Commentary for black move 15:

MAURICE ASHLEY: What intrigues me, before we take another question, and I think some of our audience, even after you explain we have the human powers of understanding, intuition, chess knowledge, chess history on our side, you can still say --

DB MOVE: 15...e5

MAURICE ASHLEY: And a move has actually been paid. Deep Blue has played e6-e5, staking its claim in the center of the board and looking to open up the position with both sides' kings still in the center of the board. Talk about unconventional chess strategy, Yaz, that's a very interesting idea by Deep Blue, opening up the board. Once you clear the center pawns, the action gets very sharp, and pieces start to fly around in the center, and then the computer's powers of calculation start to become almost more important than strategical factors in a position. This move, it's just taken it to Garry with the black pieces.

YASSER SEIRAWAN: Exactly. And the move e6-e5 works for both sides in this case. In the first situation Garry is a happy camper because the position is going to open up a little bit more so that this bishop on e2 and the bishop on f1 are going to -- c1 are going to have a big role in the future outcome of the game. From the computer's perspective, it's gaining space. The pawn on e6 is not as good as the pawn on d4, keeping in mind that the pawn on d4 controls two of its spaces, c5 square and the e5 square. So by playing e6-e5, the computer says, "I get to trade an inferior pawn for a better pawn. I open up the position." And this is by the way a scary part of the computer. I've played by the way a lot of, lot of computers and one of the things I like to avoid is an open position. And if I'm forced to play an open position, which is what this position is going to become, I want to go in with a big advantage, because I'm going to get out-calculated.

MAURICE ASHLEY: That's a scary move. It actually had the awed dassty to play it. The computer is not audacious or anything else, it just makes moves. And it's not afraid of this position, it knows it can calculate -- well, doesn't know it can /KPHRAEUT, but it can /KPHRAEUT 200 million positions a second. (Audience laughter.) And societies just analyzing, number crunching, number crunching, and Garry's got to feel the heat. You look at him now, head in hand, he's intense about this last move because it's not a passive one, it's not a casual one, the computer didn't just castle, Deep Blue didn't just say let's be safe, let's develop our game a little bit, we are playing Garry Kasparov." (Audience laughter.) There's none of that, just made this move. And every move, there's so many moves this game, it seems almost as if the programmers have tweaked with Deep Blue a little bit and put it oi ahyperaggressive drive because every move is centralizeing, we've seen the e5 square used by a knight. Now we see h5, now another aggressive move, e5.

YASSER SEIRAWAN: Absolutely.

MAURICE ASHLEY: It's a different -- is seems almost like a slightly different Deep Blue, a bit more aggressive, more like the Deep Blue of game one, Yaz, when it was very aggressive in its choice of moves.

YASSER SEIRAWAN: Well, let's give kudos to IBM and the team led by C. J. Tan. There is no question that between Philadelphia and today we're seeing a much, much better program. It's calculatetive abilities have doubled like we've already spoken about. More than that, they've polished the software program, they've played positionally better, the openings are better, even its endgames are better. And that game four, by the way, I've been told that as soon as the rooks were exchanged, the first pair of rooks were exchanged that Deep Blue had that position in its Deep Blue and knew the game was going to be a draw, if Garry Kasparov played perfectly.

MAURICE ASHLEY: Right, if.

YASSER SEIRAWAN: If Garry Kasparov played perfectly. We're seeing a much, much stronger computer. And just on that note, by the way, this match is being sponsored by IBM, they're putting up a 1.1-million-dollar prize fund, they're making this whole event possible, and please join with me in giving our thanks to IBM for organizing this great match.

MAURICE ASHLEY: Let's go to another thing. The fact that Kasparov cannot see all these millions of positions per second, and yet he still is able to beat Deep Blue?

YASSER SEIRAWAN: Yes.

MAURICE ASHLEY: He's still able to get an advantage in many games even though, admittedly, he wasn't able to press those home? What is it? What is it that allows that?

YASSER SEIRAWAN: It's very, very simple. The computer is seeing billions and billions and even trillions of possibilities a game. Garry Kasparov is not. What Garry Kasparov does is what we call a selective search. In a position where there are 48 possibilities, he doesn't look at every possibility and says, "No, there is one, conceivably two good moves, and I like this one better." And instead of searching five moves down the road, or ten moves down the road or 20 moves down the road, he searches selectively a pattern that he knows is going to lead to a won position for him. He has an extraordinary experience of endgame positions that he uses as a reference. He says, "Uh-huh, I know I can win that ending. I'm going to gear myself towards it. He sees an incredible number of tactical positions almost like at his fingertips, says no, no, no, that's an attacking win for me. I know that that's the kind of position I love and I can win." So he aims for a particular type of position using a very, very Nair, very, very narrow selective search. So Garry, in his allotted three minutes, he may only look at 15 or a hundred moves. But guess what? Those are the primary 50 to 100 moves. And when it takes Deep Blue three minutes and billions of calculations to come up with the same move takes Garry a few seconds.

MAURICE ASHLEY: I guess another analogy to use is driving a car. If we were to program Deep Blue to drive a car, it would analyze every possibility, the lady who might run into the street all of the sudden, the driver behind us who might press down on the gas and instead of going 30, go 40, instead of going 40, go 60. And all these calculations would come in. But when humans drive a car we sort of just know what to do. We know that we expect certain things to happen and we follow -- within just that limited perspective, and sometimes we do have crashes, as happens, because we are not expecting the unexpected, but most of the time we're able to engineer ourselves in that kind of simple situation where the rules are known to us, the strategies, if you will, are already known to us and we sort of employ that all along the way. And that basically gets us through. It those unusual situations, though, that we would wish we could have seen everything in advance and calculated it all out, and Deep Blue is /TAOEUBL say, "Uh-huh, this is an exception to your rule, and I'm going to beat you because you don't know it."

Real-time text commentary is made possible by LiveNote, Inc. and Vincent Varallo Associates