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
February 17, 1996
GAME 6
KASPAROV LEADS 3-2
White Garry Kasparov
Black Deep Blue
MR. NEWBORN I'd like to welcome everybody here to the DEEP BLUE versus Garry Kasparov ACM Chess Challenge. This is game number 6. We're probably not much more than five to ten minutes from starting, I think. The game will start at 3 o'clock sharp. It will probably last between 3 and 5 hours. We will have with us for commentators on the game 2 of America's greatest sportscasters.
Mr. Seirawan Chess sportscasters.
MR. NEWBORN Yasser Seirawan, who is on my right. Yasser himself has been 3 times U.S. open chess champ and remains 1 of the top players in the United States for 20 years now.
Mr. Seirawan 5.
MR. NEWBORN At his side will be Maurice Ashley, who has yet to arrive, but will probably come in in a couple minutes. Maurice has almost reached the grandmaster stage, the stage of approximately 30 or 40 chess players in the United States, and he will assist Yasser in the commentary, which I think you guys will really all enjoy. They turn this into a real sporting contest. And chess is a sporting contest except you're not moving the body around.
Mr. Seirawan Using the mind.
MR. NEWBORN Hope you all enjoy it. I'm going to turn the mike over to Yasser in 1 second, and you should feel free to ask questions of him. He will invite other people up here from time to time to commentate, and I hope you all have a good time.
Mr. Seirawan Thank you, Monty.
(Applause.)
Mr. Seirawan Once again, my name is Yasser Seirawan. I'm an international grandmaster of chess, and I had wanted to try to put the event today in a little perspective prior to its beginning. This is an historic first. This is the first time in the history of chess that you have a range world champion, Garry Kasparov, competing against a computer in a world championship regulation match style setting.
The event is a 6 game match. We have played 5 games thus far. Garry Kasparov has won 2 games. He won games 2 and games 5. He lost the first game, giving pause to people such as myself who felt that Garry was going to win the match decisively. Games 3 and 4 were both drawn. So the match currently stands in Garry's favor 3 to 2. Today is the sixth and deciding game.
Monty just informed me that the computer has some kind of indigestion. Perhaps it has eaten too many chips. Weak weak. Thank you. And the match is supposed to begin at 3 o'clock and we have a couple minutes before 3 and you've been seeing all of the people there in the room as they're hovering around the computer wondering what they can do to bring it up to speed. Again, in a regulation world championship time control, the time control is 40 moves in 2 hours for each side. So that is to say the computer will get 2 hours to consider its first 40 moves as will Garry Kasparov. And of course they're welcome to analyze on 1 another's time.
The computer has been playing very, very quickly, has very rarely gotten into anytime trouble, whereas that's not the case with Garry. In game 4 he was really in a great deal of trouble both on the board and on the clock and he just barely managed to save the game.
If the game isn't over after 40 moves, they play a secondary time control which is 20 in 1. That is to say, after Garry makes his 40 move, he will get a whole new hour to make his next 20. If the game isn't over after 60 moves, then each side will get a half an hour for the rest of the game. So the maximum the game can last is 7 hours.
A couple of points about the world of chess. I was going to help here using the screen directly behind myself to show a little bit about the world of chess, but somebody stole my mouse. Being mouseless, I cannot assist you in learning a bit about the world of chess. But 1 thing I can explain is perhaps Garry's position in the chess world.
Before I do that, I would like to take this opportunity to introduce my colleague, international master Maurice Ashley. Maurice has worked with children in the Harlem New York school district. He's a great chess coach and a great chess commentator, and it's my sincere pleasure to have Maurice working with me today.
(Applause.)
Mr. Seirawan We have a bit of a problem. The computer has indigestion.
Mr. Ashley Oh, boy.
Mr. Seirawan So there's going to be a delay. We don't know how long the delay will last. Okay, it shouldn't be too long. I just point out that in world chess championship style events the clock always begins on time. And even if my opponent isn't here, they start his clock. For example, in game 5 Garry was a few minutes late, but promptly at 3 they began his clock. So I'm not quite sure why the computer is getting some special considerations for not being able to play, but --
Mr. Ashley It's not quite 3 yet, I think. That will do it.
Mr. Seirawan I'm fast.
(Audience commentary.)
Mr. Seirawan Indeed. The question was, isn't there a rule that they have to start at 3 and I do believe that they have to start at 3. And again I'm not quite sure why they're getting the allowance time, but again, as you saw there, there was a lot of people in the room.
Today and throughout the commentary we will be inviting experts in the field of computer science. We'll borrow, drag some of the IBM crew to come here and speak about the computer and what they've managed to create. As most of you probably are aware, IBM has made a tremendous breakthrough in both the hardware and software of computer chess programming and the DEEP BLUE monster sees 200 million positions a second. And I've been told that in the space of a mere 3 minutes, the computer can calculate 50 billion positions. Meaning it has no chance. It's up against Garry Kasparov.
Mr. Ashley Well, Garry has been known to calculate a position or 2 in his time. But the point we make all the time is although the computer can sometimes calculate deeper than Garry Kasparov, it doesn't necessarily calculate better than Garry Kasparov in that he's able to get to the mate of the position with his great experiences, great intuition, his knowledge of the game. Immediately he knows what couple of moves to look at and then start calculating those options instead of looking at maybe 40 different possibilities in the position and find that 38 of them are completely random and you don't need to be looking at them. DEEP BLUE, the better part of those 50 billion positions would I say that 40 billion -- 40 billion 990 million of them are simply irrelevant and so go ahead and look at them, but they're not going to get you anywhere because they're simply not the best variations to be looking at. So Garry Kasparov -- it sounds almost -- you know, totally daunting to the human to have to suffer that possibility as Garry Kasparov now walks into the hall.
(Applause.)
Mr. Ashley But just finishing up, it's not a big deal for Garry and has Garry has shown over the last 5 games, he's more than a match for DEEP BLUE's amazing analysis -- analytical ability.
Mr. Seirawan That was 1 of the things that Garry was most concerned about before the start of the match is he never had gotten an opportunity to analyze any of DEEP BLUE's style of play or games, if you will, because they had never been published, whereas conversely DEEP BLUE had a wonderful data base of well over a thousand of Garry's games. So you can be sure that the computer had every known Kasparov game plugged into its data bases. And we have begun. 1. Nf3 d5.
2. d4 c6.
3. c4 e6.
Mr. Ashley The clock has started. Garry Kasparov just setting himself up and now beginning with the move knight F3.
Mr. Seirawan Garry has repeated this same opening move in all the games he's been White, games 2 and 4 and game 6 now. And the computer responded immediately as it has in all 3 times it's been White with the move D5.
Mr. Ashley Indeed, the preparation for DEEP BLUE is very good. It has a specialist, grandmaster Joel Benjamin, who is really an expert among grandmasters of playing against computers and he has put in the opening book basically for DEEP BLUE helping the team of computer specialists who are not necessarily strong chess players, helping them to come up with a nice opening strategy against a guy like Garry Kasparov. So the moves have proceeded along quickly and repetition basically of the same 3 moves of the other games. The game started off knight F3, D5, after Garry Kasparov played D4 the quick response was C6 by DEEP BLUE. Garry Kasparov instantly played the move C4 and again as in the prior game --
Mr. Seirawan Game 4. 4. Nbd2 Nf6
Mr. Ashley Game 4, E6 by DEEP BLUE. And now Garry Kasparov has played knight B to D2 as he had done in game 4 which then we thought of as an interesting move, original move, a quiet move in our opinion, but he proceeded to get a very nice position, Yas, with this move knight B D2. DEEP BLUE in response continuing as it did with the move knight to F6.
Mr. Seirawan We're seeing a repeat of game 4. The opening defense for DEEP BLUE that Joel Benjamin has chosen is known as the slav defense. It's distinguished by the strong pawn in the center of the board, Black's pawn on D5, supported by the pawn on E6 and C6 as well as the knight on F6. So the central strategy in Black's opening is to maintain this wonderfully powerful pawn in the middle of the board. And what Garry did in game 4 was he worked to try to get the move E2 E4 in. Eventually he did and we know he got an advantage but later it slipped and he nearly got -- fell into a trap and lost game 4. But that was a draw.
Mr. Ashley 1 would anticipate some new move on DEEP BLUE's part because in that last game Garry seemed to be working it. He had a really nice position and I'm sure he'd be glad to get back into that same exact position if the computer were willing to do that. 5. e3 c5.
6. b3 Nc6
7. Bb2 cxd4.
And he has played E3 on move 5, E3 as he did in that game. And if you can guess in the next few moves -- well, sorry, we can't. Because DEEP BLUE has instantly responded with the move C5. In the prior game it had played bishop to D6, but here the move C5 immediately. And Garry Kasparov had already anticipated this possibility because he has also responded immediately with the move B3 and things proceeding at a wicked pace as DEEP BLUE is playing knight C6 instantly in response and Garry Kasparov playing immediately with bishop to B2 and it looks like they're playing speed chess
8. exd4 Be7.
Mr. Ashley Now DEEP BLUE playing C captures D4 and Garry playing E captures D4. So no surprises here for either side.
Mr. Seirawan Other than the speed of play because, again, the idea in a world championship regulation match is that you get an average of 3 minutes per move. These guys are using 3 seconds a move.
Mr. Ashley Bishop to E7 by DEEP BLUE, its last move.
Mr. Seirawan I'm anticipating that Garry will now play bishop D3. In the last few moves which I haven't gotten a chance to speak about, they're distinguished by these two pawns in the middle of the board by white.
The C4 and D4 pawns. This structure oftentimes dissolves into what is known as a hanging pawn structure. In case that occurs I'll speak about that in a moment.
But what these two pawns in the center of the board do give white is an advantage in in controlling the center area, the pawn on D4 controlling two squares, the C5 and E5 square, while the pawn on C4 controls the D5 and B5 square. This gives white an advantage in space. And this is very important because you can build up your pieces behind this wonderful wall of pawns and giving Garry good attacking chances.
This is distinguished, for example, black's last move was bishop E7. Notice that this bishop currently on E7 only operates on the F8, A3 diagonal. When white plays a move like bishop D3, the bishop would control the squares from B1 to H and the bishop would kind of eyeball that pawn on H.
So for the early stages you have to give white, Garry, a bit of a plus.
Mr. Ashley I'm somewhat surprised, aren't you, Yas, that Garry decided to play so quickly in the opening. I understand that he knows the position and he's comfortable with the position he's going but it seems almost a bit of bravado on his part to say you guys didn't surprise me at all with this variation. I, in fact, anticipated it. And I'm wondering if that was evenness, because after all he's not playing against the DEEP BLUE team, particularly grandmaster Joel Benjamin, who has helped with the announcements of the openings but he's playing against the computer and the computer simply doesn't care how fast he moves because it's not going to be intimidated by it.
Mr. Seirawan But Garry does. And the reality in this match is because the computer calculates so deeply, so well, virtually -- and indeed, all of Garry's moves are anticipated. So it's almost like the computer has a response ready.
But for Garry we've been seeing him getting into time trouble fairly consistently after the opening is done and we go into the late phases of the middle game. So Garry has to nurse his clock, do a little bit better with his time management and I'm sure he just wanted to get off the early moves. And so that's why he played them so rapidly as we've seen.
Mr. Ashley Well, just to recap for those of you who have just joined us and also since the game is going live over the Internet, we would like to just do a brief recap so that in case we missed any moves nobody -- everybody will be up to speed. So why don't you just walk us through the opening as those moves happened so quickly anyway we didn't get a chance to say anything about them.
Mr. Seirawan As in games 2, 4 and 6 Garry opened up with the same opening move knight F3. This is a kind of a foxy opening move. It's a noncommittal opening. The idea is that the game may take on an English formation or a queen/pawn formation or a king/pawn formation.
The computer quickly responded D5. Garry D4. So now we have a queen/pawn opening. The computer, C6, which is a move that gives this opening a Slav reputation, the move C6 is one of the oldest defenses in chess. Kasparov, C4, E6, a little bit of subtlety on the computer's part, having not developed its pieces but just its pawns in the center of the board. Kasparov used this move in game 4.
It's not the most active move, knight on B2. It locks in that bishop on C1 and it will cause white a little discomfort in developing it. Immediately came the move knight F6, E3 and now the surprise, the move C5. This move represents a loss of tempo.
What the openings are all about is developing your pawns and pieces as quickly as possible and getting your king into a safe position.
So the computer spent two moves. It played first its C pawn to C6. Then it played C5. So this represents a loss of a tempo, giving white an opportunity to gain development.
Kasparov B3. The computer, knight C6, bishop B2. We have a fianchetto from white, the bishop on B2 very nicely aiming itself in the center of the board. Then we had a resolution of the center pawns, C takes D4, bishop E7. 9. Rc1 0-0.
And now Garry has played instead of the move bishop D3 rook C1, developing his rook. So this is the position on the board and Garry spent a whole 8 minutes to get it, whereas the computer only a minute and a half.
Mr. Ashley But certainly Kasparov had to be wondering about the preparation for this particular position. It was clear that DEEP BLUE had been well primed for this position and no reason for him to keep moving very quickly and I think this last move rook to C1, considering the computer has used all of a minute thinking about the move so far, was probably a new move, a move that had not been considered, a move like bishop to D3 was eminently natural.
And instead of doing that, which may have been met with an instantaneous response on DEEP BLUE's part, Kasparov has played the move rook to C1 and now DEEP BLUE seems it's on its own thinking about moves for the first time.
Mr. Seirawan Exactly. Let me just speak to the move rook C1. Let me say right up front, I'm very, very biased in favor of the human player. I want to see the human player win. So if I speak so often on all the possibilities that Garry has, please forgive me because that's what is going to happen.
Mr. Seirawan The idea of the move rook C1 is to bring the rook into the game. And sometimes opening up the C file with the move C takes D5, the rook would be beautifully positioned. Also, a strategy behind the move rook C1 -- and this is very clear, that what Garry Kasparov has been able to do throughout the match is outplay the computer strategically. Not tactically, mind you, but strategically. So the move rook C1 is a prelude to a move like C5 in order to gain space on the queenside.
Another point of the move rook C1 is White has a very nice system of development by bringing his bishop to D3 and sometimes dropping the bishop back to B1. Then the rook has already developed and is not enclosed on the A1 square. So this is just a nice comfortable developing move, leaving it up to the computer to decide how it will play.
Mr. Ashley I like that last point of yours. Why don't we take rook C1 back 1 second and show instead if bishop D3 had been played. You talked about dropping the bishop back to D1. Some in the audience may not realize what point you were making. But here the move knight B4 is what you were talking about, might have been a bit of harassment.
Mr. Seirawan Exactly. Then the bishop would have been forced to retreat to B1 square, and then the rook on A1 would be kind of entombed, if you will, behind the pawn on A2 and the bishop on B1. So Garry's last move, rook C1, was in anticipation that bishop B3 -- bishop D3, bishop B1 would be coming shortly.
Mr. Ashley So in fact a very natural move by Garry Kasparov just dealing with 1 simple annoying possibility. Right now DEEP BLUE could castle, which is a very, very natural move and then bishop D3 would follow. Are you surprised at all by this particular type of position? Do you think that it's the kind of position that DEEP BLUE will play very well?
Mr. Seirawan Well, this is a key point. We all know that Garry has the lead. So the first prize is $400,000. The prize for the loser is a hundred thousand dollars. In case of a tie match, we're going to see $250,000 go to each side. So for Garry and for the computer this is $150,000 swing game. But the computer is playing its position as objectively as it can. And we were joking last night that the IBM computer programmers, knowing the need to win, should program the computer to go all out; that is to say, to raise the contempt factor in the computer's program so the computer plays all out for the win.
But, unfortunately, it's a nice story. It doesn't work. Ken Thompson of BELLE, he told me that it would be virtually impossible for the programmers at this late stage to tweak the computer and make it play all out for the win. Now, you and I as human beings know, listen, we're losing the match. So who cares if we lose 1 more game. I mean $150,000 --
Mr. Ashley A loss and a draw are the same.
Mr. Seirawan Exactly. Well, let's just take our shot and swing for the fences. But the computer obviously will not play like that but rather play what it considers to be the best moves, an objective position.
Mr. Ashley Would you consider that a weakness on the part of the computer also? A human being would switch with this idea of flexibility, we're very flexible, we could decide what to do in a given situation and we'd also show some cunning. You know, we would get the kind of position which may not be the best position but the kind of position that would be very difficult for your opponent to just make an easy draw out of.
In fact, castles has been played by DEEP BLUE after 6 minutes thought. Very surprising. 6 minutes on the very natural move as it had been put out of its book and Garry Kasparov now at the table and we're expecting a very natural bishop to D3. But again this issue of flexibility, the computer lacks flexibility. How does that affect it when it's playing a game? 10. Bd3 Bd7.
Mr. Seirawan In the sense of match strategy that's true. It more than makes up for that by the fact that it's not tense, it's not nervous, it doesn't sleep difficult at night and it's not disturbed by any noise that may occur at the table. So there are pluses and minuses there. I just wanted to say towards the position, chess, obviously when at the start of the game the game is completely in balance. There's a perfect equilibrium at the board. So you have 2 equally opposed armies facing 1 another. So in order to win at chess, what a player has to do is create an imbalance. And again from the computer's perspective and from the perspective of the DEEP BLUE team, from that point of view, they should be happy.
We do have a move bishop D3.
Mr. Ashley Garry has indeed played the move bishop to D3.
Mr. Seirawan What I mean is we've got a little bit of an imbalance already. The first imbalance that we have is the pawn structure. On the kingside White has a structure of F2 G2 H2, 3 pawns. Black has H7, G7, F7 and E6 pawns. So Black having a majority on the kingside, White having a majority on the queenside, pawns on A2 B3, C4, D4. Black A7 B7 and D5 pawns, so there's already an imbalance. And from the computer's perspective, rather from the perspective of the team, that's the only way that DEEP BLUE could win a game like today, is to create an imbalance that somehow it could manage to work in its favor.
Mr. Ashley We saw the significance of the imbalance yesterday.
Mr. Seirawan That was a beautiful example.
Mr. Ashley If you saw the game yesterday, you saw Garry Kasparov in typical Garry Kasparov style carrying out a brutal attack against the computer's king and causing it to make certain significant weakening moves. But again it was that imbalance with the pawn majorities on 2 sides, the 2 on 1 on the left side of the board versus a 4 on 3 on the right side. And Garry Kasparov made his majority work beautifully.
Mr. Seirawan Beautifully. And throw in the other dynamic relationship of the bishop versus knight. In that particular game yesterday, the computer stuck its knight in the middle of the board and allowed it to get itself all penned up, whereas Garry wheeled his bishop around and it became the more effective piece. So in the 2 imbalances yesterday Garry got a chance to use his pawn majority as well as his bishop. So so far Garry was able to play it strategically. And I'll just talk about that for a moment so we can understand a little bit how the computer thinks.
The computer uses -- woops, we do have a move.
