is a thought, maybe bishop E4 is a thought.
Mr. Seirawan Well this bishop F5 check certainly looks tempting it. Virtually the only move is king B8. Now we have a move like queen E5 check.
Mr. Ashley King A7. King A8, no, king A8 no. Bishop E4 would be deadly.
Mr. Seirawan So king A7. And now how can I put on the finishing touches?
Mr. Ashley It's not easy, is it? It doesn't seem as if it's forced, the exchange of queens. It looks like you're going to have to use a move like bishop E4 sooner or later because Black is trying its level best to avoid the exchange of queens. It's so funny that for so many moves you said no trade, no trade, no trade, right, and Black is thinking trade, trade, trade, trade, trade, and now White is going I really do want to trade. Come on, let's trade. No, no, no, that was a long time ago. I've changed my mind, buddy.
(Audience commentary.)
Mr. Ashley Bishop F5 check. In bishop F5 check, king -- yeah, king B8. Queen to D8 check, king A7. There goes a way. After king A7 in this position, queen D7 does trade queens.
Mr. Seirawan Excellent.
Mr. Ashley Which is an excellent maneuver and does work. That in fact does work. So it looks as if the variation is forced.
Mr. Seirawan So we think that we now have it.
Mr. Ashley Which 1 do you want?
Mr. Seirawan I want to go back to the game position, position at hand. 69. Bf5+ Kb8.
Mr. Seirawan So he sees it. Bishop F5 check has been played by Garry, forcing B8. Now comes the nice move. 70. Qd8+ Kb7.
Mr. Ashley Queen D8.
Mr. Seirawan Queen to D8 check, forcing once again a move like king B7 or king A7 and then to both we have the answer queen B7, and I think -- well, Hans isn't here, but I think --
Mr. Ashley Oh, there he is.
Mr. Seirawan What do you think about that? D8.
Mr. Ashley Here it comes queen D8 check.
Mr. Seirawan Hans, would it be time to resign?
Mr. Ashley No. Hans's computer needs to be down a rook.
(Audience commentary.)
Mr. Ashley Hans Berliner says they don't believe in resigning. That's the first commandment in the Ten Commandments of computer chips I shall not resign.
Mr. Seirawan Well, it certainly is a truism that no one has ever won a game of chess by first resigning. 71. Qd7+ Qxd7.
Mr. Ashley And now after after this move king to B7, Kasparov has played queen to D7 check, absolutely forcing the exchange of queens and this position looks like toast.
Mr. Seirawan Well, 1 of the things I'm not looking forward to is seeing a silicon chip cheerleader. I like the human cheer leaders. They're looking good.
(Audience commentary.)
Mr. Ashley Fritz is not excited about this position, and I am in shock. 72. Bxd7 Kc7. 73. Bb5 DEEP BLUE RESIGNS.
Mr. Seirawan Well, I think that Fritz doesn't realize that this particular ending is so absolutely easily won. And I'm sure that the moves bishop D4 or something else is what Fritz would have centered on. But from my perspective as a human being I sit there and I say listen, all I have to do is win the game. All I have to do is win the game. If I see a winning ending, I don't have to calculate anything else. It wins, it's over, that's it, I don't have to be brilliant anymore. The end game is winning. So for Garry, he's absolutely certain this ending is winning.
Mr. Ashley Kasparov knows full well he would win this position in his sleep. Against a human opponent, Kasparov would be pulling some faces like you would not believe.
Mr. Seirawan Well, now he will start to put his watch on.
Mr. Ashley He's saying come on buddy I got a dinner date. Let's finish this now before I really start to beat you up.
Mr. Ashley He knows full well that the computer plays on and there's nothing in its program to tell it to resign at the moment because there is it's openly 2 pawns down.
So king to C7. Kasparov, I tell you if you put 1 minute on Kasparov's clock he would win this position. He could make the moves instantly. And look at him now. He's hovering over the bishop. He's eager to move. I can do anything.
He played a move bishop to B5. And the watch is back on the wrist folks. The watch is back on the wrist. That's your man Garry.
