Man vs. Machine
[Sorry, the video for this story has expired, but you can still read the transcript below. ]
ELIZABETH FARNSWORTH: Next tonight, man versus machine, Round 2. In one corner, weighing 176 pounds, chess grandmaster Gary Kasparov, age 34, born in Baku, Azerbaijan, world champion since 1985 and considered by many the greatest player in the history of the game. And in the other corner, weighing 1.4 tons, “Deep Blue,” or rather “Deeper Blue,” the new and improve RS-6000 SP supercomputer, designed and operated by five IBM scientists and able to examine about 200 million chess positions a second.
Because it needs to be in a chilled room, the computer is linked by phone to a terminal at the site in New York City, where the contenders are in the middle of a re-match. Big Blue instructs the IBM scientist sitting at the terminal which moves to make. In last year’s six-game match Kasparov won three games, the computer one, and two were a draw. The score so far this year: even–one to one. Kasparov triumphed in a close match Saturday. But Blue came back to win convincingly yesterday. The contenders play the third of their six-game re-match Wednesday. The winner will take home or back to the computer lab $700,000.
ELIZABETH FARNSWORTH: For analysis of the match to this point we’re joined by C. J. Tan, who heads the team of IBM computer scientists that built Deep Blue, and Frederic Friedel, a computer chess expert who serves as Gary Kasparov’s technical adviser. Thanks for being with us Mr. Friedel, what happened yesterday? How did Deep Blue win so decisively?
FREDERIC FRIEDEL, Kasparov Adviser: We don’t know. We are still studying the game. We spent all of yesterday evening and this morning studying it, and we haven’t come up with the definite answer. It played very strategically.
ELIZABETH FARNSWORTH: And you were surprised by that, right? You knew it had great tactics, since it can make so many–can analyze so much every second, but strategically?
FREDERIC FRIEDEL: Yes. Computers are not supposed to be able to do what Deep Blue did yesterday. It played some moves which were very, very human. And if this is the case, if it can play in this way, then we have a very, very difficult task.
ELIZABETH FARNSWORTH: Mr. Tan, what did you do to it? You can’t feed it spinach or Wheaties or anything. What did you do to it?
C. J. TAN, IBM: Well, we didn’t feed anything except electricity. What we have is a deeper machine. Deep Blue can search deeper and also sharper because it now has much more chess knowledge. So you combine the speed with the experts’ knowledge we imbedded to the system, and you have a very formidable chess player.
ELIZABETH FARNSWORTH: But Kasparov won Saturday. Did you do something to it between Saturday and Sunday?
C. J. TAN: Not really. Saturday we were playing black, and Sunday we were playing white. So basically–
ELIZABETH FARNSWORTH: And depending which one, you either go first or not, and there’s a certain advantage, right?
C. J. TAN: Right. So the strategy for each of those games was a little bit different.
ELIZABETH FARNSWORTH: Mr. Tan, explain what Deep Blue is doing when it plays. It’s not really thinking, right?
C. J. TAN: No. It’s really doing computations. It starts with a specific chess board position. Then you search through all possible alternatives, and what, based on that search, he looks to the future very far and then based on its knowledge of chess, then you take the best move.
ELIZABETH FARNSWORTH: And its knowledge of chess depends on what you all–you and your five colleagues–
C. J. TAN: That’s right.
ELIZABETH FARNSWORTH: –have put into it.
C. J. TAN: Exactly. So Deep Blue, really as a machine, does calculations, and the preparation of the game begins on the scientist behind the machine.
ELIZABETH FARNSWORTH: And you’re not a chess player, right, but one of you–one of your group is?
C. J. TAN: Well, our chess member is a relatively good chess player, and others are just computer scientists, and basically this is a scientific research project. However, we have employed as a consultant a former U.S. chess champion grandmaster Joel Benjamin. What we have done is try to capture experts’ knowledge in specific field, and in this case chess, and that–program. We hope to use the same kind of model that we can apply that to other areas to apply this technology.
ELIZABETH FARNSWORTH: Mr. Friedel, what’s it like in the room? You’ve been in the room. How does Gary Kasparov play against this? He can’t read his opponent. Even though there’s somebody sitting there, he’s just–he’s just doing what the computer tells him.