Mr. Ashley Bishop to D7, developing on an interesting square. The bishop on D7 doesn't look so active to me. This last move seems rather passive and 1 above imagined that this bishop would have been better placed on the H1 diagonal, this long diagonal this way. But it was very difficult to execute immediately with a move like B6.
Mr. Seirawan Well, okay. I'll speak about that and I'll come back to how the computer thinks. Garry Kasparov immediately responding to the move bishop D7 with the move castles, completing 1 of the cardinal principles in chess. In the opening get a safe king. So both sides having castled --
Mr. Ashley And I'll just pause for a moment because I think Garry has realized that bishop to D7 was something of a concession on Black's part because the watch is on. And for those who haven't seen this, I've seen Garry Kasparov play almost 50 games and he always starts the game with his watch on, and when the watch comes off, it is always, always at a position when he feels it's time to get down to business. If the watch never comes off, that means that you lost trivially. The watch comes off first. When he's losing, the jacket comes off. His jacket came off only once, and that was game 1. So the watch is off, and I think that move bishop D7 was the culprit for that. 11. 0-0 Nh5.
Mr. Seirawan Absolutely. Also, we know that the watch comes on when Garry knows the game is over. So the watch is off means Garry is ready for showtime. I was just going to say that going back just 1 or 2 moves because we've got a very critical situation developing early in the game, quite surprising to me. Again we have a nice little space advantage for White because of the pawns on C4 and D4, a nice central setup for White, which gives Black's pieces a little bit of a mobility problem. Most specifically this bishop on C8. Notice that it's entrapped by the pawn on B7, as well as by that pawn on E6. So that's not an active bishop. Conversely, the counterpart on the light squares is a wonderful bishop, that bishop on D3. I had expected, and as a human player I would have played, the move B7, B7 B6, with the idea of putting that bishop, just as Maurice suggested, fianchettoing the queenside bishop. The bishop would have been far stronger on the A8 B7 diagonal.
If after B6 White had captured with the move C4 takes D5, opening up an attack with the rook and with the pawn, then there's this nice little rejoinder knight B4 attacking the bishop on D3. And Black would have been able to respond with knight on B4 takes D5.
Mr. Ashley Yeah. That was indeed a more natural way to play, with the move, as you said, B6 and putting the bishop on a lovely diagonal. It played unnaturally with the move bishop D7. And I saw your head go back a second there, Yas, after you saw this last move by DEEP BLUE, the move knight to H5.
Mr. Seirawan I'd love to see Garry Kasparov's reaction to that move.
Mr. Ashley I want to see his face. 1 of the cardinal principles of chess is that 1 does not put a knight on the side of the board because the mobility of that piece is so reduced.
Mr. Seirawan Exactly.
Mr. Ashley And DEEP BLUE playing some funky chess to say the least
Mr. Ashley The knight going to H5.
Mr. Seirawan Whoa, boy.
Mr. Ashley That would look like an absolute beginner's move if we were sitting at the chessboard. You'd be licking your chops after seeing bishop D7 and knight H5. And Kasparov now coming back to the board.
Mr. Seirawan I'd like to see his reaction. This is going to be good. He's trying to be a poker face.
Mr. Ashley He's trying to stay cool. But you know.
Mr. Seirawan There's a great deal of happiness there.
Mr. Ashley He doesn't believe that last move. And one can only anticipate the world champion will be trying to take this computer out.
Mr. Seirawan Well, there's the old expression in chess a knight on the rim is dim. And I'm sure that you yourself, if you were working with some of your students there, Maurice, and they did that there would be a quick slap on the hand, never, never, never do that kind of thing.
Mr. Ashley Oh, yeah.
Mr. Seirawan What deep DEEP BLUE's last move -- I'll talk about this for a moment because I really love Joel Benjamin's analogy. Sometimes the computer does really strange moves and he says it's almost like he's a coach of a basketball team when one of the players throws up a 30 foot shot and as a coach he's going no, no, no, no, no and suddenly it's a swish and he goes yes, great move, great move.
So I'm sure Joel's immediate reaction to the move knight H5 is, oh, my goodness what have we created. The move knight H, first of all most importantly, it cost a tempo. That is to say black has used a move to move a piece that is already developed as opposed to trying to bring a rook into the game with rook C8.
Mr. Ashley It certainly agitated Kasparov because he's doing all kinds of motions at the table as if to say this is not even rational, how do I beat it?
Mr. Seirawan Exactly. The move knight H5, what the computer must be aiming for is to try to relocate that knight right back into the center with knight F4, putting some pressure against the bishop on D3.
White could play G3 just neutralizing the move knight H5, or just allowing the knight to come to F4 by playing a move like rook E1, new move, knight to F4, and bishop back to D3. When that knight on F4 -- pardon me. Bishop back to B1. And that knight on F4 is really inconsequential. It does nothing to assist in black's development.
Mr. Ashley Well, you know it's interesting. Because if you were a human player in this position you may not be thinking that that move is so great, but at the same time it's almost like a poker player's move or it's almost like a bluff, is really what I mean to say. Because, you know, you say I'll play this move knight H5 and you know Kasparov only wants a draw.
If the draw were offered now, Kasparov's hand would fly over the table so fast he'd knock over a few pieces. Because he knows a draw wins the match. He's not going to try to prove he's the best and he's the king and I have to win this game to show that every single chess game I play I'm trying to win.
Kasparov is very professional. And professionalism means win the match. However you win it. It doesn't matter. So Kasparov is quite content with a draw in this game and he'll win the match with the draw.
Now, this move knight H5 and bishop D6 and knight H5 together, okay, bishop D7, I mean, if you play bishop D7, Kasparov might say, I know that's a weak move; I'm not going to worry about it too much. I'm just going to keep moving.
You play knight H5 now the bull is going to start -- how many red flags you going to wave in front of me? Because nobody is suppose to play that move in front of Kasparov and draw. No human anyway.
So match situation notwithstanding Kasparov is almost being lured out of that draw shell with this move, but I can't envision the computer thinking about the things that we're thinking about, like all this kind of complex strategy. It's just making a move it thinks is a good move.
But this certainly is a call to Kasparov to win the game because if he doesn't then classical chess principles will have been broken and the world champion will have done nothing about it
Mr. Seirawan Well, a lovely story. Grandmaster Victor Korchnoi, who is really one of the grand old men of chess, three time challenger to the world championship, 64 years old. Victor had a student, a trainer who was working with him, Dutch grandmaster, John Vanderveel; and John suggested a move and Victor kind of gave him a cold look and said, John, before you break the rules, you should first know them.
And this is a move just as Maurice has suggested. It violates the classical principle of controlling the center. So this move knight H5 moving away from the center, you know, Garry is going to say, I have a got to punish you for this.
Mr. Ashley It's controlling the center. It's not moving a piece twice as you mentioned, not developing another piece. You know how many you going to break with one move?
Mr. Seirawan Exactly, exactly. So here Kasparov is really giving pause and you can see him going into a deep think over this position. He very much wants -- I just know Garry. He just -- he's thinking, boy, you know, you're really asking me to kill you. Just asking.
Mr. Seirawan Asking for it, absolutely. And Garry, again, is looking very deeply into position. He wants to find some way of refuting the move directly. And that's one of the difficult things about the computer, is it's not going to make a move that has an instantaneous refutation.
By instantaneous, I mean in the next 5 moves. But it may make mistakes that will have long term weaknesses or long term problems for a long game, ten, fifteen moves later it becomes apparent.
Hey, wait a minute, back here on move ten or move 9, rather, knight H5, move 11, knight H5, I made a mistake. That only will become apparent later. So we have a situation now Garry spending some time on his clock and I'd like to talk just a bit about how the computer works.
We know that the computer looks at 200 million moves a second. And an introduction into that comment First of all, what we have here on our -- on the screen directly behind Maurice and myself is our Fritz screen. Fritz is a software computer program, I think, running on a Pentium, which will help us analyze the game.
Mr. Ashley One of the strongest we should say in the marketplace today.
Mr. Seirawan Indeed. And here on the lower right side of the graph, what you have is suggestions by Fritz. He's looking at 26 of 34 or -- this is the 29th move of 34 possibilities that it's calculating. The 0.19 that you see is its evaluation table.
We say that in chess the material count, a pawn is worth one point. Currently Fritz considers that white's position is 19 one-hundredths of a pawn superior to black's. So the computer like's Kasparov's position. This depth reading that you're seeing here is 11 of 11. That means that the computer is looking at 11 ply.
For every move that Kasparov makes that's one ply. When the computer responds that's a second ply. So the computer is looking now at a depth of 8 of -- well, let's just say with 11 of 11. That basically means that the computer is looking at 5 and a half moves into the position. Down here we have this nice little graph P, where this graph is representative of the -- of who has an advantage. The green meaning Garry started with an advantage. There seems to be a little blip here that was probably myself playing around with the computer. And now you see that Fritz, after a deeper calculation, considers the position better for Garry. 12. Re1 Nf4.
Mr. Ashley Before you go on, Yas, explaining that, we should note that Garry has in fact played the move that you suggested, rook to E1, not trying to refute this move knight H5 immediately, but rather continuing to develop his pieces, finding the best squares for them and he knows that in due course the weaknesses of the move knight H5 will reveal themself. And I think that was definitely a professional move on his part.
Mr. Seirawan Absolutely. The move knight H5 will -- the weakness will only become apparent probably ten moves down the road. So what happens, how the computer thinks is very simple. It looks at every single possible move, every single possible legal move for its opponent. Then it considers every single legal possible reply to that, to all of those possible moves. And then that keeps going on like that, and therefore the great calculative powers of the computer come into effect. And because it can look at 200 million moves a second within the framework of 3 or 4 minutes, it can look 5, 6, 7 and sometimes as long as ten moves deep.
This is known as its event horizon. It's like a layer, one layer on top of another, on top of another. And eventually it can only go so far.
Now, the human being knowing that the computer can only calculate to a certain depth, call it for argument's sake 8 moves. The human being can say, well, I'll just look ten moves ahead or I'll look 12 moves ahead. It requires a great deal of calculative power on the part of the human being but he can do it in that way outfox the computer.
Mr. Ashley Also trying is a question asked to Garry yesterday. I found his answer remarkable. Someone asked him how far ahead he thinks in a given position. And this is a question that grandmasters are asked a lot because the average person thinks that a grandmaster's great strength is his ability to see deeply into a box in this way of calculating deeply, see all these moves ahead.
Some people think 20. Some people say 40. No, he doesn't do that. But Garry said well, only 3 to 5 in a given position. And you have the world champion, the best player in the world making a statement like that only 3 to 5 and we know that DEEP BLUE sees further than that immediately.
How does Garry defeat DEEP BLUE then if he's not looking further than 3 to 5 in most positions and in some only 7 to 8?
And before you answer that DEEP BLUE has played the move knight to F4 in response to rook E1 attacking Kasparov's bishop, one anticipates one of two moves B to B1 or bishop to D1. 13. Bb1 Bd6.
Mr. Seirawan Before I do answer the question, this next move I think will in many ways define Garry's approach to the match. Against a human I think he would play bishop to B1 in a New York minute. And he did it. And he did play bishop to B1. Bishop to B1 keeps in reserve this wonderful attacking option against the pawn on H7. Against a computer he may have considered the move bishop F1 just being a little bit more safe and secure. But he remained true to his nature and played bishop to B1.
Mr. Ashley I think this clearly shows that he wants to keep his possibilities of winning the game more so than thinking more for the draw. Because bishop F1 was a very solid kind of move, protecting his king. But now bishop to B1 says, yeah, I got thoughts.
Mr. Seirawan I got thoughts. I got nasty ideas.
Now again this is a very -- it's a bit awkward and I don't mean to say that Garry was being disingenuous with his answer of only seeing 3 or 4 moves ahead because what Garry's moves -- we've talked about the way the computer thinks. The computer's think is brute force. It sees every possibility. Garry's approach to chess is what is known as a selective search. He only looks at 3 or 4 moves because he knows based on his experience and in chess that these are the absolute best moves.
And if you deviate from the moves that he thinks are best for you, you're making a weaker move in his judgment. So he doesn't have to look at a hundred million moves. On the contrary, he only has to pay attention to the 3, 4, 5, ten most crucial ones. But that is one part of the judgment factor.
The other is that Garry is one of the best prepared grandmasters in the world, if not full stop the best
Mr. Seirawan He has a marvelous range of openings. He brings an incredible number of creative ideas to chess openings. He's written a lot of books about the openings. So what makes Garry so formidable is you have 3 phases in the game of chess. You have the opening phase, the middle game phase and the end game phase. Garry has a marvelous ability to marry his openings with his end games.
In other words, from the openings he plays he knows what kind of end games will result from them and he knows how to win those end games. So literally he comes with an overall game plan to each and every game and he is so formidable. So he may only be looking 3 or 4 moves ahead in the position but his experience tells him what's going to happen in another 30, 40, 50 moves. Big, big advantage for Garry.
Mr. Ashley Indeed. We should also make this point clear by taking a look at the last move. DEEP BLUE, before we get to that, has played the move bishop to D6. Let me just get to that. DEEP BLUE has increased the scope of this bishop and put the dark squared bishop on the Square D 6. But if we go back to the situation after DEEP BLUE played knight to F4 attacking the bishop on D3 in this position a human player immediately goes to the important factor in the position. The bishop on D3 is under attack. It has 4 possible moves. Well, 4 possible moves that don't lose material. Bishop to C2, bishop to B1, bishop to E2 and bishop to F1. Now, of those moves 1 move allows it to capture bishop to E2. That's a move we would think why give up a good bishop for this knight that's been moving around the board. That would be the only calculation. We wouldn't even calculate any further. DEEP BLUE would look at that position and look at millions, millions more beyond what we just thought. All we thought was bishop E2. I am not going to give you my bishop. Boom, full stop.
Now, the move bishop to C2, notice, Yas, didn't say anything about that. It also -- it retreats. If gets out of the way of the knight. But at the same time it blocks the rook from going -- the rook on C1 from going anywhere and it also stops the queen from being able to go to the C2 square. The move didn't even occur to Yas. I mean the same thing with me. I was only thinking of the other 2 moves. It didn't enter into our field of vision. DEEP BLUE would have analyzed that 1 millions of times before making a decision.
Not only that, DEEP BLUE would have analyzed moves like H4, leaving the bishop hanging. It would have analyzed other crazy moves like A3, A4, B4, C5, C takes D5. It would have analyzed all of these possibilities. And it would have been looking at all of them deeply, deeply, and then it would come back and say oh, that's all rubbish. My 2 key moves are bishop B1 and bishop F1. In other words, despite that great ability to calculate, you're not using it in a most effective and efficient way. Because you're analyzing everything and everything is not good. Most of it is garbage. It's tough enough to play 1 good chess move without looking at all the moves. Most of those moves are bad. Really, Yas, you might have to say that both moves are of equal value. It's just the kind of position you get from the 2 moves. And that was the decision of Garry Kasparov. Bishop to B1 pointing at the king more aggressive. Bishop to F1 protecting your own king, more passive.
We make this value judgment and it's probably all Garry Kasparov did when he said I'm going to attack today. I'll play bishop B1. And that took him all of 2 seconds to decide on. So that's the fact of selectivity as you talked about on the part of Garry Kasparov and the part of the human chess player, that makes it so difficult for the computer with all this capability to still defeat the best players.
Mr. Seirawan Well, indeed. And obviously the great computer scientists, chess computer scientists today both at IBM's labs and in fact labs all over the world, they know that the selective search, if you will, gives the human player great advantage. So what they're trying to do is work with their programs to put together selective searches. And I know a number of them have been very successful in a selective search. Now, that's the 1 that does get square 1. Because in the brute force method there are more possible moves in a game of chess than there are atoms in our known universe. So it's virtually impossible for the computer to analyze all possibilities
Mr. Seirawan But if they can get the computer to look at what Garry is doing, a select search, then they can take that great calculative power and suddenly the computer seeing just the best moves analyzing only the best moves, and then we'd be in deep do-do, deep do-do.
At this point it's poll time.
Mr. Ashley Oh, no, poll time.
Mr. Seirawan We have just gotten out of --
Mr. Ashley Yas likes polls, people. He could be a politician. 14. g3 Ng6.
Mr. Seirawan Out of the opening phases. We're moving into the middle game. And we know the audience is always right.
Oh, we have a move by Garry, so I'll just pause a moment.
Mr. Ashley Garry Kasparov has played the move G3 in response to bishop D6, attacking the knight on F4 and hopefully the computer will make a selective search and move the knight.
Mr. Seirawan Okay. That was good, Maurice. I liked that. But it will look at a check too. The situation once again, Garry needs only a draw or win and he's got the match. The computer has to win.
I'd like to invite all of those people who think the computer is going to pull off the upset to tie the match to please raise their hands.
We only have just a very, very few.
Mr. Ashley Out of 400 people in the room, only 3 hands went up.
Mr. Seirawan Now, at the start of the match virtually everybody thought the computer was going to win. Now the human has won respect.
Mr. Ashley They see Garry has the lead. They go with the winner.
Mr. Seirawan All those who think Garry is going to draw the game because that wins the match, it's the professional thing to do.
A fairly respectable number.
And then I assume the rest of you think Garry is going to trounce the computer.
Yes, that's right. You can all stay. Very good. Thank you very much.
Let me just talk a bit about the game at hand and then I'll be opening up the floor to some of your questions. We have a very large audience here today and I thank you all for coming. I know it's bitterly cold outside in the snow and everything. Obviously you've come to see an historical treat. The move G3 places pressure on Black's knight forcing the knight to move. Either the knight can move forward with knight H3 check, after which Garry would just move his king up and attack the knight, for it to move again. A greater probability is the computer will just simply repeat the knight back to knight to G6. Then you just take a look at the pawn structure for a moment. This bishop on D6 is an active piece, but this pawn on G3 blocks the diagonal H2 to B8. So the pawn on G3 does very nice work in limiting the bishop on D6. Remember, chess is a game of strategy, very complex strategies, where 1 player is trying to build up his possibilities while denying the activity of the opponent's pieces. Taking a look at the 2 bishops, the bishop on D6 and bishop on D7, especially that bishop on D7 is a Septilicus like I've spoken about before.
Mr. Ashley Somebody asked me to spell this word. I said I didn't know how to spell it; Yas made it up. I don't know what they drink in Seattle. Hear it's Starbuck's coffee, but I have my doubts what they're putting in that coffee. But that's the Septilicus? Because they thought it was a chess term but apparently it's a term you guys use in Seattle.
Mr. Seirawan Absolutely. Well, it's a very simple idea. You have certain pieces that are just bad, just plain bad. When a piece doesn't have any activity or scope for activity, it's a septic piece. So therefore you get a Septilicus. We do have a retreat, the move knight to G6 has been played. And Garry has a fascinating option here that I've been spending a little time looking at. 1 of the things that we always talk about is the imbalances and we have a majority of pawns on the queenside for White. Notice these pawns on A2 B3 and C4 and Black has but 2, A7 and B7. 1 possibility which is quite an attractive 1 is to play the move C5, attacking the bishop on D6. There would be the move, for example, 15, bishop C7, and then White could try the move A3. The purpose of the move A3 is to put this queenside pawn in motion with the move B4. That gets a nice bind on the queenside. And that works very well in human competition where you can just kind of grind your opponent down by pressing him, by pressing them against the wall.