Mr. Seirawan Now what his idea is will be a very simple 1. For example Black will play king D6. He'll put his bishop on D3. Black will try to put his king over to make some defense. All White will do is set up his positions with pawns on F4 G5 and H6. He'll do that very straightforwardly. He will just play H5. There is virtually nothing the computer can do. And after he reaches a position basically like this, I think then the computer will calculate it's like 4 or 6 or 7 pawns down and then it will give up. But what the computer doesn't understand is that when it calculates it's only 2 pawns down, this position that's on the board now is absolutely inevitable. This is going to happen. And there's nothing that the computer can do to stop it.
Mr. Ashley That is the position with the pawns on F4 -- sorry, F4, G6 and H5.
Mr. Seirawan It's unstoppable. So I as a human being say he can't stop me from having this position. It's an easy win.
Mr. Ashley And in fact they have resigned. Garry Kasparov has won the game in fabulous style, clear cut play, absolutely a world championship performance. What a magnificent game. I'm sure someone will point out to Garry that we're eagerly awaiting him to make an appearance.
And he's showing the operator now just why he was losing, pointing out some points there. I'm not sure Murray Campbell even plays chess. No, he does, actually. And really a spectacular performance by Kasparov, just controlled chess all the way through. And he really showed why and how you beat a computer. So fabulous indeed. And now the match tied at 1, 1 with a game coming up on Tuesday. So the match has really gotten some interest now, tied at 11. What do you think though about Kasparov's chances with the Black pieces coming on Tuesday, Yasser?
Mr. Seirawan Well, just to address that, first of all, it is it's really nice. I was sweating cannonballs. Tomorrow I have to go on the McNeill-Lehrer show with C.J. Tan and I just had this horrible, horrible feeling I was going to go on the show talking about how much superior the human being was and then being behind in the match. So thank goodness that Garry saved me at least that. So that's a big relief. I think that for Garry the first 2 games were crucial.
Mr. Ashley A very happy Garry Kasparov, by the way. We see him.
Mr. Seirawan He needed to know what the computer was capable of. He needed to see the computer on offense and defense. And I think that for the rest of the match Garry has got it. That's it. That's it.
Mr. Ashley You know, humans, we're so predictable. Yesterday we were thinking oh, God the guy is dead, poor Garry. And now he wins a game, yea, he's on top of the world now. He's going to win, no problem. Well, it did look very convincing. If Kasparov keeps this kind of idea, the idea of prophylaxis, don't give the computer any chances, make the position simple, keep it under control, then he's looking good. He's really looking good. And I think he's well aware of that. He's learned from his mistake, which is a fabulous, fabulous human characteristic to learn from our mistakes and he learned from his mistakes in yesterday's game and now he's played this game wonderfully, wonderfully. What technique. And I think this is really going to go down as 1 of the great games in chess because of his marvelous technique when using a very small chessboard.
Mr. Seirawan Exactly. And I just would add that that's one of the things that I said to start -- at the start of the match that, this was going to be a very, very legitimate match.
(Standing ovation.)
Mr. Ashley We were worried about you after yesterday.
Mr. Kasparov Yeah. Yesterday was not a good day. And first I have to congratulate the IBM team for a tremendous, tremendous job they've just done. What I discovered yesterday probably is now clear to everyone. Now for the first time we see the computer at chess and quantity becomes quality because the number of the moves this monstrous machine can play in fact prevents it from making bad positional mistake within reach of its calculation. And yesterday I think the move D5 and B3 that was so human but what I realized that for machine it was simple because it never lost A pawn within the tree of its calculations. Now first time we are planning not with just computer but with something that has its own intelligence because the depth of its calculation gives certain positions understanding. Even as you see today, as we saw today, machines don't understand many things. But only if it goes beyond the depth of its calculation. But yesterday I was really amazed by the degree of the machine and at the end I tried to very hard that the attack would succeed against any human and that's exactly what I told at any press conference. You have threats the machine comes. But if there are no threats the machine doesn't care and will continue. It's business. And maybe I could still defend better yesterday without playing at 1 point play rook D8 and keep the knight on D5. But what's happened after opening yesterday it was the worst situation you can imagine in this match. The number of pieces was limited. And then computer had the queen. This is, you know, unbelievable, queen and knight in computer hands. It's unbeatable weapon because it attacks. It finds the shortest cut to any weakness in your position and it just doesn't hesitate, doesn't have any doubts, it's not scared by your illusory threats. And that's why it was absolutely worse. And it was a massacre which was well deserved because I probably had to get better position at the opening. But I would have to believe that machine was doing so many good moves, destroying my moves. That's a real masterpiece. But today in fact my idea was just to play, just to recover because it's difficult to sleep after such a game. And that's a real consequence that I faced an opponent that has no emotions whatsoever. And in fact I just have a few ideas I believe in the future that will have to be taken into account because I just realized that in this match it's a handicapped match because computer has much more time. It can't think while I'm thinking. And at the same time you know human brains need some rest. Now probably the time should be corrected a bit --
Mr. Ashley Garry Kasparov asking for a handicap?