FREDERIC FRIEDEL: Exactly. This is one of the problems. Actually he must ignore this person because if this person looks worried or nervous or anything, it’s for a different reason. And normally chess players look at their opponents before they make a move and after they make a move, they glance up to read body language. Here he shouldn’t do it because he’ll get false clues.
ELIZABETH FARNSWORTH: How did you train him for this match? How did he prepare?
FREDERIC FRIEDEL: Well, I’m his computer consultant, and I help him on the computer side. It’s interesting that Kasparov uses lots of computers to train, whereas, C.J. Tan just told us they use a lot of human beings. We have three computers in the room, and we are analyzing positions. We’re trying to simulate what Deep Blue is going to play, and we have a program which–you know, which is exactly the same as Deep Blue, except it runs a thousand times slower, unfortunately.
ELIZABETH FARNSWORTH: What do you think, Mr. Friedel, are Gary Kasparov’s relative advantages and disadvantages against this computer?
FREDERIC FRIEDEL: I think they’re perfectly clear. He can play strategic chess. He can plan very, very deeply. He can use his experience to decide which moves to play. Remember, he’s looking at two moves per second. The opponent is looking at 200 million moves per second. And he has to use these faculties. That’s why we are very upset or very disturbed by the fact that Deep Blue was playing extremely strategic chess yesterday.
ELIZABETH FARNSWORTH: Is Gary Kasparov doing something? Is he changing something for tomorrow?
FREDERIC FRIEDEL: I wouldn’t be able to tell you, not with C. J. Tan sitting next to me.
ELIZABETH FARNSWORTH: Mr. Tan, how do you see the computer’s advantages and disadvantages vis-a-vis Mr. Kasparov?
C. J. TAN: Well, the computer’s advantage is speed and the precision of calculation, however, whatever knowledge it doesn’t have, it will not be able to make a precise accusation, so the advantage also is it’s being a disadvantage. Namely, the computer cannot do anything that you have not told it to do.
ELIZABETH FARNSWORTH: And Mr. Tan, you don’t like the man versus machine metaphor, do you?
C. J. TAN: No, I don’t.
ELIZABETH FARNSWORTH: Why? Tell us why.
C. J. TAN: Well, this is really not man versus machine, as Mr. Friedel has just told everybody. Both sides uses computers. Both sides use technology. So it’s really man and computers together to solve a specific problem. Now, here what we are seeing is two different modes of using technology, and that tells us in the future what different ways of using technology to compliment our strengths and weaknesses. Their strength is Kasparov’s knowledge and expertise. Our strength is our tool is better, so we compliment each other, and that’s how we can use technology to solve our problems, help humankind to have a better life, a better future.
ELIZABETH FARNSWORTH: And, Mr. Tan, you mentioned earlier the application–the potential applications of Deep Blue’s learning through this process of dealing with chess.
C. J. TAN: Right.
ELIZABETH FARNSWORTH: What are they? Specifically, what might be applicable?
C. J. TAN: Well, we, the machine, the R-6000 SP we use for Deep Blue is already being used in many areas. The job researching, weather forecasting, oil exploration, what we have done with Deep Blue is to further those capabilities by specific application of specific knowledge in certain areas. So a lot of the things we’ve done already is to develop a system where we can discover drugs much faster based on similar type of technology we have developed.
ELIZABETH FARNSWORTH: You can discover how a drug works?
C. J. TAN: Discover new drugs.
ELIZABETH FARNSWORTH: New Drugs.
C. J. TAN: So we can shorten the time of a drug and go on to the market, and that literally translates into lives being saved. So there are other areas we’ll be looking after this event’s over, is to use this technology to have a better handle or to economic modeling out of the financial model, so help us to have a better handle on the financial situation.
ELIZABETH FARNSWORTH: Interesting. And, Mr. Friedel, in your mind, what are the implications of a Kasparov loss in this match, or perhaps he wins this time and the next time Blue is even better, what are implications of that?
FREDERIC FRIEDEL: Well, I think it will happen. The question is only whether it will happen this year in May 1997, or I believe in maybe five or ten years. And the implications would be that for the first time in history something which requires intelligence in an area which requires intelligence in human beings machines become better than human beings. I think people will look back in 50 years at the time when Kasparov or some other world champion is beaten, has a very auspicious moment.
ELIZABETH FARNSWORTH: Well, thank you both very much for being with us. We’ll be eager to see what happens tomorrow.