刚看到一篇对卡斯帕罗夫的采访。里面有一部分是关于国际象棋和AI。我读了以后觉得很有意思，摘抄一下。 COWEN : Now, given this, does this mean even very, very good chess engines will never play games as beautiful as human beings? Because you were so excited by this combination, in some ways, it may have led you down inexact paths, but it nonetheless led you down those paths, and we got the beautiful game that we did. KASPAROV : Now you move back to these things, chess computers, and there’s certain things that people should realize. I hate talking about these things. We say in Russia it’s using a “bird language,” because you’re asking me questions and I’m not sure that — 99 percent of our listeners — they understand exactly what we are talking about. The one thing for people to understand is that chess is, you may call, mathematical infinite game. The number of legal moves is more than number of atoms in the solar system. So machines cannot solve the game. You cannot expect machine to play e2-e4 at move one and announcing mate at 16,455 moves. But machines could work the game of chess from the end. Now we know that machines mathematically solved all positions with four pieces, like king and queen, versus king and rook. All positions with five pieces, all positions with six pieces, and now seven pieces. Seven pieces, it’s on the way. I’m not sure it’s all solved. We’re talking about 100 terabytes. Obviously, eight pieces will be already just insane number, and the game of chess’s ultimate endgame with 32 pieces. That’s why, maybe, machines will get to eight or nine moves, but that will probably be the end, even for the immense computing power that you can expect in next five, ten, twenty years. COWEN : Will AI write beautiful music, or is there something about...? KASPAROV : But I want to finish this because what we discovered in this process... I wouldn’t overweight our listeners with all these details. I don’t want just to throw on them the mass information. COWEN : It’s amazing what people will enjoy, though. You’d be surprised. KASPAROV: In some of the positions, like there are certain seven-pieces positions, when the win — and we’re talking about a forced win — can be reached within 500 moves. Now, 500 moves, I remember, I looked at some of the positions. Even at six-pieces positions... COWEN : It’s not intelligible, what’s happening, right? KASPAROV : It’s no intelligence at all. It’s just pieces moving around. There’s a certain position with king, two rooks, a knight on one side, and king, two rooks on other side. It said mate in 490 moves, first mate. Now, I can tell you that — even being a very decent player — for the first 400 moves, I could hardly understand why these pieces moved around like a dance. It’s endless dance around the board. You don’t see any pattern, trust me. No pattern, because they move from one side to another. At certain points I saw, “Oh, but white position has deteriorated. It was better 50 moves before.” The question is — and this is a big question — if there are certain positions in these endgames, like seven-piece endgames, that take, by the best play of both sides, 500 moves to win the game, what does it tell us about the quality of the game that we play, which is an average 50 moves? COWEN : It means we’re clueless in the entire universe. KASPAROV : Exactly. It’s an interesting philosophical question, and I have to confess, I don’t know the answer. COWEN : Imagine how bad our politics would look. KASPAROV : Listen, that brings us to another point. Maybe with machines, we can actually move our knowledge much further, and we can understand how to play decent games at much greater lengths. Going back to computer chess, and I don’t know how much time you spend looking at the computer matches. COWEN : Quite a bit. KASPAROV: I think you can confirm my observations that there’s something strange in these games. First of all, they are longer, of course. They are much longer because machines don’t make the same mistakes so they could play 70, 80 moves, 100 moves. So 70 to 80 moves is a normal length. By the way, it’s still 70, 80 moves. It’s way, way below what we expect from perfect chess. That tells us that machines are not perfect. Most of those games are decided by one of the machines suddenly. Can I call it losing patience? Because you’re in a position that is roughly even. There’s maneuvers, and they go around. The pieces are all over, and then suddenly one machine makes a, you may call, human mistake. Suddenly it loses patience, and it tries to break up without a good reason behind it. COWEN : My driverless car will do this someday. KASPAROV : That also tells us — that’s an interesting observation, from also my experience — that machines also have, you may call it, psychology, the pattern and the decision-making. If you understand this pattern, we can make certain predictions. It’s not all predetermined. There’s so many things that can go one way or another. That means that we will not be run by these mystical AI. That it’s not perfect. It’s far from perfect. Still there are so many errors it doesn’t know how to cover, even setting aside a simple question whether the brains can function effectively separately from the body. What does it mean that you have brains that are not part of our moving body? Can you separate it and still have the same effect? My answer is I don’t know. That’s what I believe is good because that tells us that there’s so much to learn in this process. For those who are predicting that AI is just around the corner and it’s going to wipe us out, they don’t know what they’re talking about. COWEN : It’s more like a future of inscrutability and periodic surprises. KASPAROV : We also know that the doomsaying has been always a very popular pastime when it comes to technology because it’s easy. It’s us against them, the race against the machines, the war against the machines. You can sell it. It’s important that people will stop looking at this either from utopian point of view: “Oh, we’ll just get together. We’ll fly to Alpha Centauri with machines. They will serve us within this endless space flight.” Or, to the contrary: “It’s a Skynet that is going to kill all of us.” It’s a process. It’s a process, and we should be objective. This is not too optimistic, not dark pessimism, but something that is more human, which is, let’s get objective. Let’s assess our chances, and let’s realize this is a process of developing further human civilization. Compared to the fears that people had about machines in 19th and 20th century, we’re not pioneers. 原文在这里 https://medium.com/conversations-with-tyler/garry-kasparov-tyler-cowen-chess-iq-ai-putin-3bf28baf4dba
标签: 卡斯帕罗夫