Mr. Ashley Even more so it gets into a sort of blocked position which traditionally computers do not play very well. We know that the great forte of the computer is the ability to calculate, and that usually predominates in very open positions where there are no pawns blocking the middle of the board, but here with pawns in the middle and Garry Kasparov's ability to now play the move C5 and block the position some more, 1 would think that DEEP BLUE wouldn't play such a position as well as other positions and that would be the kind of situation Garry would really relish the most.
Mr. Seirawan Exactly. We've spoken about this earlier. I just ask how many have come to visit some of the earlier games?
Oh, a very respectable number. Thank you very much for coming. We must be doing something right.
Mr. Ashley We didn't chase you guys away.
Mr. Seirawan But 1 of the things that Maurice and I have both commented on very frequently in our previous discussions is the idea of open, semiopen and closed positions. The closed position is a very simple 1. That's where there are a lot of pawns on the board. Each side starts with 8 pawns. And if the position gets blocked and there's not an exchange of pawns, this is known as a closed position. Closed positions mean a great deal of maneuvering back and forth and a lot of strategical decisions have to be made. When you have a single exchange of 1 pawn, you have a semiclosed position or a semiopen position. What we have now, White is missing his E2 pawn and Black is missing his C7 pawn. So the position is still semiclosed or half open. Those are the terms that are used. When you have 2 pawn exchanges or more, then you have an open position. An open position, the variations are a lot more forcing, and it is felt popularly so that in an open position the computer does very well, although we saw an open position yesterday that the computer got completely killed in. So that's another point.
I was just going to say that Garry, I am absolutely certain once again against a human opponent because Black has wasted so much time with the move knight H5 knight F4 knight G6, 1 of the things that you like to do when your opponent breaks the rules is punish the opponent. And 1 move that absolutely Garry would in a blitz game, his instinct would be to play the move 15, H4, the idea being to continue with a pawn push up the board, H5, forcing that knight aside and then getting a chance to put this bishop on H7 in action.
Let's just see 1 possibility. H4. If there came the move rook C8, we would play H5.
Mr. Seirawan And this is the type of variation. The knight would be forced to go away with 16 knight on G to E7. And now you have sacrifices, for example, like bishop takes H7 check, that don't work. King takes H7. Knight G5 check. King G8. H6. Again, this variation wouldn't work. But it demonstrates that from afar white's pieces could jump all over black's king.
In a move like H4 would be a preparatory move towards a direct attack and I'm sure Garry is thinking about the choice of H4 or the choice of C5.
Mr. Ashley Most definitely those two look like the most promising continuations of the position. I really like the move H4. It very much goes along with Garry's style. But it may not go along with his pregame strategy.
Mr. Seirawan His approach to the match.
Mr. Ashley Which is to say draw the stupid game and win the match.
Now, that move H4, that says I'm trying to win and I'm ready to burn some bridges in order to win this game. And that's a critical decision. It's also possible, though, you must admit, Yas, that maybe he'll play C5 and block the center first and A3 with planned expansion, with seeming planned expansion, on that side after something like bishop to C7 and then go back to H4 plan with the center properly blocked.
And that would be consistent with chess strategy of blocking the center before you attack on a wing.
Mr. Seirawan Exactly. Well, Maurice pointed out one of the cardinal laws of chess is when you're attacking on a wing or when facing an attack on the wing, play in the center because that's the best way to undercut or underscore an attack on the wing.
So a move like C5 first closing off the center, preventing a counterattack in the center would be something very natural, but again Garry is one of those types of players who likes a fluid position with a great deal of options. And it would be in his nature to play a move like H4.
Also, I just was looking at it. There's another very nice point of the move H4.
Whoops, what did I do? H4. White could then combine with a knight salle of his own which would be to play knight to G5. 15. Ne5 Rc8.
Mr. Ashley Kasparov has chosen a third option in the game. It comes as a big surprise to us here. He's played the move knight to E5 instead of playing either H4 or C5. Knight to E5, after some thought. And this move comes as a surprise, Yas, especially since those exchanges that were -- the exchange that this offers with the bad knight on G6, what we saw as a bad knight on G6, didn't look so bad for black. And this looks like kind of a noncommittal quiet type move, looking for exchanges by Kasparov.
Mr. Seirawan Well, this is a move that also is a bit of a hallmark of Garry's strength. When he goes on the attack, he loves to invite everybody into the party. Okay? We want everybody involved in the party.
And the move knight E5 does two things. First of all, it does encourage that after, for example, a move like knight on G takes E5, D takes E5, D takes E5, the bishop may move, for example, to B4, bishop to B4, having moved a knight from F3, Garry has given himself the possibility of bringing his queen over to the king side with the move queen H5.
What he's also done and we haven't spoken about it to this moment is this bishop on B2 has gained a little bit more activity now that the pawn on D4 has moved up the board to E5. I think that what the computer -- the draw back in my view to this move knight E5 by Garry is that it does give the computer something to do. Now the computer is facing direct threats and it will probably respond very well from here for a few moves. So I like the move H4 because it kind of kept the position unclear.
Mr. Ashley Well, there are times where Kasparov plays moves that we didn't divine. So we will leave it to the world champion to show the real points to this move in a few moves. And if he loses, you know, we'll be able to get the I told you so's in there.
But otherwise knight E5 for Kasparov centralizing a knight may indeed turn out to be a good move. We will see.
Mr. Seirawan And Fritz certainly does like the move. Fritz has increased Garry's advantage from the opening. It now likes Garry's position by 34 one-hundredths of a pawn. So certainly from the computer's point of view maybe this was the nastiest move for Kasparov to play.
At this moment we'd like to open ourselves up a bit to some of your inquiries. I've seen some hands go up. Do not be shy.
(Audience commentary.)
Mr. Ashley Before you answer the question, Yas, DEEP BLUE has responded immediately to the move knight E5 with a developing move rook to C8. And the question from the gentleman is do computers have ratings and would you care to postulate DEEP BLUE's rating.
Mr. Seirawan Okay. First of all, let's talk about the concept of a rating. For all of those of you who are chess players you should know that there's a ranking system and what is known as a rating system. It's a 4 digit rating, and, scientifically speaking, chess ratings really are a very good measurement of a strength of a player. We know that Garry Kasparov has held at 1 time the highest rating in the history of chess players. And these ratings go back many, many decades. So we know that Garry is considered, scientifically speaking, to be the strongest chess player of all time. And his rating currently today is 2775.
Before the match Dr. C.J. Tan of the IBM team was asked to postulate on the rating of DEEP BLUE, and C.J. said, well, now that it sees so many moves, the rating of the computer might be 3,000. And all the chess players in the audience took a deep breath. They said what? 3,000? We're talking about -- poor Garry. And then when the computer won the first game, I had an audience member come up to me and say oh, by the way, Yasser, the computer's performance rating of today was 3200. Get out of here. I didn't need to hear that, 3200. From that point on Garry said that he felt from games 2, 3 and 4 that he was playing against a program that had a 2700 rating. And then yesterday he was disdainful. He was contemptuous. Because he had offered a draw in the game that had been declined and 5 moves later the computer had a loss position. And he said, my God, the computer played today like a 2300, 2400. So at times the computer plays extraordinarily well; in Garry's words, like God, or nearly perfect. And at other times it plays like it's never read an opening primer.
Mr. Ashley Well, we should qualify that a bit because 2300 and 2400, that rating is in fact a very strong rating, stronger than 99 percent of all chess players in the world. In the USCF, in the United States Chess Federation, a rating of 2200 and above qualifies you as a master. And you get a certificate. You get to put it on your wall and forever you're a master. Ralph Kramden would have loved that certificate. Now, if you have a 2400 rating and above, you're a senior master. Now that's not an international rating. 16. Nxd7 Qxd7.
On the international level there are titles such as international master and international grandmaster that you get from playing in these competitions, but most international masters and grandmasters are over 2500 United Chess Federation rating. So that is of a much higher standard. But even 2500 would be absolutely no match for Garry Kasparov. He's played the move knight takes D7.
Mr. Seirawan And he just whipped off that move.
Mr. Ashley It looked like a bad bishop. But just finishing the point, for somebody to play a 2300 strength against Garry Kasparov, the difference of 400 rating points means that you can play him a hundred games and you might win 1, even if you're a master. A hundred games against Garry, you might win 1. That's the significance of a 400 rating point difference. Just a hundred rating points difference suggests that you may not win a game in 5 games. It may be a 4 to 1 score, in Garry's favor, a hundred rating points. So as you can see, the lower the rating goes, the worst it's going to get. Now the computer Garry said is playing 2300, what he's saying is that compared to my level of play, my professional top level of play, it wasn't close to chess strategy. We're not trying to say that it's -- that 2300 players are bad chess players because indeed they're not. They're masters. But they're not of a level of a guy like Garry Kasparov.
Mr. Seirawan Indeed. And to answer the gentleman's question most directly in my postulation, in my gauge of the computer -- and I do want to emphasize this is only based on the first 5 games that we've seen -- the computer plays certain positions extraordinarily well and I would have to give it about a 2600 rating. Then we saw, as we saw yesterday, most especially, it would lose a game it shouldn't have and also in game 2 it had an ending that Garry felt most grandmasters would have easily held, human grandmasters.
So my belief at the moment is in a tournament rating we're probably seeing a 2575, 2600 at the moment. And of course we know it's only going to get better.
Mr. Ashley Frighteningly so.
The gentleman in the back.
(Audience commentary.)
Mr. Seirawan The gentleman's question is, I mention in passing the idea is that there are more moves in chess than there are atoms in the universe and it's such an awesome statement to think about it because atoms are the smallest things in the universe. And you say well how big is the universe? There must be a quadzillion, quadrillion moves in chess and that can't be.
And the reality is, is not only is it so but it is scientifically proven. And again, forgive me if I don't know these figures by heart. I think the number of atoms in the universe is ten to the 67th power, where the number of moves in chess are ten to the 106th power. It's not even close. Those are big, big numbers. Lots and lots of zeros.
Mr. Ashley It's one of the reasons why we're so attracted to the game because so many possibilities can occur, so many unique and interesting possibilities can occur. But it also speaks to what obviously we couldn't do, is unlock all the mysteries and all the secrets of the game.
The move queen takes D7 has been played by DEEP BLUE. But if we didn't have a system, if Garry Kasparov didn't have a system, if he didn't have a strategy, if he didn't have ideas that he knew and principle that he knew worked in all chess positions, he could not possibly compete with a computer. But because he has all of chess history behind him, all these principles that have been shown to work from experience and have been validated in countless numbers of grandmaster games, he is more than a match for the computer.
He can come up with ideas that looking at all these possibilities will not divine. And we have to really read it. The top chess players of the past with making these discoveries, such as which Wilhelm Steinitz, Aaron Nimzowitsch and the whole hyper modern school. There's wonderful literature about the whole development of the history of chess and chess thought that has progressed over the last 150 years or so. Very fascinating reading material if you ever wish to look it up.
But the fact that that history is there, Kasparov can really say that he rests on shoulders of giants and what makes him the great player he is today and what makes it unnecessary for him to look at all these innumerable possibilities in a chess situation.
Mr. Seirawan A very close friend of mine Dr. John Nunn of England wrote a book last year, wonderful book, highly highly technical book on the secrets of pawn endings. And so this book was published in 1995 and I happily sat down and I just opened up the book and said, okay, John teach me something I don't know.
I opened up the book and the first thing he does is he quotes an end game position from an Arab manuscript from the 11th Century A.D.
I'm looking at this book and I'm saying, my God, this is analysis, classical analysis, 800, 900 years old that some Bedowin Arabs sat down and wrote on lamb skin how to win a particular type of end game. And that analysis was as valid then as it is today. So literally you have generations and generations of analysis built upon. And Garry from a very early age has been studying his classics. So he really knows what's going on.
(Audience commentary.)
Mr. Ashley Where is the game actually taking place in relation to this room?
Mr. Seirawan You go down the hall, you take a right. It's just a few doors away. We're sandwiched in between with the press center, and there's also the room where all the worldwide web activities as well as DEEP BLUE's backup, BABY BLUE, is in the room. So there's this huge shuttlelike area with space station types of computers all humming and whistling away and a lot of activity. And then you're all welcome of course to go into the room. You, of course, see it on your screen. What you'll see is a stage, hush, hush, very quiet and you can sit down and take a look at what's going on.
(Audience commentary.)
Mr. Seirawan Even now as we speak.
Mr. Ashley Despite the fact that Yas did say you can go into the room, it's nowhere near this size. In fact, it can only accommodate about 12 spectators at a time, maybe 20. So you --
(Audience commentary.)
Mr. Seirawan Oh, I apologize.
Mr. Ashley Then Yas was wrong about that 1. Only during the first 5 minutes of play 1 had access and I guess they don't want any Garry fans going in there and shouting "go Garry."
Mr. Seirawan I apologize.
Mr. Ashley So it's only open to the media. So those who were interested in going in, keep your seats.
The gentleman was asking a question.
(Audience commentary.)
Mr. Ashley A good question by the gentleman. Why did Garry Kasparov resign the first game with so much at stake, $400,000 first of all and humanity, second of all.
Mr. Seirawan Well, that's always a tough question of course because the gentleman said he was an inexperienced player and he didn't understand because even in the final position in his experience he says to himself boy he's still got some chances he can kick around a bit. Why did he resign. 17. Nf3 Bb4.
Mr. Ashley Before you answer that, Garry has played the move knight to F3.
Mr. Seirawan Okay. Developing a piece.
The answer to the question is that from Garry's perspective, he had no chance. Grandmasters, when they resign, they do it with a heavy heart. I mean he wasn't happy about having to resign. So he felt that he was in a position that he couldn't come back from. Sometimes -- in fact, I always recommend this to students of chess and Maurice most certainly does as well. When you're inexperienced, just play the game out to mate because then you can gain some experience. But in Garry's view that particular moment the computer was going to play it perfectly and he couldn't get out of it. So he felt he was lost even though it wasn't perhaps completely obvious to everybody that he was going to lose it. But he was.
Mr. Ashley On a grandmaster level a single pawn we say is like a difference of 20 points a basketball game. 2 pawns, 40 points. Garry was about to be down a third pawn. When it's a 60 point difference in a basketball game, you get up and start heading for your cars to go home. Garry didn't want to have to suffer that to the finish and we have a wonderful opportunity in chess that says I can just stop the game instead of playing on. And actually it's disrespectful to play on in positions like that. Actually it's respectful to your opponent to continue because you know the guy has the skills to win a position like that in his sleep. And that position was really just completely dominating for the computer.
Mr. Seirawan Speaking of chess etiquette, it was wonderful.
Were you all here for yesterday's game where Garry won? Did you see the ending part of that game? It was absolutely great. What happened was Garry's watch goes back on, he starts looking at his watch. He's got a winning position at the board, but the computer team of IBM wasn't ready to give up the game yet because they wanted to see, well, maybe we can still save it. So Garry is sitting there tapping his fingers on the table, looking at his watch, come on, come on, come on, resign already; and a few moves later they did because then it became obviou. I think Joel Benjamin, the grandmaster on IBM's team, threw in the towel.
Mr. Ashley There's no real etiquette there, but...
Mr. Ashley There was 1 game, 1 story I remember and I never could validate this story. I remember I read in a chess book about a guy who kept playing on in a loss position and his opponent had a heart attack and of course he won the game. Now, the comment was, the comment by the person writing the book was of course this kind of behavior won't endear you to your colleagues. Now, still I don't know the validity of the story. I remember reading it in a book. Would I love to know if it indeed were true. But this is the kind of thing. That's what would have to happen for Garry to not win yesterday's game or for DEEP BLUE -- well, it can't have a heart attack obviously but not not be able to win the game. For a grandmaster the technique is so obvious. It's so obvious. You give Jack Nicklaus a 6 inch putt. It's obvious. He doesn't even have to pick the ball up because you know he's going to get it in. Only in the case of a hurricane wouldn't he get the ball in. I'm sure Jack would argue yes, I would. So you just have the respect and you say yes we know it's a winning position and you stop.
A move has been played in the game bishop to B4 by DEEP BLUE, attacking Garry Kasparov's rook on E1. Bishop to B4. 18. Re3 Rfd8.
Mr. Seirawan And after bishop B4 this -- I don't think that this is a particularly effective move because it does of course contain the poisonous threat of capturing the rook, but after the simple rook lift, rook to E3, getting out of the direct attack, the rook is better positioned on E3 than it was on E1. So White -- Garry is going to benefit from this early attack.
Also, I'll go back and say now Garry in fact has quickly played the move rook E3. He has used 40 minutes for his first 18 moves. So much better clock management by Garry today by the way. But now I'll go back and say DEEP BLUE has to be wary of the idea A3 and C5 getting that queenside bind that I had mentioned earlier. Because now it would come with tempo.
Mr. Ashley What about the move C5 first so that the whole A3 B4 business would happen with tempo.
Mr. Seirawan Yeah, bang bang bang bang bang. Absolutely.
Mr. Ashley And it's interesting, though, that exchange by Garry Kasparov, knight for bishop, what do you think about that? Because I'm intrigued. We go back a couple of moves to the position where Garry Kasparov played his knight from E5, and he took -- after rook 15, rook to C8, he immediately played the move knight takes D7. That bishop was, as you called it, a Septilicus bishop.
Mr. Seirawan Septic bishop.
Mr. Ashley A bishop in normal chess parlance that is just blocked in by its own pawns. That bishop was not really doing much. White had a knight that was really centralized. What was the point behind that maneuver and why does it look as if Garry's position is still looking very promising at this stage.
Mr. Seirawan Well, 1 of the -- again very difficult relationships in chess is understanding the bishop versus knight conflict. And 1 advantage that the bishop has over the knight is that as pawns are exchanged, as the position gets reduced, diagonals are opened. So that in the end game the bishop tends to be a more effective piece.
Mr. Seirawan What Garry could have been saying to himself when he made that I would say slightly unorthodox swap of knight for a bishop is he was saying that you're right, Yas, at this moment in time the position is not good for that bishop, but it may have a future role to play and therefore I'm going to capture it now.
We do have a move by DEEP BLUE.
Mr. Ashley Yes. Rook F to D8. Rook from the F file has been moved to the D file, D8. And the computer has been pursuing a singular strategy of attacking down the D line in its last couple of moves. It's cleared out the D file, moving the bishop off D6 and now moving a rook to the D line, and it seems as if it has intentions of exchanging off pawns in the middle of the board so that it can attack white's D pawn down the D line.
It seems interesting but it doesn't seem as if it's going to be very effective, especially if Kasparov now plays the move C5, then that beautiful battery of queen and rook on the D line for black is now just running up against a brick wall with a pawn on D5.
Mr. Seirawan Would be totally ineffective. In fact, I was really surprised by DEEP BLUE's last move because he, DEEP BLUE, the computer, is completely underestimating the effect of the move C5 where Garry would have a perfect opportunity to clamp down on the queen side. And I think the move C5 is going to be an awesomely strong move. It will have a very long term effect in the game.
