Commentary for white move 15:

MURRAY CAMPBELL: It's sort of an evolutionary process over the last year orse. Doesn't it keep adding knowledge, teaching it about different kinds of pogs -- teaching a computer to play chess is drastically different from teaching say Tal to play chess. Teaching it -- teaching a child you show them a few examples of a good idea, and they will after a while be able to generalize and say oh, the principle behind that is this, and in similar positions I can do the same thing. In building a system like Deep Blue, it's much more difficult because if you just give it a rule, it will follow that rule, and if there are exceptions, and you haven't told it about the exceptions, -- exceptions, it will still follow the rule and potentially make a mistake. So it's a very difficult process to put chess knowledge in a chess program and it takes a great deal of time. And over the past year we've seen the level of play of Deep Blue slowly go up and more and more knowledge has been incorporated into the program. It's impossible to get all the relevant chess knowledge into a system, a chess program. There's just way too much, too many exceptions, too many rules. GK MOVE: 15 b4

MAURICE ASHLEY: Kasparov has moved b4, avoiding the pawn storm. You see that move a3 he played so many moves ago has now shown its usefulness. I did see a question from the audience? AUDIENCE MEMBER: I agree with your analysis for Tuesday's game. The green bar, according to Fritz, is slowly rising, but it only got to .94, which is not quite a pawn but very close to it. What was Deep Blue's analysis of its number for the closing position, when Garry resigned?

YASSER SEIRAWAN: Just to /RAOEPLT the -- repeat the question, what the gentleman had said is that he had watched the Fritz toolbar, and it had only risen to the point that it considered white to be just a point ahead, which isn't what we might describe as resignation territory. What was the valuation by Deep Blue?

MURRAY CAMPBELL: I believe that Deep Blue was a little more than a pawn ahead, maybe a pawn and a half. We wouldn't have resigned if we were playing. It might not have seen Qe3 -- perhaps it would have -- but we wouldn't have resigned right away.

MAURICE ASHLEY: When do you decide to resign for Deep Blue? Does Deep Blue have a function that says "I resign"?

MURRAY CAMPBELL: No. We will not let it do that. I can show you positions where one side is a queen ahead and it would still be a draw. So -- and Deep Blue would be able to recognize it. So we make the decision for it, but it rests very strongly on what its evaluation of the position is. For example in game one it thought it was down 500 points. A pawn is a hundred points so it thought it was down a rook. So at that point there's just no point in playing on.

MAURICE ASHLEY: Yeah, as you were talking about the pawn storm on the king-side and now this is switched. Why don't you give a reason for Kasparov doing this.

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