Bronstein on computers and humans

by admin on February 16, 2008

The Russian-language chess website www.chesspro.ru has a really cool interview with David Bronstein, the former world championship candidate who died in December 2006. Mostly Bronstein talks about the differences between human and computer chessplayers. The interview was conducted in 2002 and 2003 by Yuri Vasiliev. In his preface, Vasiliev says that Bronstein read over and approved every word, so “This is his voice.”

So here’s what Bronstein has to say (as translated by me):

“A human sees a position as he sees a dark road by lantern light… At first you see the position clearly. After three moves it becomes for you somehow a little bit foggy. After five moves you see only the contours of the position. You sense that something is there, over the horizon.  You don’t know exactly what it is, but you’re very curious. For me, that is really the charm of chess. But the computer sees any position as clearly as you see the first move. This is why I protest against computers. A human thinks between moves, but a computer places the position on the screen — and sees.” [Emphasis added.]

A great example of the difference between thinking and seeing was my recent game in the Santa Cruz Cup. I thought very deeply in the position, seven moves deep … but by that point I could no longer see the position. So one moment I thought my opponent’s rook was on c6, and the next moment I thought it was on c7. A computer would never become confused in the same way; it would know the rook was on c6 with the same certainty I would if I saw it there.

It’s remarkable to me that chess players, no matter how accomplished they are, always see better than they think. That’s why blindfold chess is hard. That’s why we have all had the experience of thinking a long time over a move, touching a piece to move it, and in that very instant seeing why the move is a mistake. This phenomenon has always mystified me, but Bronstein’s comment explains it. Seeing uses a different part of our brain than thinking. Humans are actually very bad at thinking, at putting one logical step after another one. But we are phenomenally good at seeing — better than any computer, in fact.

So here’s a thought: if we could only tap into the immense power of our visual cortex, and use it to process chess moves instead of images, we could probably outplay computers even today. Unfortunately, we can’t do that, because we can’t override millions of years of evolution and reprogram our brains.

Later, Vasiliev asks Bronstein if humans are doomed when they play against computers. Bronstein replies,

“Of course. To play with a computer, hoping to beat it, is the same thing as beating your head against the wall. The harder you run up to it and hit it, the more obvious the result will be. More blood. Yours.”

So you might expect that Bronstein would think that we shouldn’t play against computers at all. But no!

“We must. Because you’ll never find a better teacher. In the first place, it brings humans down to earth: don’t stick your nose in the air, don’t think that you can see something, you’re not a magician! Secondly, by its actions, which are absolutely free of prejudice, it can teach us not to be afraid of difficult positions… The computer suggests to humans new technical methods and helps teach us not to think like a stencil. For example, ever since the time of Capablanca it has been accepted that it is not good to move the same piece twice in the opening. Or that you shouldn’t bring your queen out early in the game. The computer is phenomenal at refuting these human misconceptions, which have become firmly embedded in our brains. It takes ideal advantage of the strength of the queen, making it do unbelievable pirouettes on the board. Humans cannot play that way. That means we have to learn. The computer teaches humans to sense the geometry of the chessboard better. It reminds us that it isn’t necessary to play with plans and ideas, but it’s possible to play one move at a time… Can humans learn? They can. There are many capable students. Yes, man has something to learn from the computer. He must always keep it available as a partner. Not making it into an enemy, not trying to slay the Dragon. I could end on this optimistic note if I didn’t know for a certainty that they won’t be satisfied with this. They are the people who have an insatiable desire to slay the Dragon, and the programmers who want to prove that their Dragon can subdue the imagination of Man.”

Lots of interesting things here, some I agree with and some I don’t. The last sentence especially: I really think that programmers have never been interested in “subduing” man, as Bronstein puts it. That’s sci-fi stuff, out of the Terminator movies. His comment about slaying the dragon has more truth to it; I confess there were times, when I played the computer, that some kind of primitive endorphin-releasing system has kicked in and I wanted nothing in the world more than to beat the machine. (I think that’s also what lies behind video-game addiction; you always think that if you played just a little bit better, you’d win.) Bronstein is telling us to stay away from this primal urge to beat the machine, and I think he’s right.

Also, I don’t think that the computer has proved that Capablanca was wrong about moving a piece twice, bringing the queen out too early, etc. But humans are too “stencil-like” in their thinking, as Bronstein said. We read Capablanca’s sensible advice, and then we think that in every position it’s wrong to bring the queen out early or to move a piece twice. The advice was sensible, but the dogmatic application of it is not.

Finally, although most of the interview was about computers, Bronstein said a few things about the King’s Gambit that were music to my ears. (As you know, I’m a King’s Gambiteer.)

“I think that after 1. e2-e4 e7-e5 the most noble move is 2. f2-f4. The King’s Gambit, in my opinion, is the only opening that makes it worthwhile to play chess. If both players set a good tempo and aren’t afraid of putting their kings in danger, it is the most enjoyable. No one has proved that 2. Ng1-f3 is the best move. Philidor said that the best move in this position is 2. Bf1-c4. And it’s even more unproven that White’s third move 3. Bb5 [after 2. Nf3] is the best. Kasparov won many games as White with the Scotch Game. It’s more aggressive, although it does have some problems.”

The only opening that makes it worthwhile to play chess! How about that!

Print Friendly, PDF & Email

{ 1 comment… read it below or add one }

dribbling February 24, 2008 at 12:31 am

I was once introduced to Bronstein and spoke with him very briefly. He came across as completely unpretentious.

Speaking of computers and humans, I saw Smallville (Nakamura) playing bullet against a computer on the ICC. Smallville was rated something like 3300 and the computer 3600 plus! I think, I’m not too sure. If you have already seen this, please stop me. 🙂

What Smallville did was he played to block the position completely and was willing to sacrifice a great deal of material to do so. It was funny to watch the computer a piece or two up moving around aimlessly behind its own pawns while Smallville did the same in order to draw – thus improving his rating – by the 50 move rule.

I watched several games, but it all happened so fast I have no idea of how he did it. It seems to me that Smallville has found a flaw in the computer’s evaluation algorithm.

It also appears to me that Nakamura is a genius.

Reply

Leave a Comment

{ 1 trackback }

Previous post:

Next post: