Stork: What was it like when you first saw 2001?
Campbell: I remember very vividly, I was a kid, I went to a movie theater and came out saying "wow, what was that?" Especially the closing sequence. I didn't know what it was, but I knew it was cool, and the technology at the time didn't impress me I didn't know enough to know how difficult some of that stuff was. I knew that space travel was difficult, but I didn't know that the computer technology was so difficult.

Stork: Can you tell us about some of the history of computer chess?
Campbell: The very first chess programs were laughably weak, they played far worse than even a beginning human player, but sometime in the 60's they started to improve as machines sped up, and sometime in the early 70's it was discovered that if instead of trying to emulate the way that people play chess people program computers to play chess the way computer play best, and that is brute force. Looking for all the possibilities, being very systematic. Suddenly there was a big leap in the level of chess play and the strength continued to increase as computers got faster. And so, throughout the 70's and 80's, computers became stronger and stronger, eventually reaching the Grand Master level, in the late 80's, and then as we worked on improving the speed even further the technology, we were able to reach the level of the world champion in the Deep Blue matches.

Stork: You mentioned that it is very different from the way human play, how do humans play?
Campbell: Humans and computers have a drastically different style of play. Human beings have, especially the best human beings, have this ability to look at a position and intuit what are the best moves, the two or three best moves, the moves that are most worth looking at and they'll examine them, and often based on gut feel more than anything play the best move. Computers, on the other hand, look at all of the possibilities, as many as they can, as deeply as they can, and apply a fairly rudimentary evaluation to the positions , no where near like the sophisticated evaluations that people have. And make their judgement that way, so it's quite a contrast.

Stork: How does Deep Blue play?
Campbell: Deep Blue begins by looking at all possible sequences of moves down to some depth, say, four moves ahead, and each of the positions that could possibly be reached, it evaluates them in a fairly complex way for computers, it comes up with a number, this is how good the position is, and then it chooses a move based on a four move search. Then, if there's time, it'll look 5 moves. 6, 7, and so on, and it'll maybe end up looking 11 or 12 moves ahead, all the possibilities again, each time choosing the best possible sequence for it, assuming the opponent plays a reasonable sequence of moves, and then in the end, there is one move that will pop out as the best.

Stork: But how does it tell the quality of one of the board positions that it might ultimately obtain?
Campbell: One of the most complex parts of Deep Blue is what we call the evaluation function. That's the piece that, taken the whole chess board, will come up with a single number that represents how good it is. Positive numbers are good for white, negative numbers are good for black, and it takes into account a whole range of factors. First of all, the number of pieces on the board for each side, but more subtle things like the placement of the pieces, the arrangement, the weaknesses of the king position, there's in all, 8000 different features that comprise the Deep Blue evaluation function.

Stork: What's the role of humans in helping Deep Blue tell what's a good position?
Campbell: Deep Blue by itself can't have the raw features. It knows what the features are and the chess positions that are interesting, but it doesn't know how to weight them, what's most important, what's unimportant in this situation. So, the Deep Blue evaluation function was completely tuned by hand, by human beings. We had to look at a chess position, see if the Deep Blue was making a reasonable move, and then, if not, find out why not. And tweak the evaluation function to correct that. Now, the problem with that is you can fix one problem and create 1000 new problems, so that's what makes it difficult, and that is our 1 regret. That Deep Blue wasn't a learning system, that it couldn't learn by itself how to play chess. It had to be taught every step of the way.

Stork: Tell us about the first tournament between Gary Kasparov and Deep Blue.
Campbell: In 1996, DEEP BLUE played the first serious match between computer and the world chess champion, in this case Gary Kasparov. It was a 6-game match under tournament conditions which was again unique. The world champion had never played a computer under tournament conditions before. And, it was exciting from our point. Of view because the very first game, DEEP BLUE was able to win which again was unique under tournament conditions. Unfortunately, DEEP BLUE didn't go as well after that. Kasparov showed that he was able to adjust his style of play. He perceived some of the weaknesses of the system, and was able to, in the end, win the match by a score of 4 to 2.

Stork: So he was psyching out, or figuring out and learning about Deep Blue, but he was also famous for psyching out his opponents. What do you think that had to do, or didn't have to do in this kind of a tournament?
Campbell: Kasparov is well known for his ability to psych out his opponents. To make them play less than their best, and he's a very intimidating presence. Obviously this is completely lost on Deep Blue. It had an affect on me sitting across the chessboard from him, but I was still aware enough to be able to move the pieces on the board. That's ok. However, his ability to abstract out some weaknesses, detect weaknesses in the play of Deep Blue, that's something that only the top players could have done, and Kasparov was able to do that very well in the first match.

Stork: So, now tell us about the second match between Kasparov and Deep Blue.
Campbell: In the second match, we had about a year to prepare for that match, against Kasparov, and we had seen a lot of weaknesses in the first play of Deep Blue, so we had decided that Deep Blue was not searching fast enough. It was looking at about 100 million positions per second. In the first match, we were able to increase that to about 200 million without any significant effort. But the key was increasing that amount of chess knowledge which we were able to do by working very hard with a grand master for a year, and in the first game of the match, Kasparov was able to win by playing this very anti-computer style of play. But we weren't worried because he had to work very hard to win that game. He had to play extremely accurately and it showed that in the rest of the match Deep Blue won the second game, and then a few more draws. He was working extremely hard to play each game and the 6th game came along he wasn't thinking as clearly as he might have and he made a poor decision and he lost the game extremely quickly probably the fastest game he's ever lost in his life.

Stork: What particularly unusual or dramatic happened in game 2?
Campbell: Game 2 of the match with Kasparov was an amazing game. It was probably the best game ever played by a computer. It played in a very human style, and that perplexed Kasparov, even to this day. He didn't understand how a computer could play in such a human style. But the interesting point is that right at the end of the game, when it looked like Deep Blue had completely sewed up the game, it was over, Kasparov just gave up and walked away. But it was discovered later that night, by numerous people analyzing together on the internet, that Kasparov in the final position could have had a draw had he played a very accurate sequence of moves, and I think that also had an impact on Kasparov's psychological well-being.

Stork: So, it seems like Deep Blue psyched him out.
Campbell- Unintentionally, Deep Blue was able to psyche out Kasparov.

Stork: What was unusual or dramatic about game 6?
Brooks: Game 6 was the critical game of the match. The winner of the game was the winner of the match, and going into the match, that game, it had been a tough match, he, you know, it had been five games that had been played so far. It was very tough on Kasparov, and very early in the game he made an extremely risky decision, a decision that many people have since questioned, and that is he allowed Deep Blue to have a very strong attack. Deep Blue had to give up a knight to do it. Most of computers would not have understood that position, would not have realized that the white pieces (Deep Blue was playing white). The white pieces had the advantage. Deep Blue knew that, and it took advantage, full advantage, of that, and was able to win the game in only 19 moves.