Mr. Kasparov It's a handicap today because the machine can calculate. It eventually uses more time than you do because it goes on and on and on. And we don't have such a problem, electrical cable. Or maybe you should switch the machine while I'm thinking. And also I think there's a very important question, you know, which bothers me a bit. You know, when you ask mathematician to solve a problem, theoretical problem, you give him a calculator. He has something to use the knowledge that was collected by the previous mathematicians. Now, we cannot compete with computers memory. I mean our memory is not excellent. It's so tiny.
Now obviously we had problems preparing for the game. The question is why we do not have an access to an opening data base or ending data base. That is something the computer has not created. Now, it has a big advantage because it has unlimited memory. And that's why for the future matches I believe, you know, that we have to give certain advantages or certain access to human being just to equalize its chances to see how man plays against machine but not how machine plays the man having all the knowledge of us, the chess players.
Mr. Ashley Garry, you defeated the computer today in magnificent human style.
Mr. Kasparov Yeah. But, first of all, it took only 6 hours. And second, you know, it was very funny game because I -- first of all, I couldn't --
Mr. Seirawan Let's take a look.
Mr. Kasparov I couldn't figure out what was happening in my game. In '85 I played a match and that was a game that won the match. But I couldn't figure out what happened in that game because I wish I had an access to a data base to look what I did in '85 when I played. I think the crucial moment happened when Black played E5 queen B8. And suddenly I realized it didn't care about White's weaknesses.
Now probably instead of B4, I mean definitely computer didn't take on B2, the pawn on B2 because, you know, then I wanted to play queen B8. Of course, it didn't take on B2 because then it's simply exchange and queen comes on B4 after exchange. Again the damage was within its calculation. That's why, you know, I didn't think it would do it.
Mr. Seirawan But I think that was very impressive of the computer to recognize that it's busted.
Mr. Kasparov After queen D7 it's losing material. And of course I could play B3. White is still slightly better because it's difficult for Black to play central pawns. But then I just wanted to check whether or not B4 is -- you know, whether the computer would take on D4 or not, you know, how good is positionally. Now we took on B4 without hesitation.
Mr. Seirawan Yes, it was tempted.
Mr. Kasparov It was tempted because again there was no damage, no real damage. Unfortunately now the threats are so old the computer had to give up A pawn. What I realized, just look at this -- it was working. I spent about 5 minutes to move queen D7 and it played queen C8 almost instantly. Then computer made another bad move rook B8 queen A4 and the bishop C3. Bishop C3 is a terrible move. Of course bishop D6 was a terrible move. Just having bishop on B8. Still White had an advantage. But after bishop C3 I think the advantage was almost decisive because now bishop C3 takes B8, queen C7. And here again I made a very human mistake, you know. I saw rook B1, and I would get similar end game with A pawn which is probably winning technically. But, you know, normally when you -- in our mind we play H6 because pawn is hanging, and then I play queen F3, queen should go H6, queen F3, queen should go on D7 or whatever it is, and then who comes on B1 with a decisive effect. And that's a decisive effect because we have the same position. But even rook is playing and Black, of course, is absolutely defensive. When I played queen D3, I mean I just saw -- and I think this position okay. This position should be drawn. White has some tiny chances, tiny chances. But obviously it's a draw because I don't think White has a serious -- you know, can make serious progress against a human player. And it's okay. Now I took pawn H7. What else I could do? Rook B2.