And after C5 I think the computer will be in real trouble. The last move of the computer rook D8, just as Maurice, was saying was to open up the D file and to put some pressure against the D4 pawn. So therefore, in my view it would have been better for the computer --
Mr. Ashley Yeah. Exchanging for the previous move.
Mr. Seirawan To begin with D5 takes C4 as opposed to 18 rook on F to D8 which it has just played. And again, I would say Garry is going to jump all over this opportunity. He's going to play C5.
Mr. Ashley C5 definitely looked like a princple move, very logical in this position. And then he could initiate those attacks we were talking about. Again, the idea of C5 being very much a part of this position and then H4 and H5 and all those ideas could definitely follow.
So I have real doubts about this game for black. It's not the kind of situation black wants to be in needing to win a game.
Mr. Seirawan Exactly.
I'd just add another comment to the heart attack story because that heart attack story by the way is a very, very famous one and there's also a corollary to it, and that was there was a chess tournament and a fellow, you know, was so excited. You know, he had a heart attack and the doctor was called over and he said, well, what do you think, Doc, is there anything you can do for me? He said, Nah. But it didn't matter because his position was lost anyway. It didn't help.
(Laughter.)
Mr. Seirawan And one other story, very quick one, one of my favorites. Has anybody heard of the great grandmaster from Poland Akiba Rubenstein? Wonderful, wonderful chess player. Historical player who had a beautiful pure style, very classical style. But the story is told by George Kultinowski. And again, this is a little bit tongue in cheek, but on Akiba's death bed the doctor came to him and said Akiba I've done everything I can do for you. I'm sorry there's nothing more I can do to help you but you're going to die. Is there anything else I can -- that you might want? And Akiba made a motion to say to the doctor, Come, I'll whisper something in your ear and the doctor leaned over Akiba's mouth and Akiba said I'd like a second opinion.
(Laughter.)
Mr. Ashley Chess players don't like to resign.
Mr. Seirawan Chess players never like to resign, going back to what you said in game 1.
Gentleman with his hand raised.
(Audience commentary.)
Mr. Ashley Will DEEP BLUE offer a draw or tender its resignation?
Mr. Seirawan No. The computer will not. The engineers in fact make that decision for the computer. On the table of evaluation we see that Fritz believes that Garry's position is indeed very nice, giving it forth-one hundredths of an advantage. If that ever got up to, say, 3 points or 4 points, then the computer would still play on.
But the engineers say it's getting too much and they'll resign on behalf of the computer. Yesterday, however, we had a very interesting set of circumstances and that was that Garry Kasparov made a draw offer. Now, this produced a great deal of excitement in the back rooms where all of the computer scientists were mulling about because at that particular moment that Garry offered a draw the computer felt it had a 3100's of a disadvantage. So the computer was telling the engineers hey I got the worst position. I'm getting off the hook here. Take the draw. And the engineers in the interest of a historical scientific experiment I believe the expression was, in the interest of furthering the historical scientific experiment, declined the draw.
DEEP BLUE went down in flames and they're still kicking themselves.
A question in the way back.
(audience commentary.)
Mr. Seirawan I apologize. Are you saying --
Mr. Ashley The question was why would DEEP BLUE allow a move like C5 which looks very, very obvious to us.
Mr. Seirawan Exactly.
Mr. Ashley Doesn't DEEP BLUE see it, or is it the programmers affecting DEEP BLUE in some way?
Mr. Seirawan Well, again, once again, this is the weakness, if you will, of the program. The program is probably calculating, as is Fritz, at least ten ply deep and probably even closer to 15 or 16 ply. And in all of its calculations the computer doesn't recognize that the move C5 is really a dangerous move because in the next 5, ten moves it's not obvious to the computer that it will be losing after a move like C5, but to the human player, the human would sit there and say, gee, I may not beat you the next 5, ten moves but I'm going to beat you in 50 odd moves.
Mr. Ashley But you know, Yas, looking at the way DEEP BLUE has been playing this game, it really looks as if DEEP BLUE has its strategical functions turned off. And I really -- I mean I don't mean to say that -- I don't mean to be facetious.
Mr. Seirawan Sure you do.
Mr. Ashley Not at all. But there are a couple moves it's played that after seeing the first 5 games, I would never have anticipated it playing. It seems to be playing on all tactics right now. Knight H5 to F4. Quick bishop D7 instead of long range B6 bishop B7. Bishop to B4 attacking a rook. Rook to D8. You know, it's like it's setting up tactics on every move. And I'm wondering do they tell it to go on extreme tactical move in a game that it really really needs to win instead of playing compositional moves. I don't know if they could tweak with it in such a way right now, but that's how it looks. Look at these moves. You didn't respect knight H5. It didn't look very good. But it had that tactical point of knight F4 attacking a piece. Bishop B4, so what? But it had the tactical point of attacking a rook. Now this move rook D8 had a tactical point of getting pressure along the D line and 1 has to wonder, before it really did look like it was seeing some strategical points, but now it seems like it's setting up minor tactics on every move.
Mr. Seirawan Well, again this is -- everything you're saying is absolutely accurate, Maurice, but do keep in mind that this comes after an education that you received over 5 games just like myself and Garry and everybody else.
Mr. Ashley No question.
Mr. Seirawan I mean basically Garry has managed now after the experience of 5 games to adjust his style. Okay. He's made some major adjustment to his style and suddenly we're sitting there saying, well, my goodness look at all these weaknesses that the computer is making. Well, it's only making them because Garry made the adjustment to uncover the weaknesses of the computer. Now, if Garry had adjusted his style in the wrong way, we might be saying to ourself, my God, the computer is just killing Garry because he's playing tactically with it and he's getting smashed. So if Garry Kasparov is smashing the computer, as we think it is today on a strategical level it's because he's managed to make those types of positions where the computer is at its weakest. So kudos to Garry.
Mr. Ashley And again we talk about the word flexibility. Humans have it. You know, we're stupid sometimes. We do 1 thing wrong even though we know it's wrong and we do it anyway, thinking maybe we'll get over this time and bam we get hit with something and we say okay I learned from it. And do the right thing the next time. We know how to do that. Garry Kasparov adjusting his style against the computer after seeing the first game where he decided to play tactically even though he came into the match knowing not to play tactically, don't play sharp with the thing, it calculates very well. Did it anyway. World champion did it anyway. Lost that first game. And since then every opportunity to mix it up Garry Kasparov has said no no no. Only quiet positions, please and he's managed to contain the computer and has been plus 2 over that period of time.
Question in the front.
(Audience commentary.)
Mr. Ashley Well, Garry Kasparov is indeed stopping on this move to think for a while and the possibility of knight to G5 has been suggested and does indeed present itself, knight G5, which is a move which tries to seemingly win immediately Yas, I mean a very aggressive kind of move, trying to bring the queen over to H5, but serious commitments in its wake in that the knight leaves the center of the board, it still offers Black the opportunity to open up that D line with pressure. The knight also -- the knight and the queen defend the D2 square. I should make that point. The knight and the queen defend the D2 square and if they leave, then a bishop can jump into D2 and attack the 2 rooks. So that is the kind of thing Garry Kasparov will be considering in depth. But I have a feeling that knight G5, unless he can really make this work, he's going to ignore it as a possibility. It's a bit too tactical in fact. It's a bit too sharp to play in this position.
Mr. Seirawan I was just going to say that's very interesting. The Fritz computer had been analyzing knight G5 and after 6 minutes in its search it determined that that was the best move. It's very strange. I never looked at it. Not to say that I was blindfolded, if you will -- we have a new move by Garry.
Mr. Ashley Well, he has instead of playing C5 which we anticipated.
Mr. Seirawan As an absolute. 19. h4 Nge7
Mr. Ashley Yeah, as an absolute, yes. And -- but actually this is interesting. Because if we go back in a couple of moves in this position, if you remember this position back here excuse me. Knight G-E 7 in response to H4 was its instantaneous response. But just going back a moment, you remember this position here we suggested 2 moves. We suggested H4 and C5. The position before -- just after Garry had played the move G3, knight back to G6. And we thought H4 or G5. And we had talked about the dynamic possibilities H4 going after that knight immediately, and I said well you know Yas C5 it keeps the position solid and then you're on the attack which is what we were expecting from Garry in this position and you made the point about Garry's desire to keep a position as fluid as possible.
Mr. Seirawan Exactly.
Mr. Ashley So here after the 3 moves knight to E5, rook to C8 -- knight to E5 and then rook to C8, in this situation we like the idea of not giving Black any counterplay at all, none to speak of but blocking the center with the move C5. But Garry, again keeping to his style of fluidity, has decided to play the move H4. And instead of taking on C4, DEEP BLUE immediately responded with the move knight back to E7. Knight GE 7.
Mr. Seirawan Right. Well, 1 of the points of Garry's move is he very much would like to advance his H pawn to H5 and to H6. This would cause a great deal of discomfort on the dark squares in front of DEEP BLUE's king. But the 1 piece that could cover those dark squares is the bishop on B4. And once again I would say that the move C5 makes a very attractive possibility. But in that case I wouldn't have used the move H4. The idea of H4, Garry seems to have committed himself to be playing for the attack and the natural followup may be, for example, a move like queen D3.
Mr. Ashley Aren't you a bit surprised though that DEEP BLUE played knight GE 7? It doesn't seem like the greatest square for the knight in the defense of this position.
Mr. Seirawan Indeed. Knight GE 7 came very quickly. Again, I would have predicted an exchange of center pawns just to open up the D file first.
Mr. Ashley And is E7 like a top square? You pointed out the F8 square for the knight where there were much better plays defensively.
Mr. Seirawan Well, there are a number of principles and witticisms and anecdotes in chess as we've heard of some and 1 of the most well known is Bent Larson's expressions when I have a knight on F8 I can't get mated. And I keep doing the wrong thing with the computer. But the computer's last move, knight GE 7 put the knight maybe on an awkward square and I was just going to say that it seems to me far and more natural that the knight might have found itself back on Bent Larson's square with knight on F8. But again I would have predicted an exchange of pawns by DEEP BLUE to preside precede this movement but again DEEP BLUE knight GE 7 very, very quickly and just to follow up what my thought was, in chess what you like to do is create targets, most specifically targets are thought of as pawns. So you like to create a weak pawn in your opponent's Garry Kasparov. But you could create a weak pawn you have to create weak squares. The way to create weak squares is to lure your opponent to advance his pawns up the board. And 1 move that could do that very nicely is a move like queen D3. Although this is as you might put it Freddie Flintstone chess, cave man chess, it has a very brutal threat, a direct threat of queen takes H7 check, snaring a pawn.
Now, if DEEP BLUE was to play the move F5, well a weakness would have been created thereby. This pawn on E6 becomes a backward pawn F the computer were to play G6, then again some new weaknesses occur, specifically the F6 and H6 squares are weak. Remember, pawns only move forward. And as they move forward they leave weak squares behind in their wake. So a move like queen D3 encouraging a pawn to move forward.
Another idea is a sacrifice that probably is certainly thinking about. After queen D3 there is the possibility of 20, knight F5, blocking this diagonal
Mr. Seirawan Comes the move 21, G4, knight takes E3, winning an exchange, a rook for a knight. But then we would pounce down with queen takes H7 check. The king, black's king, would be forced to move to F8 and then after recapturing the knight, white would have sacrificed a rook for a knight and pawn, but look black's king is about to become the hunted.
Mr. Ashley Yeah. And I'm sure Garry is looking at variations such as these. And still, though, wondering is this in keeping with his strategy for the game. A draw is enough. Does he need to sacrifice, go hog wild, play like Freddie Flintstone, bam, bam get the guy or just be calm and let the computer try to win because it's the computer that needs to win this game, not Garry.
Those lines are certainly tempting and there's no question that Kasparov sees these lines. He's a great calculator, a very inventive attacker. And so he sees these possibilities.
But again, we've seen him reign himself in time and time again, saying no, let's not attack. I tried that in game one and it didn't work. No, I don't need to win that way. I can win another way.
Mr. Seirawan Well, he was definitely playing with fire and he doesn't want to get burned again.
More questions from the audience. We had a couple in the middle. Yes, sir.
(Audience commentary.)
Mr. Seirawan The question is can they change the parameters of the program during the play. And this is very interesting. Again, I'll only say this. This is a historic first in this kind of a match. And in my view they're going to have to make some adjustments with the rules. As I understand it, the answer to your question is yes, the engineers could do some changes while the game is in progress and I think that that is one of the rule changes that will probably be made if we see a revenge match.
Mr. Ashley I wasn't aware of that, that they could change.
Mr. Seirawan Yeah, they can tweek things around during the game. Not that they would want to because they wouldn't have enough time to test it. But as I understand it that option is still open to the DEEP BLUE programmers.
Perhaps later in the day we'll get a chance to invite some of the DEEP BLUE programmers to come up and rap with us a bit. They've been very kind in the past.
Let me just say this. This of course is once again an historic event but it's not the first event that the ACM has put on, the Association for Computer Machinery. ACM has been -- this is their 50th anniversary of the ENIAC computer. They're celebrateing actually a hundred years of computing because they want to celebrate the next 50 years today. There have been events going on all over. And the ACM for the last 25 years has been putting on computer sponsored events.
I must say that if all of the chess events I've ever been in, and I don't mean this other than in the sincerest flattery, this has been a marvelous, marvelous event and I really would appreciate if would you join with me in thanking the ACM for this event.
(Applause.)
Mr. Seirawan A lot of people, a lot of people spent a great deal of time and effort and it's unfair to single out just some of what is a team of people but I can't help myself. Joe Palozzi, Monte Newborn, Terry Phoenix are 3 of the people most responsible for making it such an enjoyable event. And thank you to those 3 people specifically.
(Applause.)
Mr. Seirawan On the same note, IBM, the DEEP BLUE program, has a great team behind it. They have been working with ourselves, Maurice and myself and everybody. The IBM web site, there's a whole crew of people that have put together the worldwide web pages. The IBM worldwide web pages are getting upwards of 5 million hits a day globally in hundreds of countries around the world have come in and the match has been broadcast to them. IBM has done a wonderful job, not only with its team but with its web site and a deep congratulations to IBM for what they've put together.
(Applause.)
Mr. Seirawan And most certainly last but not least is Garry Kasparov. Garry is the greatest human chess player today. He certainly didn't have to accept this challenge. It was one of those horrible situations where the only thing that could probably happen is he could lose. I mean everybody, yourself, myself, certainly expected Garry to win, but had he lost, it would have been terrible for Garry.
And I know a lot of my colleagues don't like playing against computers, especially when a great deal of money is on the line. They think it's unfair. We get nervous. The computer doesn't. And Garry Kasparov did not back down from this challenge and it looks like he's going to win it, but please join with me in congratulating him.
(Applause.)
Mr. Seirawan Okay. At this point we'll continue with our questions. There was a gentleman there. Yes, sir.
(Audience commentary.)
Mr. Seirawan Why does Garry keep leaving the table?
Mr. Ashley I'll tackle that one. Garry is a man who's full of emotion, full of emotion. Garry -- part of Garry's strength as a chess player is his -- the fact that he can communicate what he feels at the chessboard. You sit across the board from Kasparov, I've had the happy occasion -- I should say unhappy occasion -- to do so just once. I know Yas has done so 5 times. You feel Kasparov at the board. He lets you know he's out for blood.
Now, with that there are a number of things he does. And we talk about his removing his watch. But there are many, many others that we'll talk about as the afternoon progresses that other grandmasters don't do. And you'll see him move around the board. You'll see him collapse his hand over his face. He'll go back, he'll go forward, constantly making all these moves.
Now people used to believe that Garry did this on purpose to intimidate his opponents. But he doesn't. At least not according to Garry. He says that he just can't him help himself while at the board. He's feeling all this tension at the chessboard and sometimes you'll see, for example, Nigel Short, grandmaster Nigel Short complain about Garry in one occasion where Nigel made this move that obviously Garry thought was a horrible move and Garry sat up as if he just couldn't believe the move. He's looking at it. He starts going like this as if the move stinks. He starts pulling on his nose. And then he starts looking back at other grandmasters saying come on over to look at this horrible move.
And Nigel just couldn't believe he was being humiliated like that. He proceeded to get crushed. But that is Kasparov.
People ask him why he gets up from the board. And he says I just can't sit there. Once I know what's going on, the nervous tension of the moment doesn't allow me to sit. I have to do something. So he gets up.
He has a dressing room in back that is provided for him and he goes back to his dressing room. There's more space, less tension in that situation. He has a television screen the way we do here to look at and if a move is made by DEEP BLUE then he can get up and come back quickly from back stage.
He also has a little niblets, chocolate, little drinks back there that he can go and refresh himself with.
So there's a room provided for relaxation for him during a game. And that's why he will almost always get up from the chessboard when it's not his move because there's no reason for him to sit there.
I would like you now to applaud with me please. One of the key members of the IBM DEEP BLUE team, Murray Campbell.
(Applause.)
Mr. Ashley While we're getting Murray a microphone, it is truly been a joy to participate in this wonderful event. None of us knew what it was going to be like when we came here but they have been literally working years to bring this about and we're really impressed by it all and how the computer has played so far.
Yas, I remember you had a question for Murray regarding DEEP BLUE.
Mr. Seirawan Oh, boy, we've got lots of questions. First of all, once again, Murray, I mean I know you guys were working your fannies off to get your progeny here and playing as well as it's done. It's been remarkable. Game one must have been an extraordinary breakthrough, a great deal of happiness as were you celebrating the Chinese New Year.
MR. CAMPBELL That's right.
Mr. Seirawan One of the things that we had a conversation about last night was tweeking the program during the game or during the match. Currently do the rules allow you to in some way play with the computer's evaluation, algorithms and what have you?
MR. CAMPBELL Well, during the game it's completely forbidden to change the program in any way.
Mr. Seirawan Okay.
MR. CAMPBELL That would -- obviously that could be cheating because you could bring along a grandmaster who would say, oh, in this position move the knight to the king side. And so you want to let the program play on its own and show what it can do.
Now, between games is a different matter. As people go back and look over their games after a particular game and adjust how they play or adjust what they're going to play in the next game we do that also. It's a very difficult thing to do under these circumstances because whenever you make a change you fear that you're going to -- although you fixed the one problem, you create a host of other problems. And so you're always very conservative about making changes to the program during a match like this. You really need a lot of time to test any change that you're going to make.
Mr. Ashley Have you made any changes for this game knowing that DEEP BLUE really needs to win in order to tie the match?
MR. CAMPBELL We've made some very small changes, hopefully it would produce, you know, less likelihood of a draw.
Mr. Seirawan Does that mean you've raised the contempt factor?
MR. CAMPBELL Perhaps, perhaps.
Mr. Ashley You could tell us now.
MR. CAMPBELL You know, we have increased it somewhat. Also, even though the position, you know, on the board isn't that great, from our point of view, it's at least complicated. At least there's a game, although there's a chance it's going to lose, there's a chance, small chance, that it's going to win. It's not just a dead position with nothing to play for.
Mr. Seirawan So you want it obviously to play a provocative type of game and certainly the computer did that. That move knight H5 must have sure shocked Joel Benjamin. Did it shock yourself?
MR. CAMPBELL I was a little surprised, yes.
MR. CAMPBELL I don't know how many times that knight has moved in this game, but more than I care to count.
Mr. Ashley 5.
MR. CAMPBELL That's not so bad.
Mr. Ashley From whose point of view?
Mr. Seirawan We had the question earlier from someone in the audience. They asked myself. I didn't think I had a very high opinion of the computer's rating. Now, of course, you did some testing prior to the match. Did you have any conception of what rating strength the computer might be at?
MR. CAMPBELL We don't have any idea of how the full-fledged DEEP BLUE, with the 32 node SP-2 and all of the processors -- we don't have any idea of what the rating is.
And my opinion is it can be 2 different players on 2 different days. You can never tell how well it's going to play. If you give it a position that it likes, that it can deal with, it outperforms any person in the world, I think. And if you give it certain other kinds of positions, a B class player can outperform it.
So human ratings are based on the assumption that the variance on ratings -- and we have a move.
Mr. Seirawan We do have a move.
MR. CAMPBELL The variance on the ratings is about 200 points. And I think that assumption doesn't apply for computers, and so the rating scale is perhaps not exactly appropriate. You have an average rating, but the variance is much larger.
Mr. Seirawan In other words, a human rating is a poor way of trying to measure the computer's strength.
MR. CAMPBELL Well, when you say the rating of the computer is 2600, it could be anywhere from -- you know, for a human player, you would say on a given day they'll play between 2400 and 2800. For a computer, you could say 1800 and 3,000, something like that.
Mr. Seirawan I just want to talk about the move. I don't want you to leave. 20. a3 Ba5.
Mr. Ashley The move A3 has been played, Yas, attacking the bishop on B4. 1 can anticipate that bishop moving 1 of 2 options open to it.
Mr. Seirawan Bishop A5 would invite the B4 C5 maneuver, as we've seen before. And bishop back to D6 might invite the same type of thing, C5 as well as B4. We had expected Garry to maybe try the move queen D3.
And for those of you who have been looking at the screen, there's also that interesting sacrifice bishop takes H7 check. A classical sacrifice. Greek bishop sacrifice, as it's called. I'm sure the computer had anticipated bishop takes H7 check, but I notice when I put it into Fritz, Fritz kind of liked the move a bit.
Mr. Seirawan One of the questions that I had, of course, was that DEEP BLUE's predecessor was DEEP THOUGHT, the strongest chess computer program that I have ever seen, and now of course you have the new program DEEP BLUE. How much stronger is DEEP BLUE than DEEP THOUGHT?
MR. CAMPBELL We've run tests between early versions of DEEP BLUE and DEEP THOUGHT, which is a small scale version of DEEP BLUE, but we've never run any games between full DEEP BLUE that's playing here today and DEEP THOUGHT. So we don't know. But the earlier version was beating DEEP THOUGHT more than 2 to 1. That was a very small scale version of DEEP BLUE.
Mr. Ashley Before we take anymore questions, I don't want anything to suddenly surprise us. Something we may not have anticipated, especially since it may be something positive in the position, but that sacrifice you talked about, bishop takes A7 check, does look like a classical 1 one, the kind of sacrifice that Garry Kasparov might readly do in a speed game against a human player, because it seems to give such a strong attack.
Maybe DEEP BLUE has seen the sacrifice, if it is actually good, and might play a strange move in this position. I'm talking about this sacrifice that you just pointed out on the previous move. If you will, Yas, explain this sacrifice.
Mr. Seirawan Okay. Well, this is known as a Greek bishop sacrifice. The idea is to draw the Black king out into the middle of the board or to simply open up the Black king shield. So you play bishop takes H7 check, the king recaptures, then you follow up with a move knight G5 check. This opens up the way for the queen on D1 to enter the game in a big hurry.
Normally speaking, the king drops back. If the king is forced up the board -- go back just a moment. If the king is forced up the board like, for example, king G6 or king H6, it's really wide open.
In other words, that Black king is in the middle of the board and, normally speaking, it would drop back. So knight G5 check, the king would drop back with king G8, and now we whip out the old queen H5. This threatens to capture the pawn on F7 and it also threatens to penetrate with queen H7. There's probably 2 defenses, knight F5 and knight G6.
I was just looking at the move knight G6 and I then further continued with queen H7 check. The king is forced to go to the F8 square, and now either rook F3 threatening to play queen takes G6 or threatening sometimes to capture twice on F7. This looked quite promising.
And so this bishop H7 check is a well-known tactical mechanism with lots and lots of books written about it.
Mr. Ashley Now, Fritz 4 believes that this position is completely winning for White. The other move you mentioned was knight F5. What does that lead to?
Mr. Seirawan Okay. We're again with the same sacrifice. What was that 22, knight to F5? Instead of 22, knight G6.
And this is all analysis and speculation. But after knight F5, again that quick penetration with queen H7 check, king to F8 obliged, and now a move like rook F3 once again, just setting up for various tactical cheapos, a move like G4 is on the horizen and so forth.
Mr. Ashley Certainly a sharp and unclear sacrifice.
Mr. Seirawan Speculative perhaps, but certainly the kind of thing that Garry was considering.
Again I'd like to take advantage of Murray's presence. Do we have some questions for Murray? We have a gentleman at the aisle there.
(Audience commentary.)
Mr. Seirawan Let me repeat the question. The opening moves came up very quick and furious, what was the reason for that? I will add to the gentleman's question. It seemed like you had difficulties at the start of the match. Perhaps you could answer both.
MR. CAMPBELL There were some problems with communication at the beginning of the match but before the game, just before the start of the game, everything was running.
The opening in this game, we did not want to repeat game four, we believed that Garry got an advantage in the opening of game four, so this was a way to vary.
Mr. Seirawan C-5 was a novel play.
MR. CAMPBELL C-5 has been played before.
Mr. Ashley Yesterday's game was certainly a setback for you in the position of the match. What were your thoughts on yesterday's game, particularly on how DEEP BLUE played after the draw?
MR. CAMPBELL Before the draw I thought it played fine. It must have been a psychological draw for us. It did not play particularly well. This is in part, although it is a serious match, it is also in part for us a learning experience, and we certainly learned something from that game.
Mr. Ashley What do you feel you learned? Because we had thought that ideally, generally one could say the computer plays extremely well in open positions and yesterday it was in open position, yet it did not handle the concept of balance of majorities in an excellent way and ended up losing very quickly.
MR. CAMPBELL That's it. It did not handle the concept of the balance of majorities in a good way and that has to be looked at. It is not something that we can fix in five minutes. That is something that requires serious work.
(Audience commentary.)
Mr. Seirawan To repeat the question, is the computer able to change its evaluation function during the game to perhaps go for a draw or something?
MR. CAMPBELL Yes, it recognizes what the position is like at the start of each move and decides if it needs to reevaluate some of its evaluation terms and it will do that, and I think most programs do do that these days.
(Audience commentary.)
Mr. Seirawan The gentleman raises the issue of the fact that you have 256 proprietory chips all working in parallel processing to address that as well as what you are intending to do to improve your computer for the next generation.
MR. CAMPBELL The chip that is the heart of DEEP BLUE was designed by Fun Fon Chu, who is the operator right now. It has taken some time in the designing. It is capable of searching up to two million chess positions per second by itself, so it is capable of quite strong play from a single chip. When you combine them together in a multi processor like we have, that is a very difficult programming problem.
It has been a problem in research science for 20 years and it is not easy to solve. The problem is you end up not getting the maximum speed that you think you would be capable of getting. If you add up 256 times two million, that is up to almost 500 million, more than, so it is very difficult to get that kind of speed. So one of the problems in a program like this is to get an efficient use of those processors.
As for the second question, what the plans are, we just have to sit down after this match and reevaluate where we are going.
One of our goals for this project was to take what we have learned from building this machine and apply it to other problems as well, and so I think we are going to begin doing that, but I think DEEP BLUE is still going to be improved in the future.
Mr. Seirawan We will invite just a few more questions.
(Audience commentary.)
Mr. Seirawan The question is, if the computer program is able in the future, better luck next time, by the way, to get to the stage where it will be equal to everybody, will that be the death of chess?
MR. CAMPBELL I don't think so. Personally I believe that when I play chess or -- and I know that there are grandmasters who could mop the floor with me, but that does not distract from my enjoyment of chess, so the fact that there is a computer somewhere in some research lab -- that could theoretically beat every human in the world I don't think will affect how people enjoy chess. And it provides for possibilities of advances such as that that brings the --
MR. RIGHT Sorry to interrupt, Murray.
MR. CAMPBELL Maybe I should get going.
Mr. Seirawan Murray Campbell.
(Applause.)
Mr. Ashley Murray is going back to see what is going on in the other room because he is concerned about what is going on with his baby. Something is certainly happening. The operator talking about something happening on the screen. We have seen technical glitches like this before and they have been resolved expeditiously in every single case.
We see here arbiter Mike Valvo, who is overseeing the event, standing right now behind the operator and he is making sure nothing is wrong.
DEEP BLUE has been thinking a very long time on this move, which is not typical of DEEP BLUE. DEEP BLUE usually comes out with three minutes a move. But so far we have had a very long conversation and DEEP BLUE has still not moved. You have to wonder if anything has happened, or if DEEP BLUE has been affected by the phone lines. That has happened before. The weather plays a part.
At the moment after the move A3 attacking the bishop -- we should say that DEEP BLUE is not in Philadelphia. DEEP BLUE is right now in Yorktown Heights in New York, so the problem of the phone lines connecting now to the monitor here pops up as an issue. Sometimes things could happen. Whether or not Garry has people working in back, we might have to go to DEEP BLUE's laboratories if we see some of Garry's friends doing something. I doubt it.
But certainly DEEP BLUE has spent a very long time in this position. And I was extremely concerned about this position as a matter of fact, Yas, because this possibility of bishop takes A7 check is a serious one, and if -- we saw it happen yesterday. DEEP BLUE saw a win that we hadn't. We knew that Black's position was getting very, very good, and DEEP BLUE obviously saw a win that we hadn't seen and it proceeded to weaken his position in such a way as to make the win, in our minds, easier even though it prolonged the game.
Mr. Seirawan As you said, it saw a win that it wanted to avoid and the cure was worse than the disease. It ended up playing a weaker move.
Mr. Ashley To our eyes.
Mr. Seirawan To our eyes that made it easier for the human to win.
I'm thinking about what DEEP BLUE has been thinking about for so long. It is obviously seeing so many billions of positions, it may have gotten worried.
(Laughter.)
Mr. Seirawan What the move A3 has done is cast a spell towards the bishop on B4. The bishop on B4 always had that covert ability to come back to the F8 square as a measure of defense. Move A3 may push that bishop away from a possible kingside defense.
So what I was taking a look at is, in this position, if DEEP BLUE was to play the move bishop to D6, bishop to D6 -- I would like to look at a sacrifice, C-5, again the intent being to drive the bishop away from any possibility of defense to the king.
So C-5 forcing the bishop to retreat, queen D3 is a possibility now, but again the sacrifice bishop takes H7 check is still an attractive one. There is King takes H7, knight G5 check. That would be the 23rd move. King G8. Queen H5.
Woops. What did I do? Help me. With A3, this is a new variation, Maurice. Sorry. He did the same thing that I did. We are learning through this match as well.
Murray Campbell has in the meantime slipped off of our stage and gone to be the computer operator, and this is the -- this would be a position after the move queen H5. I'm anticipating the move knight to F5 thereafter, queen H7 check, king to F8, and once again this kind of little nasty move rook F3 snuggling up to the King side.
This could be something. Again, I'm just looking at this position quite briefly as a possibility. DEEP BLUE is not. DEEP BLUE is looking at this position as an absolute probability and making exhaustive calculations.
For example, there is a very easy mate in the position, just to show how dangerous this position is if the move knight takes D4, bishop takes D4, knight takes D4, queen H8 check, King E7, rook takes F7 mate.
That is move 29, but it began after move 20, A3. So it is a nine-move combination, which makes it 18 ply, which was perhaps a little beyond the computer's event horizon.
Mr. Ashley It depends. In some situations the computer will look that deep. This is the danger we talked about before. It can happen. It can happen that in some situations the fact that the computer can see so deeply makes it play worse moves now.
The point is that as it is sitting calculating, let's say it sees a force mate in ten, let's say it sees it. Ten moves down the line, it whipped out this variation and said, you know, if I play such and such a move, which may seem to us very natural, I lose by a move force mate in ten.
It is not going to play that move. It will back up and look. And if it sees the force mate in ten coming, it is going to say what can I do to stop that mate. That may involve giving away a piece now. It may never have occurred to Garry that the checkmate was in the position, but since the computer sees it, it will say let me give him a piece, and Garry will say thanks. For no reason at all.
And, of course, Garry will be able to convert that piece to a win with no problem. But because the computer sees so deeply, it will make a bad move, which to us seems like a bad move now. That happened yesterday. It made this move F3, king G2 and F3 that seemed inane, but it had apparently spotted a win for Garry. Easy win as far as it is concerned, losing pieces in a worse way, so it came up with this move F3, which to our eyes was completely ridiculous, and Garry was happy to win a piece by force in that game. He played very easy moves and couldn't piece.
The way to win was very complicated because we had not seen it yet. Maybe Garry would have come up on the win as the position proceeded along, as he increased his position, but maybe he would not have.
And a human player would have gambled. A human player would have said, well, let me go for some counterplay and see what happens. Let me mix it up. Try to confuse the guy. The computer does not play those kind of games. The computer said that's a win. I see it. So my opponent sees it, so therefore let me make some random move.
If the computer has spotted this sacrifice bishop H7 check, if this is what happened. We don't know if it indeed spotted that. If it retreats the bishop to one or two natural squares it loses by force. It will make a random move right now. Thankfully, it hasn't.
And finally it has made the move bishop to A5. And that move is a very natural move, now going to the D6 square, where it might have created some congestion, but that was a possibility.
We hope -- if they in the operating room, the folks in the operating room can hear us, we hope that someone will come back to the stage and brief us quickly on what actually happened just for this last few minutes that DEEP BLUE has been thinking. Someone will be here shortly to let us know what was going on.
But after much thought DEEP BLUE has finally played the move bishop to A5, and Kasparov is at a crossroads. He has very simple moves. A move like B4 is screaming to be played, chasing the bishop back, followed by C5 is definitely a very natural move in the position. Ideas of B5 certainly in a position.
So Kasparov with a very solid and forceful way of playing this position. The expansion, Yas, on the queen side as well as the speculative way of sacrificing a bishop, one would expect Garry not to sacrifice in a crazy way, but instead to play this other continuation which looks very powerful.
Mr. Seirawan I think Garry is being tempted by a multiplicity of possibilities. There is knight G5 and a very nice move queen D3. He has a lot of riches, if you will. And I notice that once again Garry has slipped behind on the clock. We have only played 20 moves. He currently has spent an hour and 12 minutes, and he will now get into his 21st move.
Mr. Seirawan Let's welcome Dr. Feng.
DR. FENG I went back and Garry is thinking I should wait there until I finish the moves. The monitor that we are using is an energy saving monitor so it went blank. So I went back and I am typing the move, A3 and somehow the A did not get recognized, it recognized it as 3. And it was a command to process number 3 and then I type F3 and the machine say I see a repetition and then say there is a beep. And it stopped beating after that. And then I check with the guy in the back room. They are saying that could cause the program harm and I checked with Valvo and we restart the program.
Mr. Seirawan So they had to restart the program and get it back up to speed. How much time did that cost the computer on its clock?
DR. FENG About ten minutes. And those are lost.
Mr. Seirawan Just to be clear, stay with us please, ten minutes to reboot everything and then ten minutes of lost thought time, so DEEP BLUE got hit with a 20 minute --
DR. FENG It would have played the same any way.
Mr. Seirawan It got hit with a 20 minute penalty, it seems to me.
DR. FENG It is expecting him to sacrifice.
Mr. Ashley Would you?
Mr. Seirawan Garry is not hearing us, but does the bishop to H7 sacrifice in DEEP BLUE's opinion work out to white's favor.
DR. FENG Would carry to a draw.
Mr. Seirawan It calculated it to a draw to sacrifice so it is not losing for the computer but it is good for you to get a draw. 21. b4 Bc7.
Mr. Ashley Kasparov has not sacrificed. He played the move B4, but the question is asked how do you know that the computer was planning to sacrifice on H7?
DR. FENG It printed out what the variation was going to be.
Mr. Ashley So that is actually on the screen that he was looking at. It's telling the variation the way Fritz does here. And the sacrifice of the bishop that we were talking about may not win but at least to a forced draw and that would be good for Garry. But maybe Garry thinks it is not good enough so he has played instead B4.
Mr. Seirawan I have another question.
Your colleague, Murray Campbell, was giving you the lion's share of the credit for the new chip that you more than anybody were responsible for its design and imprinting. Could you speak a bit how this chip differs than say, other chips that everybody might be used to and what was the breakthrough?
DR. FENG There is really no other chips. I'm the only guy that designs this kind of thing.
The chip that I have is a complete chess machine. If you look at old design, it is more like bell in safety and so on.
Mr. Seirawan Could they buy the chip that you created? I would like to buy one, that is for sure.
DR. FENG It is not in print. I don't know what is going to happen. That is not in my area. That is not a decision for me to make.
(Audience commentary.)
Mr. Seirawan She would like to know is DEEP BLUE limited on the amount of time it can spend on a move?
Mr. Ashley Before you answer that, DEEP BLUE, after the move B4 has played bishop back to C7.
DR. FENG Usually the time, we average the time that is for the time controls, so given the time control for this match, in the first 40 moves it was spending about three minutes.
There were cases that he would spend more time. When he accumulated move time, because opponent had been guessed, he would spend more time. And the other thing is when he find a new move that looked likely to be good, he might spend time to verify that that is the case.
And another possibility is the panic time that happened in yesterday's game. When the machine sees it is in deep do-do, it will spend extra time as well. 22. c5 Rde8.
Mr. Seirawan I'm going to allow the audience to ask you two more questions, but we will say that Kasparov after DEEP BLUE's 21st response bishop C7 has finally at last gone for the big clamp. He has played 22 C5, which sets his majority up on the queen side. And it will continue to press black's pieces on the queen side.
Now, you note that black's queen on D7 and the rook on D8 are rather inactive because that pawn on D5 is in the way.
Mr. Ashley Black's pieces are suffocating in their own camp without too many moves, and DEEP BLUE will have to come up with a way of breaking out of this lock, this prison that has been created by its own pawns and white's own pawns and we will see how it will try to do that in a few moves.
(Audience commentary.)
Mr. Seirawan The question was --
Mr. Ashley Before you do, DEEP BLUE has responded quickly with the move rook from D8 to E8.
Mr. Seirawan The gentleman says that he uses program trading for commodities and they will look at all kinds of things. Moving average, data bases, historical data, trends and pattern recognition and things like this, but does DEEP BLUE look at any of those other things or it is just solely a calculating machine?
DR. FENG It is based on pattern as well, the calculation, you would recognize because in future that is important. You would recognize whether you had a passed pawn or good pawn structure and you might take into account your material imbalance. And so on. And you calculate all of those numbers together.
In the end you do have number crunching but before you get a number, you look at the patterns first. 23. Qd3 g6.
Mr. Ashley Kasparov has played the move queen to D3 setting up a thinly veiled threat. Queen takes H7, the old one mover, and I'm sure DEEP BLUE will spot this.
(Laughter.)
(Audience commentary.)
Mr. Seirawan The question is does IBM have any plans to challenge other, let's say, lesser grandmasters than Garry Kasparov in a similar match?
DR. FENG There has been some talk about playing the Israeli national team a long while back. I don't know if it is still on or not. The thing was delayed because we didn't get machines up and running at that time, so it might still be going on and some other people are interested in playing some other matches and things like that, but I don't know the detail. It is too early to say anything specific.
Mr. Seirawan I would like you to join me in thanking Dr. Feng for joining us.
(Applause.)
Mr. Ashley DEEP BLUE has responded quickly to that move by playing G6. All of Black's pawns are on light squares and not a single one of them controls any dark squares. And I can envision this white knight in its greatest fantasy on the F6 square.
(Laughter.)
Mr. Ashley If knights could move like that for one move.
Mr. Seirawan Knights do dream, by the way. I must say that I was very impressed by the computer's last move -- pardon me, the last would moves by the computer. 24. Re2 Nf5.
Mr. Ashley Before you go on, Kasparov has played very quickly and responded with the move rook to E2. So it seems that he knows full well that he is under a little time constraint. He should move a little quicker. But still, rook E2 a very interesting move.
Mr. Seirawan In the position after 22, C5, the computer played the move rook to E8. And this was a very, very good move because what has happened is Black's pieces are getting all bunched up in the center. And most especially this bishop on C7 has little to say.
So I think the computer had anticipated that Garry was going to play. So the move rook E8 was very nice because now at least he opened up a possibility that the bishop might be able to come back. So kudos to the computer for that. The move queen D3 threatened the instant queen takes H7.
We saw G6, and now -- I did it again. We just get our position set back up here. There came the move A3 at the point, yes. Queen D3 D6 was played. Garry stepped back with the nimble rook E2. The move rook E2 was in anticipation of the move DEEP BLUE very quickly played, knight F5. The move knight F5 would have attacked the move on E3. So Garry made a preventive move before the computer could make a threat.
Now we see the move knight F5. And again, Garry has a very nice ideas here. What he would like to do is to take advantage of these weak side squares. The move knight F5, a provocative move by the computer. Kasparov could certainly consider a move like G4, trying to push the pawn up the board. He could consider a move like H5. The idea of H5 would be to exchange pawns on D6. Or he could simply try to clamp down on the position with a move like rook E1. 25. Bc3 h5.
And indeed he has decided to go for activity. Kasparov has played the move 25, bishop C3. We saw this especially, Maurice, in game five where Garry connived his way to utilizing his majorities very wisely. And the move bishop C3 appears to me to be setting up the move B5 and continuing to gain space on the queen side.
So bishop C3, a very quiet little move saying I have the position well in hand. You, the computer, are going to have to find a way out of the bind that I have you under and what are you going to do about it?
Mr. Ashley Yas, I know you have written many instructional chess books in addition to being editor of Inside Chess magazine.
Mr. Seirawan Commercial break.
Mr. Ashley I have to give you credit for that. In addition to being editor, where you have written some very fascinating articles on the topic of space, you have written some basic books, the Winning Chess series. And one of them talks about the four principles of chess, but space is the one that seems to be predominating here, isn't it? Garry has such wonderful spacial control on both sides of the board.
Mr. Seirawan Exactly. As Maurice was saying, in the earlier works that I wrote, I wrote a series of books to help people understand the various phases of chess. And one of the primary skills that a beginner has to learn to play a decent game of chess is to understand the elements.
There are four. The first and most important one is a material. When are you a begin game and you are playing and somebody asks you how are you doing, you say well, I'm up a rook. That answers the question. I'm winning. I'm up rook, up a pawn. So being ahead material is a very important element in chess.
The second, of course, is development, bringing out your pieces. Thinking in chess, in war, like terms you have your army, so you should deploy your army, develop your army. If all of your soldiers remain in the barracks, they are not going to be be effective. So thinking in warlike terms, development, time, very crucial elements.
The third element that helps us devise our plan is pawn structure. In this position what I mean by that is, white has a lovely pawn structure on the center queen side.
We do have a move by DEEP BLUE, so I will interrupt myself. H5 has been played, and that is interesting. That was a very pregnant pause, Maurice.
Mr. Ashley The move has its point, the main point being to prevent Kasparov to be able to advance a pawn. So from that positional vantage point, it looks reasonable.
But we have been talking about dark squares. And all of White's pawns -- or Black's pawns, I'm sorry, on that right side of the board are really showing that weakness of not controlling the dark squares. All of them on light squares, only one pawn in the whole position on a dark square, where it can control a dark square.
So Kasparov has to be thinking about how to exploit this weakness. Maybe it is not possible because those pawns create a fortress. I can envision some future possibility of that knight being traded off the board for White's bishop, and the dark squares around Black's king could be a cause for concern.
Mr. Seirawan What we are seeing here is an interplay of pawn moves, as Maurice was saying. The pawn on H5, it does have the benefit of preventing the move G4, so securing the knight's position.
Let me complete the idea of elements. So you have your material. You can win the game by just getting ahead on material, trading off the pieces and dragging your opponent down in an end game. You could win the game of chess by quickly developing your pieces, opening up a possibility of a rapid attack.
The third way that you can beat your opponent is through the use of pawns and pawn structure. We have talked about majorities and talked about how you can have a double pawn formation or even a triple pawn formation.
And the fourth element, and one of the most misunderstood elements, is space. When you see two boxers in a ring and they are fighting it out, one of the things that the trainers like to see is his boxer standing in the middle of the ring because, by controlling the center, he has more options of which way he can move| left, right, center, forward or even retreat.
If you think about the chessboard, it is a 64-square board. At the start of the game it is 16 pieces versus 16 pieces and 32 squares for white and 32 for Black.
We have a move. 26. b5 Nce7.
Mr. Ashley Garry has proceeded to control even more space on the left side of the board, playing the move B5 attacking the knight on C6. That knight will move either to E7, possibly, possibly to the A5 square, where it might thinking about going to C4. So two options for this knight, either A5 or E7.
Mr. Seirawan When you start the game, I draw your attention to the middle screen. You have to draw an imaginary line. The lower rectangle from the squares A1 to A4 to H4 to H1, that comprises White's space. So White literally, his space is this space, the first four ranks starting from the back rank.
And Black conversely has the other half of the board. If you note, what happens, White controls with this pawn, the C6 square and the A-H, and with the C5 pawn. And the D5 pawn controls the the E5 squares.
If you count up all the squares that White's pieces control, you will discover that White has a large advantage in space, and why is this meaningful.
Well, when I have more squares that my pieces can move to and you have fewer squares, you are going to get cramped, squeezed, and literally, as in wrestling terms, I will get a headlock on you, a Full Nelson, and you won't be able to move.
So a wonderful stratagem against computers is to take as many squares as you can, totally limiting the computer's possibilities and suffocating the computer.
Mr. Ashley Kasparov understands space better than anybody in the world. He is not thinking about the attack.
Mr. Seirawan Just about every game Kasparov has had the opportunity to play flashy chess and the first game made him decide resolutely, I will not carry out some crazy attacking idea.
The computer has played knight from C6 to E7, but if you look at that, we were talking about that attack on the right side of the board. Kasparov has allowed the computer to completely repel that attack. Now it has set up such a strong position that we are not really thinking about that as familiar anymore. What we are thinking about is the wonderful space that Kasparov controls on the other side.
So Kasparov is playing more or less strategically. It may turn into a big attack. But this is not a position that Kasparov is thinking about losing. I would say A4 just getting more space.
Mr. Seirawan Continuing to march up the board.
Mr. Ashley It is going to be tough to find moves for black. The moves have become more difficult because you don't have anywhere to move.
Continuing your boxing analogy, it is like having a guy on the ropes. If you are on the ropes it is tough to get around the guy. You are going to have to fight your way out. It is tough in a chess game to do that because the other side has the same kind of pieces you do so you can't overwhelm the other guy with the same amount of force so since it is equal force.
The guy that has the other guy pinned in is always going to have an edge. And here Kasparov knows he has DEEP BLUE stuck on two ranks. Only one position beyond the third rank and it is the knight on F5. But if you look at Kasparov he has three pieces beyond the second rank and he could bring more pieces if he felt like it because his pawns are not in his way. They have been moved up and are not in his way. He could keep controlling more and more space in the position.
We should make a note, however, that Kasparov has 38 minutes left on his clock. He has to make his 27th move, yet to make his 27th move, which means that he is just fine as far as the time is concerned.
Mr. Seirawan The only problem is really sometimes, you know, you walk into a video show, a video store, and you have 20,000 titles to select from. It takes you six hours to make a decision and you miss the movie. It is sometimes better to walk into a video store that has two movies, you check one out and you are on your way.
Garry has an assortment of moves, really an embarrassment of riches. He could go left queen side or right king side. He could play in the center. And although he has a lot of time to make such decision, sometimes it is quite difficult because all the decisions are quite compelling. He has a lot of good moves to decide from, choose from.
I have just suggested the simple move A4. If we go back to the position on --
Mr. Ashley I don't think this computer likes the humans today.
Mr. Seirawan It doesn't like me.
I was going to say if you go back to the position at the board white's queen on D3 protects the pawn on B5 and therefore the queen is a bit tied to the defense of that pawn. So I was going to suggest that white play the move A4 and this would open up the possibilities of the queen moving.
I would like to invite once again some comments, questions from the audience.
(Audience commentary.) 27. Bd2 Kg7.
Mr. Seirawan The gentleman has two points.
Mr. Ashley I'm sorry. Kasparov hesitated for just a moment before playing the move bishop to D2. I should say he hesitated for just a second. His hand went over the bishop as if to say I'm a little uncertain, and then he grabbed the bishop and played the move bishop to D2, so I think Kasparov senses that he has all of these plans in the position, and he didn't want to make a mistake right now. I would imagine that he is enjoying the position as the spectator, Kasparov is enjoying the position. But he has to be nervous too. And the other part of the question, was maybe you can explain the concept of zugzwang.
Mr. Seirawan Whenever Garry or a Grandmaster has a good position you definitely enjoy the position, yes. There is no question. There is no feeling like being on top in a crucial game of chess especially a game like this which is a $150,000 swing on the game. The nervousness factor has to be there. Garry does not want to make a mistake, so you did, as Maurice pointed out, have the slight hesitation on his last move.
Zugzwang is a German word and the phrase itself means a compunction to move. Imagine, if you will, that you have a position that you are worse. You are getting squeezed and it's a tough one but finally you found the ideal setup. You found what you consider to be a fortress, so you set up all of your pieces in their ideal position and you think great, now my position is invulnerable. And then your opponent makes a move that is like a pass. Your opponent passes. And then suddenly you realize that you are expected to move. You are in a state of zugzwang. You must make a move but all of your pieces are on ideal squares.
And because you must move, you have to move one of them off the ideal square. In this position, Garry has a very, very nice camp on the computer and he is definitely trying to bring about a state of zugzwang into the computer by limiting the amounts of active moves for the computer, squeeze the computer.
The last move, I will just speak to it very briefly, is the move bishop D2. We have spoken about the pawns on the king side, black has pawns on F7, G6, H5 making especially the square F6 vulnerable.
Mr. Ashley DEEP BLUE has responded with a move king to G7, shoring up the dark squares somewhat but it is never a pleasant task for the king to be the one to defend weak squares because it is so vulnerable to dual attack because it must run. This king definitely helping out bringing everything to bear is DEEP BLUE. This variation you are analyzing on the sweep, bishop G5 is possible, but F6 is a playable alternative and thanks to that king, that pawn is not going to go for free.
So DEEP BLUE with the move king to G7 playing very quickly.
Mr. Seirawan And a very natural move. I think the human being would have played king G7 very quickly as well. So kudos to the computer for playing what I think is a very good move. And I also might say that the king move also gives this knight on E7 a little bit more option.
If we go back just one move, please, yes. This knight on E7 has virtually no squares it can move to. All of its squares are blocked.
Mr. Seirawan The move king G7 has opened up the possibility. Very good move, king G7. 28. a4 Ra8.
Mr. Ashley Kasparov has responded very quickly with the move A4, continuing the march of the pawns on the queen side. And the more you look at it, the more you have to like Kasparov's spacial control. He is getting up on every move now. He knows what is happening in this position. He is not afraid of any surprises. He has the computer squeezed in a box. And there is no reason, as far as he is concerned, to think too deeply about the options in this position because Black has no real source of counterplay.
We use the word counterplay a lot in chess. You always want to have counterplay.
Mr. Seirawan Look at what we have seen. Again a 30-foot shot by the computer.
Mr. Ashley I'm not sure if this is a 30-foot shot or not, throwing the ball out of bounds. This last move, rook to A8 -- and we talk about the computer's poor play in close positions. I have to see Kasparov's face when he comes back to the board because rook to A8 is just an admission of really having nothing at all to do and not having a clue as to how to proceed.
This is what we are talking about, running out of moves and being forced more and more into a situation where you just really have nothing to do. You are being suffocated, so you start making random moves back and forth.
Rook to A8 was played, and Kasparov is fully aware that the computer is at a loss for moves and now sitting up in his chair. He won't give anything away as far as this last move is concerned, but he knows that the move really doesn't have too much of a point, although maybe, maybe we can give the computer an idea.
For example, the move A6 might be threatened. Let's say we make a random move for Garry like king to G2, for example. Now let's take a look at the move A6. This might be the sole source of missing counterplay, opening up the A file so that that rook can get into the game, Yas, so maybe we have jumped the gun and saying the computer has made a meaningless move. It needs counterplay, and this move goes in that direction.
Mr. Seirawan Indeed. The move rook C8 to A8, the reason that I would have ridiculed the move is twofold. First of all, the rook is perfectly fine on C8. Rook to A8 is what we might call an undeveloping move. You are literally taking a move back.
The other point is, we talked about majorities. So often it is an awkward concept to teach. When I have a majority, I have an advantage because I have an extra pawn on the flank or side that I'm attacking on. So that means that I have more muscle in that part of the board, if you will, than my opponent. That gives me an in-built advantage.
If I can force my opponent to only play on the side where I have an advantage, I'm going to win the game. So when the computer plays a move like rook A8 with the intent of playing A6 and at least getting an active rook, I can say, well, bravo for you, but at the same time good for me because you are doing exactly what I want you to do.
A simple response for Garry would be a move like queen B3 as a new variation and then, after the move 29, queen B3, invite the computer to go for its plan of A6. Now I would bring my bishop to a square like D3. And after they exchange A takes B, A takes B, I am really happy because all of the play, if you will, is occurring on the side of the board where I have an edge.
So although the computer might not have been able to find the plan in the position, or at least it has found a plan, the plan in the long run will work in Garry's favor.
The plan I would have chosen had I been in the computer's shoes -- we do have a move by Garry. Instead of the move rook to A8, I would have moved my knight back to G8. So let's see what has happened. 29. a5 a6.
Mr. Ashley After rook to A8, Garry has played the move A5. And boy, those pawns look nice.
(Laughter.)
Mr. Ashley A5, B5, C5. You know, really there are a lot of correlations. We use them a lot between chess and sports. And one of the things about pawns is they are often what we consider blockers. The pawns in front of White's king tend to help his king.
If you think about a quarterback in football, a quarterback likes having blockers in front of him. He does not want to see his defensive line running down the field because the rest of the guys will come get him.
The king is the same way. It likes to have those pawns in front so it can have protection. But on the other side of the board, you can expect to see some White pieces act like runningbacks and follow those blockers all the way down the board. Those three pawns just ready to open up lines for pieces, and the White pieces will just flood into the position behind them.
So Kasparov with a nice flanks of pawns, beautiful, A5, B5, C5. An overwhelming advantage for Kasparov. A position he can play easily on the wing that he wants to be on. And on every single move we have seen Kasparov getting up from the board. No reason, as far as he is concerned, to sit at the board and waste his time. I wonder if he has HBO in back.
Mr. Seirawan There is a good show going on back there.
Your last comment deserves comment. I love watching football, and from time to time the commentators do say that they are putting in new players onto the pitch. And they say, well, the two coaches are really engaged in a game of chess. I have been dying to hear a sports commentator say you take a game of chess and make it a football analogy, which you just did. And that was wonderful.
You are absolutely right. The three pawns have acted like 300-pound offensive tackles, pushing their way up the board and forcing Black's pieces away.
Mr. Ashley It is like the Cowboys, bump you off the line.
Mr. Seirawan In the meantime Garry will be able to bring his pieces behind those three brutal passed pawns on the queen side with a move like bishop to C2 and bishop to A4 trying to play moves like B6 and C6.
Let's take a look, if we will -- although this is a very handsome position for the computer to play, at least it should try to move A6. I think Garry has leaped over C6, a nifty little pawn sacrifice like C6 or the direct B6. After A6, B7, bishop D8, Black is not a happy camper because that rook on A8 is locked in for life.
So we have to go back and put the bishop back to D8. Okay? New variation. And the intent here is at least one thing that Black could try. And we have, by the way, seen the move A6. DEEP BLUE has played 29, A6. And now we were just looking at the possibility of -- should I hit that button?
Mr. Ashley Yes. Bishop to D8 is what you wanted.
Mr. Seirawan Bishop to D8. At least in this particular variation Black has one thing that he can hope for. He can try to bring his knight to this C6 square, where it is a wonderful blocker. Okay?
So Garry may have been -- may be prepared to play a sacrifice. We go back to the position at the board. We have just seen this move A6, and Garry may be contemplating a move like C6, inviting the sacrifice B takes C6, B6. In this case the bishop could come out here to the D6 square, and Garry may -- Garry has created a protective passed pawn, but he has sacrificed a pawn.
Mr. Ashley This is a sacrifice he will have to judge very carefully because going down a pawn in a game of chess on the Grandmaster level is very -- is a very serious thing to have happen to you, and it better be really good or otherwise it is going to be really bad.
(Laughter.)
Mr. Ashley How is that for a political statement?
Mr. Seirawan The idea being a positional sacrifice, this is one of those sacrifices that grandmasters are have famous for. They don't win immediately, but they have this long, long-term pressure.
And this is the type of position that Kasparov could outplay the computer. Because again we have spoken about that event, horizon, the computer is only looking maybe ten moves deep, or so forth, and it won't be able to see in the long run a sacrifice like C6 could cinch the game.