Mr. Ashley This position, someone suggested the move bishop to B1. Did that impress you at all as a chance?
Mr. Seirawan I was just wondering if you thought about it. Somebody in the audience had suggested it.
Mr. Kasparov You know, maybe it's a good idea, but what I calculate, for instance, if I played bishop B1, then put my queen on E4 after queen E6, for instance. Now it plays F5, and I understand that the end game is dead draw. But what happened later in this game, it took about 15 minutes for computer to agree to give up the pawn. It didn't want to give up A pawn, but probably it saw, you know, so many threats that it eventually gave up A pawn. Now for the human being it would take maybe a few seconds to give up A pawn because the end game would be dead draw. I hold that something could happen because clearly White has only 1 chance in this position. It's to bring Black's king to the center. Because then in the center then we can create certain threats.
Now, my hope was that for computer it should still be dead draw. And you know my plan worked for computer moved king, you know, to the center. And unfortunately once I repeated -- the second time I repeated the moves and, you know, computer won more time. Here, if I didn't repeat the moves, then computer had to solve the problem with sacrificing the pawn before time control. And I was not sure if it would do it. But anyway I still think Black had plenty of resources.
Now, it requires serious analysis when you know White could do something or Black could do something. I think the critical moment, the really critical moment, because I just outmaneuvered eventually, at that point the computer realized -- it didn't realize it. Probably calculated. I just felt it and it knew. It was really tough because the king was in the center and the reason that kept me in the game, I said, my God, if I play Black and computer won, it would kill me because it was queen and chance to attack and you know I decided fine. Let's try this. I think the mistake computer made was move queen D7 when I just won the pawn and Black got my bishop on E5. Here of course king D7 was a terrible move. King E7. Because after king D7 my bishop comes back. Instead of H5 it comes to C4, and now I just -- I realize because I could restore coronation of my pieces and I already saw the plan, you know, with 2 pieces -- my piece I paralyze its queen and I have to find a plan. Obviously Black cannot move pawns because with mate and threats I always have checks and I will take certain pawns. And you know after king E7 I think it was very difficult for White, very difficult to win eventually, if it's winning at all.
Mr. Seirawan I think that also the problem for the computer is it didn't realize its best defensive idea was to stay close to the king side.
Mr. Kasparov Exactly. You know at that time you know it's -- even if you run hundred millions a second, it doesn't help much because the number of pawns is big. That's why as soon as you get too many pawns off the board, you can calculate positions to the very end. It's really big. If computer sees the defensive line which would be proved better if human point of view, but it doesn't look better within its tree. It goes to another 1, which better defense, but longer chances. Again probably the same happened in the game because it just -- the human being would play at 1 point D3 or F4, would do something dramatic and machine just waited. Because probably it saw it was losing. It was losing by force. You know, if it tried to push any pawn, but again the human being would do something because, otherwise, if you sit and wait --
Mr. Seirawan Garry, I think you played an absolutely gem of a game and for me I always think that the player becomes very good when it loses well. The computer lost a good game. I mean you had to play very good. And that makes to my mind the computer a very legitimate opponent.
Did you expect before the match began that you would lose a single game against this computer at classical chess?
Mr. Kasparov No. I could imagine it was a game. But clearly I did not expect such a strength. As I said in the beginning, you know, we are facing now a new quality. And also we are all playing with little computers. It's on our laptops. And you know this morning I just realized that it probably works against you because you're used to certain mistakes that computer always makes. You play with the strongest chess playing programs on your laptop. Very good. But you know this computer is thousand times faster. I mean it's just -- it makes -- within 1 second it looks to the lines that the chess playing program that chess playing genius would spend ten to fifteen minutes. It's like you are trying, you know, just to play a five-year old kid and suddenly the big daddy comes. And it really creates a new, new challenge because first time you know we have to deal with something that understands chess. Not because it understands but because it sees too deep. The normal mistakes you see in the computer, just forget about them. Within its calculations you can hardly expect a different mistake.
(Audience commentary.)