Mr. Ashley Kasparov will spend some time on this position. The time he is spending now, really wondering to himself how he should play this position. The move B6 has its drawbacks. As you said the C6 square would be very weak. It has its pluses certainly because of the conquest of the E5 square, and that has to be a big question. For example, if B6 in this position and the computer were to play the move we anticipate --
Mr. Seirawan Bishop to D8?
Mr. Ashley -- bishop back to D8. We didn't want to go to B8 because that rook on A8 --
Mr. Seirawan Is locked out of the game for ever.
Mr. Ashley Bishop to D8 does have the singular drawback of giving white the square E5 and now white could play a great move knight to E5 jumping into that square, controlling a lot of squares in the enemy camp with the knight and attacking the queen.
This queen could go on a crazy raid by playing queen A4. This kind of move we consider stupid.
(Laughter.)
Mr. Ashley The reason for that is you cannot begin to attack your opponent with only a couple of pieces. If you look at the situation, black has two pieces. It plays the queen on A4 and the knight on F5, the knight on F5 being the best piece. Your queen goes into enemy territory where there is a queen, a rook, mother rook and two bishops laying in wait.
You know what is going to happen. So this move wouldn't be advised. Instead though, a move like queen to C8 is looking horrid. The reason for that is black's pieces have really been pushed back in the worse way and Kasparov has gained so much space in this position that you wouldn't be surprised if there is a knockout blow. 30. b6 Bb8.
And Kasparov has played the move B6, putting the question to that bishop where is it going to go. Is it going to go back to the B8 or D8?
Mr. Seirawan Whoa.
Mr. Ashley Instantly not allows the move knight E5, DEEP BLUE has played bishop to B8. And we were discussing the possibility of that, but DEEP BLUE thought that way too dangerous. 31. Bc2 Nc6. But now we have to wonder as Kasparov has quickly played bishop to C2 eyeing that A4 square, so bishop thinking about going to A4 and just with some nice attacking possibilities down the line. Obviously he can't go there now because it would need protection, but bishop to A4 would be instantaneous. That rook on A8 -
Mr. Seirawan Looks bad.
Mr. Ashley He has to be wondering what did I do?
Mr. Seirawan Basically the way I would look on this is the rook on A8 has been played in a coffin, and that bishop on B8 is the lid. And unless that lid gets lifted, that rook will remain out of the game forever.
Again, this is the problem with the computer. Its event horizon. It sees that it is under pressure and it doesn't understand that there may not be a win for white in the next four to six, eight moves. But in the meantime, Kasparov's perspective is I will win it in 20 if I have to, but as long as that rook remains in the corner, black will be playing a rook down.
It will be very simple for white. All he has to do is maintain the control of the C7 square, the D6 square, the E5 square, and the F4 square and that bishop can never move, Maurice. It is going to be stuck there for a long time.
Mr. Ashley The bishop can't move, the rook can't move. I remember a friend of mine that used to sacrifice on the other side of the board, and he would say, But I'm not really sacrificing because two of your pieces can't move. You are playing without those pieces. So I'm really up material. I'm using free material to mate you with.
So that is what you are looking at here. The bishop and the rook out of the game. So what if you have them.
I always use this example with my students. I teach these kids up in Harlem, and I have a lot of fun with them and I like to show them when pieces are undefended. And I show them a famous Morphy game when Morphy played in Paris at the opera against Duke and the Count. And the game is a classic showing what happened when you don't have a couple of pieces in the game.
The example I tell them is what if we have a fight and I had to fight you and I had on my side Arnold Schwartzenager and you had on your side Peewee Herman. It looks like it is going to be a done deal. But if you could somehow get Arnold Schwartzenager locked in a steel room where he could not get out, theoretically I have him on my side, but we cannot get in the fight.
So I can boast about having him but he can't lock me out because he is locked in. Maybe I will still beat up your Peewee Herman, after all, but the fact of the matter is that big advantage of yours is out of the game. That bishop on B8 and that rook on A8, yeah, you see they will. They are existing on the board. They're taking up space, but they are not doing anything and won't do anything for a long time to go so they are out of play. 32. Ba4 Re7.
The move knight to C6 has been played and Kasparov instantly responded with bishop to A4, so this zugzwang that describes a compulsion to move. I'm not sure it is going to get that far because looking at this position here, black is seriously running out of moves.
Mr. Ashley Black can't move the rook, can't move the bishop, can't move the knight on C6, so basically black has only three pieces, the queen, rook and knight on the other side and white can move everybody.
Mr. Seirawan Exactly.
Mr. Ashley That can't be very good.
Mr. Seirawan This is what makes Garry such a formidable opponent. He is very craftily, beautifully -- I consider the position almost to the point of desperaton for the computer and I'm suggesting a move that I myself do not like very much which is the move F6. This would create a major weakness of the G6 pawn, because now the only piece that protects the G6 pawn is the king.
But at least the point behind the move F6 is that black could try for some counterplay based upon the center push E5 and begin majorities coming into play with that move E5.
Mr. Ashley That move I think you think is hard to believe, opening up a position when you are behind in piece placement. It seems like the wrong thing to be doing.
There are some strange possibilities in this position. Kasparov bishop on A4 is momentarily unprotected. There was a suggestion in the audience to move bishop takes G3, which is sort of like, What the heck?
Mr. Seirawan The bishop is too bad, this coffin lid. It is a reasonable suggestion. However, it will have some very, very powerful tactical consequences.
Let's look at the move bishop takes G3 because it tries to lift the blockade. F takes G3 accepting the sacrifice, knight takes G3.
Now, again we have two pawns, Black has two pawns, but materially speaking that is not enough. I think Fritz will already be giving a big advantage, and it does.
We have a move.
Mr. Ashley Giving up those pieces. It would be one thing if you sacrificed the bishop, and will as some great followup. What do you have left? You have one knight in a position. If you are trying to move those pawns downward when white is the one with all the great pieces, it is going to be pretty bad for the side with the disadvantage.
Mr. Seirawan We do have a move.
Mr. Ashley DEEP BLUE has played the move rook to E7. Yas, this looks like part of what we were talking about, being squeezed in that box, running out of moves, not having many good moves to make.
And Kasparov did not get up from the board on the last move. He has stayed at the board, calculating. And part of that, I think, is the fact that he is below 30 minutes. And usually when you are in that red zone, below 30 minutes, you want to make sure that you know exactly what is going to happen.
He has 24 minutes and 16 seconds. He is now on his 33rd move. And now what do you think his plan of action should be, Yas?
Mr. Seirawan We just spoke about the possibility of bishop takes G3. Although we did not like it, a move like rook E7 is a way that Black is saying I can't figure out what I should be doing. I don't have an active plan. And he is kind of throwing the ball at Garry and saying, do you have an active plan? Garry is saying, sure. I will be happy to show you.
But what Garry could do with a move like queen G2 is one of those moves that once again it cements the advantage, cements the pawn on G3 and prevents daring sacrifices, good or bad, and it forces a decision.
I think the purpose of the move rook B7 at least was to get off this diagonal and in some ways threaten the move knight E5, breaking the pin. By getting the King up to G2, I avoid any kind of unpleasant checks on the F3 square.
So a move like queen G2 cementing the position and developing the pieces would be a very, very nice move. I think there are other approaches, but that is certainly a nice one for Garry to take.
Mr. Ashley I want to take a look at that sacrifice. This move knight to E5, it seems it may be giving away material because the knight is under attack by three white pieces.
But this is what we call a discovery in chess, a discovered attack. This knight by moving has done, unleashed an attack by the queen on this bishop on A4. So if you happily pocket the knight, thinking who is this idiot, then queen takes A4 could be a response.
And not to say that Black is any better in this position than the last position, but it is certainly a possibility to be thinking about and it is not clear that Kasparov wanted this to occur.
If you think let me take a queen, then Black could say, well, I get your queen. And with this rook under attack and this bishop under attack, White is going to have to come up with some moves quickly because a move like rook to C3, maybe a move like knight to C5, and this bishop will be taken on the next move.
So Black has a little trick. It does not happen often in chess position. You are merrily going along your way thinking I have this sucker cold, no problem. And then you play some quiet move and even on his back the other guy is still kicking. So White always has to be on the alert for little tricks like that. I'm not sure whether that is going to help Black's position in this case, but it is something to think about.
Mr. Seirawan This is the type of thing that makes chess an extremely cruel game. What I mean by that, to be a major star in the world of chess requires what I would simply describe as a killer's instinct. You really, really want to have to win, and you have to keep that killer instinct for a long time.
In other words, a track star may have his killer instinct but only for ten seconds and he pours it on, he or she pours it on, for those ten seconds and really does their best. It might be a 30-minute track event where they have that killer instinct.
But in these, and especially at the highest levels of the professional regulations time control we are seeing, some of those games go for five hours, six hours, and you have to keep that killer instinct the entire time.
If you let up your guard even for a single second, boom, your opponent, quick shot, wow. He is back in the game. You lose. You feel like an idiot. You are kicking yourself in the teeth. You say I was winning, I was winning. I could have been a contender.
Mr. Ashley How many times do you hear that at chess tournaments? It is the oldest song and dance story. A guy crying about how he was winning and somehow he lost, but that happens when you are not on the alert. 33. Bc3 Ne5.
Kasparov has instead played the move bishop to C3, a peculiar move.
Mr. Seirawan We will just call it interesting for now.
Mr. Ashley Bishop to C3, that move brings the bishop back to a square it was on earlier. It defends a pawn that was defended before, but it allows the queen to move away. The pawn was not under attack. So we are still trying to figure out why he did it.
It is a strengthening move. It definitely allows more of a control of the E5 square. And Black might be planning to play E5 in the future with a move like F6. We talked about that. But it certainly wasn't the move that we would have anticipated, not the most natural move. But certainly Kasparov just wants to strengthen the chain of his position more.
What if a move like knight E5, this crazy move, knight E5? What does it mean to the position?
Mr. Seirawan My take on the move knight E5 is that it puts him in this terrible pin. The bishop on A4 is getting him gummed up. So it is in Black's interest to try to dissolve this pin as fast as he can. And a move like knight E5 does it.
The White queen cannot come back to -- for example, queen D1 would be a horrible move because we do have this move knight takes F3 check. That was why queen G2 was an earlier suggestion.
But going back to the position on the board, if we see the move knight E5, I think what Garry has in mind is to try to just take overall the squares to play G4 takes E5, allow Black this move queen takes A4, now to play the move knight D4.
The idea is to encourage knight takes D4, queen takes D4, queen takes D4 4, bishop takes D4. We would have slipped ourselves into an ending, but an ending that is hopeless for Black because that rook on A8 is totally buried. The bishop on B8, the coffin has a lid and it is not going to be lifted, and White would effectively be a rook and bishop ahead. So Garry is taking the same strategical approach of space.
I like to say that space is one of the most misunderstood elements. If Garry can maintain this fantastic grip, it is like having a full Nelson bodylock. He will just thrash the computer. Any way, this is a variation and I think one that Garry most anticipated.
Another just subtle point of the move bishop C3, the idea is that Black has only one source of counterplay. This is the only chance to break the chains that bind Black's position.
So the move bishop C3 is a preventive move designed to stop the move from E5 from being played because after D take E5, the Black squared bishop, the C3 bishop of White, the Black squared bishop is perfectly aimed at Black's king. A prophylactic move.
(Audience commentary.)
Mr. Seirawan I would like to invite Dr. Hans Berliner up to the stage. He is one of the giants in the chess computing industry, if we will, please, applause for Dr. Hans Berliner.
(Applause.)
Mr. Ashley We have been up here for two hours plus so I will invite my colleague Grandmaster Yasser Seirawan to take a break.
(Applause.)
Mr. Ashley Hans, you made a very interesting point before you got to the microphone that most problems are not designed to pick up.
DR. BERLINER Their knowledge is in the form of what we call a piece square table, which says that a piece on a given square has a certain value, so, for instance, a rook in the corner there would be worth less than the rook in the center of the board if it could be maintained there.
So the rook is worth less minus most of the bishop. And in order to detect that, in order to detect the fact that the bishop is blocking the rook, one needs to look at two squares, not at one. In other words, it is not satisfactory to just say, oh, the rook on this -- in this corner is worth something and the bishop next to it is worth something.
You have to understand the fact that they are interfering with each other and you assume it may be that the A file is open so the rook can get out at any point. But if you let the computer know that that configuration is bad, it will do it and then you can be relieved and know that that is not a problem. If it can't do it you really need to know that that is a problem.
We have had this in there for about ten years and it is something that might slow the machine down a little bit but in a situation like this it would really be worthwhile to know that.
Mr. Ashley It is also possible that the DEEP BLUE programmers have this problem but they decided that playing the bishop to B8 was better than the alternative.
DR. BERLINER It is certainly possible that they have it and Murray Campbell could probably speak to that question if he were available here. It is certainly quite possible.
I would say most of the micros will not know about it and so it just depends on the degree of chess knowledge in the program and how important they think something like this is. And of course in a position like this it turns out to be very important.
Mr. Ashley We would like to invite you to ask questions.
(Audience commentary.)
Mr. Ashley Right now Murray Campbell is at the board. It is DEEP BLUE's turn to move. It is thinking about its 33rd move. It has 35 minutes and 23 seconds left. And here, interestingly enough, Fritz four has settled on the move knight to E5, the move we have been talking about. It has decided that the move knight to E5 is the best move for black in this position.
But the end game that Yasser pointed out, which may not occur by force, but that end game looks promising for white.
DR. BERLINER The more pieces that come off the more serious is this problem so that makes me believe that Fritz does not know about this trap.
Mr. Ashley We have Frederick Freiel here who is very responsible for Fritz, and I'm sure he would be raising up to the stage to answer that accusation, if he heard that one. So the question about again knight to E5. You are talking about a relationship aren't you, this relationship about rook and bishop, not the single analysis.
DR. BERLINER Yes, that's the idea. The thing is that there are positions where knowing a piece on a given square is just not enough, you need to know a set of pieces and how they relate to each other. And that's what we call a pattern and pattern detection is expensive so people say it seldom comes up and there is a whole philosophy of people that program computers that say it does not come up often enough for us to do so, so we won't do it.
Knowing that there are all sorts of positions of this type. There are situations where a bishop can get trapped independent of the rook also being trapped, like a bishop can get trapped at G3 sometimes by a line of pawn. And those kinds of things, if they are not detected by the program just assumes it has a nice bishop, but it is one that never gets out. It doesn't anticipate problems in getting it out.
Mr. Ashley There was something that I read in a book on this topic by David Norwood. The book was talking about chess and education in general but he made a point that there are some things that a computer can never understand, and that is something like the concept of never. He says that it is impossible for a computer to understand the concept of never because in order to come to understand what it is about, it has to calculate all the way to never and since it is never going to do that, then it will never understand.
(Laughter.)
Mr. Ashley That made a lot of sense to me as a layman. Is that true that a computer really couldn't comprehend the concept of never, although a human immediately gets what it means?
DR. BERLINER I have a background in computer science and I can say that that is not true. It is possible. It certainly isn't easy today to do something like that, but there is nothing theoretical that would keep a computer from being able to do this. The way we solve, the way humans solve problems like this is they go to a higher level of abstractions, so this whole idea of the rook being trapped at a higher level of abstraction, just like a pattern is a higher level of understanding than thinking about a single piece on a single square. When you go to a higher level of abstraction, you reduce the possibilities. You hope to understand the total situation. You know, the concept of never, if humans can understand it, then computers will ultimately be able to understand it too.
Mr. Ashley Why is that?
DR. BERLINER Because there is something called the theory of computer ability and there is nothing that humans can do that is considered to be outside of the theory of computer ability. That does not mean that it is not hard and it doesn't mean that it may not take a thousand years to do it but it is theoretically possible. And we tend to think in terms of what can we do tomorrow and it is not going to happen tomorrow.
Mr. Ashley What about this argument. Some people would like to know that the computer will never be able to play better than the best humans in the game of chess. There is an argument that at some point there is a ceiling on how fast the computer can go. There must be a point like we have the speed of light being the fastest that anything will ever go into the universe. There must be a point where a computer will not be able to compute any faster. And it will just stop and then you will have to look for some other type of algorithm. Where are we in seeing that?
DR. BERLINER I think we are still a long way away. You can go to parallelism and in theory I could play devil's advocate and say we can reproduce a human's pain and it can think like a human but whether that is desirable is subject to further discussion, because computers seem to be advancing so fast.
Mr. Ashley DEEP BLUE, after much thought, has indeed played this very fancy looking move, knight to E5, and it doesn't come as a surprise to us, knight to E5. In fact, we did consider and Kasparov hungry to rip that knight off the board.
(Laughter.) 34. dxe5 Qa4.
Mr. Ashley He even fumbled with it for a second.
He played the move D captures E5 and a forced move now for DEEP BLUE that we can anticipate is the move queen takes A4. I'm sure it will think for a brief movement before playing it.
DR. BERLINER Now the bishop is more constrained than it was.
Mr. Ashley Yes, since the bishop cannot move, the rook cannot move. So playing two pieces down is DEEP BLUE.
In the meantime, Fritz four initially gave black a significant advantage, a slight advantage, actually. Now it is saying the position is basically equal.
(Laughter.)
Mr. Ashley We are far, far from equal in this position. Right now black to me is down eight points.
(Laughter.)
DR. BERLINER So you can really see the effect of that pattern recognition can have in the right situation.
Mr. Ashley In this situation here, as you mentioned. The whole way to activate those two pieces, the rook and the bishop is to give up the bishop which I guess then would mean that the bishop, black is really down that single bishop. That bishop in order to get the bishop into the game is going to have to sacrifice. Fritz is changing his evaluation a little bit but not saying much different.
I would think, as you said, the more exchanges the merrier. If you can envision a position where all the pieces that is one pair of rooks, the one on E7 and another rook for white and the queens and the knights were off the board, black wouldn't have a move with those two pieces, just simply would not have a move and white could think about getting in very easily like with a move like C6.
This position looks overwhelming and completely lost for black in the long term.
DR. BERLINER There are positions where bishop would be at C8 and pawn at B7 and neither could move and those things would be locked in there rock solid for the rest of the game. And if you could recognize that you would actually be the full bishop and rook down eight points, like you were saying, and at least high-tech recognizes those situations. They don't come up very often so they don't bring in many points but intellectually it is good to know that.
Mr. Ashley We have seen DEEP BLUE think on occasion over moves like that one even though it is a forced move. As we said, sometimes it is hamstrung by its ability to think so deeply. For us we know queen takes A4, what are you waiting on, but it is calculating all sorts of possibilities that really are not significant to this position, but in some unique situations can become significant and I think that's why your programmers tell your computer to even pause on moves like this that are very obvious to us.
DR. BERLINER This is an ongoing discussion that we have in the field of computer chess. It so happens that in the approach that is being used by DEEP BLUE, with hardware there, it is possible to have many things happening at the same time, as we say, in parallel.
So it is possible to have a machine which evaluates the position normally, and sitting on the side of it are some little machines that are looking for certain kinds of patterns. And those things don't slow it down because they are working at the same time and they find a pattern. They raise their hands and say I found a pattern.
If they don't say anything, that means everything is okay. That is the style that we have used in the past. It does involve designing additional pieces of hardware, and it is cumbersome, but in a situation like this, it very much works.
Mr. Ashley DEEP BLUE has spent almost four minutes on the move and Kasparov not getting up from the board because he is anticipating this move. DEEP BLUE refuses to move. I saw Murray Campbell, the master, drumming his fingers.
DR. BERLINER It may have failed low. It may finally have seen that something bad is going to happen.
DR. FENG I think our problem is how we recognize the fact that bishop is trapped. It does not recognize that the rook is trapped. The problem is that we did not put in the value for bishop in trap, and between bishop B8. And D8 there is a ten point difference.
Mr. Ashley That is significant.
(Laughter.)
DR. BERLINER That is something you can fix, right?
DR. FENG It is a finite code, but it was not done. We did not have experience with bishop in trap values, and didn't get around to do it.
(Applause.)
Mr. Ashley Sometimes we wonder if it is the computer or the humans losing the game.
DR. FENG Both.
Mr. Ashley Certainly they have done amazing things with DEEP BLUE and there are so many issues. I know from talking with really wonderful minds, just having these two gentlemen on stage who are grandmasters, this is their field, it is a bit of an honor for me certainly.
It is a difficult thing. We have, as Yasser mentioned, a thousand years of chess experience. You think about something as basic as the human being's ability to pick up a cup, just a simple thing as picking up a cup or walking. For us it is so easy to do. We do it without even thinking. But to make a computer do that requires, my goodness, to do it naturally, so much computing power that it is off the scale.
DR. BERLINER As a good player you are always seem to be unaware of all the things you know, and then when you think you have told the computer a lot of things and it goes off and does something like this, then you realize that there is something you know that you still haven't told it. And as Dr. Hsu was saying, that is something they overlooked. And there are really lots of things that one has to consider putting in.
Mr. Ashley This is coming as a real surprise to us. Obvious move, queen takes A4, which we had supposed would have been played instantly, DEEP BLUE has thought now for about seven minutes on this move.
DR. BERLINER The explanation is fairly clear. It has to be this way, even though it is not my program, that normally it would make this move very quickly because it is the only move. So the only explanation that it is not making it quickly is that it has discovered there is a problem and it hopes that some other move will solve this problem.
But we see very quickly that there is no other move that can solve the problem. So it is essentially wasting its time. But this is just the way this search algorithm works. And because it does not really know that every other move, leaves it a bishop behind. I think that's what the real problem is. 35. Nd4 Nxd4.
Mr. Ashley So this is an interesting thing. As we said, we were able to ascertain this relationship immediately, the bishop looking at the rook on A8. It seems so trivial, why can't that just be put into a computer.
It is so easy, but these relationships that we consider simple -- DEEP BLUE has finally played queen takes A4 and Kasparov instantly responding with the move knight to D4, looking to change off the key pieces in the position.
Remember we talked about the knight on F5 being one of Black's best pieces, in fact Black's actual best piece. And one thing in chess, one thing known in chess strategy, is you trade off all of your opponent's good pieces so that your opponent will only be left with bad pieces. So you just look for the good pieces and say let me trade that off and that off, and then the inactive pieces will indeed suffer.
Here, that rook on A8 and the bishop on B8, two pieces looked out of the game. He is not going to try to attack immediately. He will try to trade off the good pieces.
For example, knight takes knight, queen takes knight seems like a very reasonable move. He would love this exchange of queens because if that happens Black would be left with only one aggressive fighting unit. And in the meantime White has three, two rooks and a bishop. That is going to be overwhelming, and Kasparov knows it.
DEEP BLUE quickly exchanging knights on D4, knight takes D4. Kasparov has two options, bishop takes D4 and queen takes. I think bishop takes will probably not play for the reasons we discussed and, in addition to that, the pawn on A5 would be hanging. So bishop takes D4 would be answered by queen tanks A5.
There is no real need for Kasparov to go in for these sacrifices, especially when he has a great possibility in queen takes, and has done so. Now that is Kasparov's 36th move. We are waiting now for DEEP BLUE's move. And we can say unequivocally, if queen takes D4, the game is just over, just over.
DR. BERLINER The thing that is interesting is that DEEP BLUE has almost noticed that when it played that, it was in deep trouble. We figured it out by conceptual means, but it figured it out by looking at long variations.
And so it is sort of interesting to contemplate how quickly it must have looked to see that something bad was going to happen. I don't think any of us would be ready to propose a variation where Black meets its demise. We argue on general principles, and that is sufficient, and it has to work that out in detail.
Mr. Ashley So the critical difference between the human being and the computer really being shown in full in this game.
And I guess that is why we are also fascinated by this match. A lot of people all around the world wanted to see this man very versus machine battle. It has been in the press. The New York Times, CNN was here yesterday. Everyone is so fascinated by this question as man tries to unlock the mystery of the human brain. And we have learned a lot certainly from the computer.
Before I take a question in the back, DEEP BLUE now finally, for the first time in this entire match, the first time, is lower on time than Kasparov is. Kasparov with 19 minutes, DEEP BLUE with 18 minutes and 39 seconds. This is the first time in this entire match. So he realizes it has a lot to think about. But I don't know what good it is going to do in this position as Kasparov begins to assert himself.
(Audience commentary.)
Mr. Ashley Since the success with brute force strategy is pretty limited so far, are there any plans for selective strategy which may prove better in the long run?
DR. BERLINER Let me say whatever the limits of brute force are they are pretty significant. I mean not the limits but the performance of brute force programs is very, very, very good. Back in 1900, Claude Shannon, who is a pioneer in electronics and things of that nature wrote a paper saying sort of tongue in cheek that we would never solve chess by enumerating all possibilities. And that has turned out not to be true at all.
DEEP BLUE is very, very close to outperforming Kasparov, and probably not very much additional work needs to be done, however, the problem of selective search is one that goes well beyond chess because people do it so successfully and in chess there is on the average of about 35, 36 moves in the average position.
In golf for instance there may be like 200 or even more moves possible at any one time, and whether the brute force approach would work there or in other endeavors that are not game-like but still happen on a human scene, it seems doubtful that brute force would win there.
So selective search is interesting, and I have another in the water in this context, but it is really still too early to speak that that, I think.
(Audience commentary.)
Mr. Ashley DEEP BLUE has recognized that the exchange of queens would be fatal, so instead it has dropped back to D7 with the queen, queen from A4 back to D7. But address the question, please.
DR. BERLINER Let me just explain, exponential explosion for those who may not be familiar with the term. If there are 35 legal moves in the chess position, if you look at every possibility there are 35 immediate possibilities and each of those has 35 answers and each of those answers has 35 answers, so the number of terminal positions that you are looking at at any one time.
When you go one deeper it increases by a factor of 35. So for instance you could be going from 35 to about a thousand and then to 35,000 and to a million or something like that, so it goes, very, very fast. That is just after a few moves. But it turns out there is something -- this is what all the early people who wrote about chess overlooked. There is something known as the alpha beta algorithm which has been discussed here on previous occasions. And what the alpha beta algorithm points out is that a move needs only be refuted one time. So once you found a refutation of a move, once have you shown that a given move is not as good as some other move that that side can play, you don't need to keep refuting it.
So that means if you really know what is happening more or less than hopefully the very first move you look at will refute the move that you are replying to.
So instead of looking at 35 possibilities you are only looking at one. So it turns out that every other level of the extremity you are only looking at one possibility and that means that in the limit you are only looking at the square root of the total effort.
So where you would normally be expected to look at a million positions, you need only look at a thousand, so that cuts it down tremendously and there are some other things that are called hash tables, which cut it down even further in the sense that if you recognize a position that has already been searched in some earlier effort to the sufficient depth then you don't need to search it again. And that happens again and again and again.
For instance, in a given position there could be a sequence A, B, C lead it is a certain position and somewhere else in the extremity there could be a sequence C, B, A leading to exactly the same position. You need to search it again. So these things are working to reduce the total effort.
Now, in chess that turns out to be significant. In golf I don't think it would be, but that is -- brute force is really sort of a tremendous scientific achievement because nobody thought it would work. And a bunch of programmers have sort of made it work and it is a fabulous tool. 37. Bd2 Re8.
38. Bg5 Rc8.
Mr. Ashley The only reason we think that maybe it is not working here is because the best player is defeating it, the best player in the world, but look who it takes to fight this thing. So certainly a wonderful accomplishment by the programmers at IBM.
Kasparov has played the move bishop to D2, and this again is typical chess strategy. After removing the active pieces, then you attack on the side where you are stronger. The pieces on the left side of the board, that bishop and that rook are simply not going anywhere. That bishop and that rook, they were out the gate.
So now Kasparov has turned his attention to the other side of the board, the right side where black's king is and he is prepared to bring his bishop on a wonderful post, the F6 square. That bishop on D2 I know Kasparov is just licking his chops thinking about how to bring this bishop in to the F square. And once that happens he is going to just have a significant force advantage on that side. I have use this example before, this analogy before about a bishop or any white piece being posted up down low and this bishop is on the go at G5, now looking to get into the F6 square.
And it is going to have a deep post in a position, sort of again the basketball analogy of being posted up into a position where nothing can be done about it. You throw the ball to Shaquille O'Neal. That bishop is about to become Shaquille O'Neal.
F6 is such a rosy square, but I'm enjoying it because Fritz four believes that black has an advantage. He has never seen check.
DR. BERLINER The thing I was just going to say is this position looks to me of being more for white than the position they were discussing yesterday as to whether DEEP BLUE should resign. There was more to be slated there. This one is so obviously won.
Mr. Ashley Fritz four has changed its evaluation to say only a slight advantage. Clearly this advantage is huge than is evident to us with no calculation.
DR. BERLINER Let me say a few other things about chess competition between humans and machines. It is frequently the case and this has happened to me certainly, that when a machine first comes out, people try to play it as if another human, and they find out it is a lot stronger than they originally thought and they get very much afraid of it and they say we don't want to play it or borrow it. And then somebody comes along and they try some things like Garry is doing and they find out some of the weaknesses, and the next thing you know your rating from last year is going down because everybody is learning how to play against it.
So that's what Garry is doing here, except I should say it takes a top notch player to control the elements like that. 39. Bf6 Kh7.
Mr. Ashley DEEP BLUE did play rook from A8 to C8 pressuring a pawn that it has defended twice. The thing is that DEEP BLUE has nothing to do. We talked about the dark square weaknesses a long time ago.
Well, that knight is gone and nothing defends the dark squares now and we talked about that bishop when Yas said the bishop could go to D8 and maybe protect that other side of the board, that bishop on B8 unable to help out in the critical sector. And now white's bishop has jumped in after the move rook to C8 to F6, and I don't know if you noticed but Garry did that flick of the wrist when he put that bishop on F6 as if to say, Take this, baby. That's a bishop.
(Laughter.)
Mr. Ashley Feeling very good about his position is Kasparov. He knows he is in complete control. He is leaning bakc in this chair and the watch. There goes the watch.
Garry knows this one is out. DEEP BLUE played the move king to H7. Garry's watch is back on. He knows he does not have to think about anything. He is about to make his 40th move. His head is laxly going back and forth. He is loose and he is relaxed.
I like the move G4 in this position, very attractive when one side is completely tied down. You just attack on the other side and G4 seems to open up the board in a wonderful way for white and I would not be surprised if this move came.
He has other options but he is thinking how to win on that king side now.
The king is by itself and he has a bunch of wolves waiting to come and eat up the king.
Kasparov has played another move, the move C6. This is very interesting. The point is not clear to me.
(Laughter.)
Mr. Ashley Interesting, it offers black the possibility of seemingly winning pawn with the move rook takes C6. I have a feeling though that, yes, the point is to dominate along an open line. Kasparov is not going to sacrifice on the right side. He wants to see a move like rook takes C6. If he can get his move he will play rook to C2.
Indeed DEEP BLUE has instantly captured with the pawn. Garry is trying to open up a line so he can get in the game but after pawn takes, this move seemed a lot worse because now Kasparov can play the very simple move rook to C5 and just look at black's pieces. They will never, ever dream of getting out of this position, ever.
(Laughter.)
Mr. Ashley I hope that is a concept that the computer understands.
(Laughter.)
Mr. Ashley I believe that Kasparov can simply attack on the other side by playing the move F3, G4, opening up the position. I think he has -- the fact is black can do nothing. Black has no moves. The bishop can't move. The rook can't move, nothing can move. This move by Kasparov --
DR. BERLINER C6 keeps the queen from getting out.
Mr. Ashley This of course is a fantastic move. The move I was talking about, G4, was an interesting one, and I think that Garry will get back to it. This is a typical way of playing this position,
Garry would have to worry with a move like queen to B5. This creates some kind of counterplay in that the queen is attacking the rook, attacking the pawn, putting some pressure on this pawn also. That is one thing that you don't want in a chess game, is to give somebody you are trying to come get any counterplay at all. You want them to have no moves.
The move C6 keeps the black queen out of the game, and after B takes C6 I think rook to C5 and all of black's pieces are in tune, all of them.
(Audience commentary.)
Mr. Ashley Rook to B1 is also a possibility. The only thing about that that I'm a little concerned about is after queen to B7 preventing B7 black now threatened the move C5 any way and you are going to have to stop this move. So why bring your rook over when you don't really need to.
In fact, black is tied down to the B7 square anyway. This queen is not going anywhere because of the threat of B7. Rook to B1 is certainly a possible move, but I think also rook to B2 is also suggested itself, but I like this kind of move. It's impossible to move the pieces, just impossible. That is just one move that brings attention to itself. Certainly though Garry is enjoying himself.
I would like to thank Dr. Hans Berliner for coming.
(Applause.) 40. c6 bxc6. 29 Mr. Ashley Welcome back, Yasser. 40. Qc5 Kh6.
Kasparov has chosen to put a piece on C-5, but he has put his queen to C-5, blocking again the possibility of this pawn's advance and DEEP BLUE responded instantly with the move king to A6. And what is the Seattle term for this position?
Mr. Seirawan Horrible or lovely depending on if you are Garry.
I was going to say go back one move, if you will. I thought that, first of all, it was very interesting that DEEP BLUE did go for this tactical sequence knight E five and we discovered all the moves. Going back to this position, I thought the simplest position for Garry to win the position was to play the move rook C2 with the intent of just playing the move queen A4 and recapturing the pawn on C6, and that was deadly. Instead Garry has done something that also could be expected. Woops. He has played queen C-5 with the idea of just keeping the position, keeping all the pluses of the position.
DEEP BLUE having no moves, if you will, has played the move queen H6, and now this bishop on F6 is just a monster piece. It sets up a beautiful mating net. 42. Rb2 Qb7.
Mr. Ashley Kasparov on every move now making that flick of the wrist, enjoying his moves and now he has played the move rook to B2. It seems as if the move queen B --
Mr. Seirawan There is a beautiful bishop on F6, sets up this wonderful mating next over here against Black's king. So I expect that Garry may go for the direct win with a move like rook E4 intending to play G four, opening up the kingside, and once that Kingside becomes open, all of Black's pieces huddling over here in the cold on the queen side won't be able to get over to save their beleaguered monarch, so I think that Garry has the game well in hand, and rook B2 is very nice.
Mr. Ashley Tying everything down. I'm surprised that Garry is holding back, but he knows what he is doing, getting everything he can on one side before he busts open the position on the other side. He may not need to because he can also win the C6 pawn back at will, practically by force, so DEEP BLUE would anticipate that he would play a move like queen B7.
What about rook to E8 that drops the C6 pawn immediately, so that is going to be bad. DEEP BLUE needing a move and queen B7 would put the queen on such a horrible square. Well, well, DEEP BLUE is in deep trouble.
(Laughter.)
Mr. Seirawan We have the same position. Garry put his watch back on. I was out doing some interviews when I saw that and it is just so beautiful. That is so typical of Garry, and as an opponent who I played Garry five times, having lost twice, won once and two draws, when that watch goes on like that, it can be so intimidating. It is like oh, no is it really that time already.
(Laughter.) 42. Rb4 resigns.
Mr. Seirawan So lots of ways for Garry to win the position. He knows it. And he is just indicating to the operator, you know, hey, good movie on tonight.
Let me just say what we have in store for tonight. Following the game, the IBM team -- and we have seen, by the way, queen B7 has been played.
IBM team -- and we have a rook B4, as predicted, going with this idea of busting open -- I like that, busting open the kingside with the move G4. Following the game, there will be a press conference that we won't be invited to attend, but we will certainly watch. Our mikes will be off so we won't be able to ask questions, but we will be able to follow the press conference.
Both Garry and the IBM team, following the press conference, which will last about 20 minutes, they will then join us in this room and give us further insights to the match, the strategy. Again, this is the final game of the match, so it is over after this evening and also invite just a few questions.
Obviously we won't be able to get all the questions in, but that is the festivities for tonight.
Mr. Ashley Yas, I have to say that I'm feeling pretty good about this win for Garry because when asked by our friend here, Dr. Frederick Freidel of Fritz 4 fame, what I thought the final score was going to be, I say 4 to 2. He has the paper, but I said 4/2 Garry. So you have the written word on that. Is that what it says there, Fred? That is that guy Maurice right there. He is right there.
Mr. Seirawan Don't I have a piece of that pool? Some early match predictions.
C.J. Tan, the developer, the lead programmer of the DEEP BLUE team, said three to three. That was what the lead programmer thought.
David Levy of the ICCA, the International Computer Chess Association, predicted 5/1, Garry.
Joe DiBlasi, the president of the ACM and our sponsor of the match, he thought 4/2 Garry.
Ken Thompson, Bell Labs, gave Garry a slight edge with 3-1/2/2-1/2.
Garry's mother predicted four to 2. Hey, mom.
Mr. Ashley Garry is drumming his fingers right now, a bit impatient with this position. He knows how easy it is going to be. He is saying hurry up and move, man. I want to beat you up now.
Mr. Seirawan Terry Phoenix, ACM, the head of the media, predicted 5/1, Garry.
Monty Newborn, more than anybody singularly responible for the match predicted the computer was going to win two and a half, three and a half.
Bob Rice, the PCA commissioner predicted Garry.
Mike Valvo, the neutral arbiter, predicted four to 2, predicted 4 to 2 for Garry.
Hans Berliner, who has joined us, predicted four and a half, 1-1/2 for Garry.
And finally Maurice, four 2.
Mr. Ashley Thank you very much.
Look at Garry. He is looking at his knuckles and fingernails as if to say move, move, move. I have this game under control, hurry up. And DEEP BLUE? I wouldn't want to move either.
(Laughter.)
(Applause.)
Mr. Ashley Garry Kasparov has won the match by a four to 2 margin, the world champion.
Game ends 646 p.m. US EST
Deep Blue Resigns!!
THE WORLD CHAMPION GARRY KASPAROV HAS DEFEATED DEEP BLUE 4-2 IN THE ACM CHESS CHALLENGE.
[ The Game | The Program | Commentary | Grandstand | Deep Blue ]
[ IBM home page | (C) | (TM) | IBM Small Planet Pavilion | Internet 1996 World Expo ]