Mr. Kasparov You know, human stamina counts as well, you know. I had the chance. I had a very good chance to win the game. I definitely made a couple of mistakes. I missed my best chance when the position was technical winning. But I had a chance, and this morning I realized that there would be not so many chances in this match. To be at this machine, you have to try extremely hard. You expect a mistake and you win technically. To win the game, you have to fight as many hours as it takes. If you have a chance, you have to go for it, and I was lucky, but I pressed my luck hard.
(Audience commentary.)
Mr. Ashley The question is did you know that the computer lost access to the opening book, or did the telephone go out?
Mr. Kasparov There was a wrong move C takes D4. I was so happy. And then, you know, it is quite annoying because when it just happens, I spent a minute or so. Your mind is working in other directions, and suddenly it is a mistake now. You come back to normal. Then I just didn't realize.
When it castled, I don't know whether it lost access or not, but I wouldn't tolerate a second change of the move because I spent already ten minutes calculating something because it is very difficult for me to come back, but it had a pretty normal at game.
But, again, I understand it is not easy for computer to anticipate or to evaluate the line which was not prepared for the game because there are many lines and from the computer's logic, highest logic, that we don't understand, maybe the position is not good. It has some other ideas than we have.
(Audience commentary.)
Mr. Seirawan The question is we have seen the first of a new generation of computers and the future is going to offer even stronger than what we are seeing, and your reaction, Garry.
Mr. Kasparov It will be very tough. Maybe it will be over one day, but as I said, I believe that for the fair match we will need an access to the data base during the game. It helps us from preparing, hours and hours preparing. You can come fresh and play the best chess you can.
No, I think that any match with super computer like that, it is as serious as the world championship because you have a real opponent, which in certain positions we have to admit it. And it is not one percent. In 10 or 20 or 30 percent of the positions it plays much better than we play, and in some positions it plays like God because it simply does not make mistakes
(Audience commentary)
Mr. Seirawan Just in case everybody did not hear the question, the question was Garry took about 20 minutes, I think, Garry, 25 minutes on the move knight C6 when, boy, there was a wide variety of choices there.
Mr. Kasparov Can you make another offer?
(Audience commentary.)
PARTICIPANT My question is after move 40 there was a problem with the computer. Do you know what happened there? Can you tell us about it?
Mr. Kasparov I think the main problem was computer was looking so deep to find out it had to sacrifice a point. Because for computer it is very painful decision because there was no -- the weaker machine that didn't make a hundred moves per second would never sacrifice a move. You can take maybe eight or ten moves, but it would create stress and only this machine to see the stress. It took some time, you know, and then it decided to give up a point, which is a very difficult decision. But I think that was the main problem. The machine was figuring out that it had to give up a point, and it didn't like it.
PARTICIPANT Are you saying that it locked out kind of?
Mr. Kasparov You will have to ask the programmers. It definitely had some problems because presumably it was that draw from machine's point of view. And that's when it happens. It takes a lot of time for the machine to figure out that it had to give up a point because otherwise it was getting worse.
PARTICIPANT Somewhere around 60 or so you put your watch back on. Did you know game won at that point?
Mr. Kasparov At that time I was quite happy. Time to go home. In fact, I was pretty sure that I'm winning when I put my bishop on C4. I saw that idea clearly, F4, you know, because that is crucial. With F4 the Black King is trapped. And when I saw it, just I saw the whole construction, and then what it took for me to place a piece on the correct squares is not to miss any tactics, but I know it's over.
PARTICIPANT What do you expect the outcome as far as the score of the match will be?
Mr. Kasparov It will be very tough. I expect it will be very big fight because I clearly see the advantage of the machine. It doesn't make automatic mistakes, mistakes by hand that we very often do, and, you know, if it gets its shot, you can forget about me in that game. That's why I have to consider seriously how to play further because these next four games will be very tough.
Mr. Ashley We would like to thank world champion, Garry Kasparov.
(Proceedings concluded at 915 p.m.)
[ The Game | The Program | Commentary | Grandstand | Deep Blue ]
[ IBM home page | (C) | (TM) | IBM Small Planet Pavilion | Internet 1996 World Expo ]