4 ms·

There was a great New Yorker piece a few months back about the new generation of computer-raised chess players, typified by present-day wunderkind Magnus Carlse

There was a great New Yorker piece a few months back about the new generation of computer-raised chess players, typified by present-day wunderkind Magnus Carlsen. The piece is concerned primarily with Carlsen and his rise to fame in the chess world -- but, being a New Yorker article, it's also quite discursive and takes a few fun tangents along the way. One of which is a dissection of computer-age chess players and chess-playing computer algorithms. The general consensus among chess grandmasters is that the best algorithms out there are undeniably impossible for a human to defeat -- but yet, they play "ugly" chess. They're using a brute-force style of AI and can beat a human player down through sheer speed of calculation, and/or perfect recall of vast databases of former games. Someone like Kasparov would argue that a computer chess master is no more the intellectual superior of a human master than a calculator is the intellectual superior of a human mathematics PhD. Finally, the article did confirm a suspicion held among some AI theorists and neuroscientists these days, which is that a combined human/AI "team" can defeat any lone human or lone AI. In the case of folks like Carlsen, the thinking is that this new generation will soar to new heights precisely because it has trained with and against computers, thus piecing together a best-of-both-worlds approach.

gnosis 15y ago
"Someone like Kasparov would argue that a computer chess master is no more the intellectual superior of a human master than a calculator is the intellectual superior of a human mathematics PhD." Except that there's a lot more to being good at math than just being good at calculation and memorization; not so in chess (at least not against computers). Computers have proven that they've got what it takes to win the game. And, ultimately, that's all that really matters in chess as it is presently designed. Of course, there are plenty of chess variants (not to mention games like Go) which are a lot harder for computers to play well. It is quite likely that that's where most serious chess players will migrate to over time. In fact, for the last decades of his life, Bobby Fischer himself refused to play anything but Fischer Random Chess (also known as Chess 960), a variant which made it practically impossible to memorize a significant portion of opening theory, since in this variant there is not just a single opening position as in regular chess, but 960 of them.
- geebee 15y ago
  Do you think chess players will migrate to Go because they don't like playing a game that they can't win against a computer? I'm nowhere near good enough to pretend to understand the motivations of a chess master, but this does kind of make sense to me. Somehow, it kills the vibe to know that computers will always win. Funny, humans still enjoy sports even though a human sprinter will never run as fast as a motorcycle. Maybe it's because we've always been physically less capable than animals, maybe it hurts more to be bested intellectually. Horses have always run faster than us, so we're primed to see the motorcycle as a tool rather than a challenge. In sports, maybe we're better at celebrating human achievement without comparing it to something else. Like most people, I don't actually think humans really were bested by computers in chess. There is a huge difference, of course - the computers are winning by following a completely different "thought" process than a human. It's an astoundingly impressive accomplishment to build a program that can beat chess masters, but because the way a human plays is so different, I think that I understand what the calculator analogy is getting at. In some ways, the "victory" shows the severe limitations of computers - a human chess grandmaster, capable of holding and evaluating maybe a dozen positions, is capable of playing evenly with a computer that evaluates hundreds of millions of positions a second. And in Go, that dozen or so positions combined with intuition, strategy, and other mental processes we consider almost subjective is so effective that a billions of positions a second computer can't even come close to winning. If go eventually falls because computers are now capable of evaluating trillions of moves per second, but the approach is still essentially number crunching, well that's an impressive engineering feat, but "human intelligence" won't be any more defeated than it was the first time, with chess. Now, if a computer can beat a human by evaluating fewer moves, ok, maybe then we've been bested.
  - jonnathanson 15y ago
    If I'm following the line of reasoning correctly, it's that chess players will migrate to a game like Go because it can't be reduced as easily to positional values and min/maxing of said values. While it's true that chess has an absurd amount of possible moves, combinations, and games, computers are showing us that chess strategy and theory are essentially meaningless in the face of brute calculational ability. From a poetic point of view, you could argue that computers are sucking the mysticism and romance out of the game. From a practical point of view, you could argue that computers are steadily making the game pointless for humans. Either way, it's a less exciting game now that it's been reduced. Kasparov seems to be an optimist, for what it's worth, and in several of his disquisitions on this topic, he argues that a computer program will never "solve" chess the way a program solved checkers a few years ago. There are too many numbers to crunch. Furthermore, there are too many weaknesses in AI's ability to form strategies or think creatively. Unfortunately, this line of reasoning assumes that creativity is necessary for success in chess. It might be in a human-vs-human game, but in absolute terms, evidently it's not.
BasDirks 15y ago
Games by Rybka[1] and Houdini[2] have often been amazingly beautiful. 1. http://www.rybkachess.com/ http://www.rybkachess.com/ 2. http://www.cruxis.com/chess/houdini.htm http://www.cruxis.com/chess/houdini.htm

gnosis 15y ago

"Someone like Kasparov would argue that a computer chess master is no more the intellectual superior of a human master than a calculator is the intellectual superior of a human mathematics PhD." Except that there's a lot more to being good at math than just being good at calculation and memorization; not so in chess (at least not against computers). Computers have proven that they've got what it takes to win the game. And, ultimately, that's all that really matters in chess as it is presently designed. Of course, there are plenty of chess variants (not to mention games like Go) which are a lot harder for computers to play well. It is quite likely that that's where most serious chess players will migrate to over time. In fact, for the last decades of his life, Bobby Fischer himself refused to play anything but Fischer Random Chess (also known as Chess 960), a variant which made it practically impossible to memorize a significant portion of opening theory, since in this variant there is not just a single opening position as in regular chess, but 960 of them.

geebee 15y ago
Do you think chess players will migrate to Go because they don't like playing a game that they can't win against a computer? I'm nowhere near good enough to pretend to understand the motivations of a chess master, but this does kind of make sense to me. Somehow, it kills the vibe to know that computers will always win. Funny, humans still enjoy sports even though a human sprinter will never run as fast as a motorcycle. Maybe it's because we've always been physically less capable than animals, maybe it hurts more to be bested intellectually. Horses have always run faster than us, so we're primed to see the motorcycle as a tool rather than a challenge. In sports, maybe we're better at celebrating human achievement without comparing it to something else. Like most people, I don't actually think humans really were bested by computers in chess. There is a huge difference, of course - the computers are winning by following a completely different "thought" process than a human. It's an astoundingly impressive accomplishment to build a program that can beat chess masters, but because the way a human plays is so different, I think that I understand what the calculator analogy is getting at. In some ways, the "victory" shows the severe limitations of computers - a human chess grandmaster, capable of holding and evaluating maybe a dozen positions, is capable of playing evenly with a computer that evaluates hundreds of millions of positions a second. And in Go, that dozen or so positions combined with intuition, strategy, and other mental processes we consider almost subjective is so effective that a billions of positions a second computer can't even come close to winning. If go eventually falls because computers are now capable of evaluating trillions of moves per second, but the approach is still essentially number crunching, well that's an impressive engineering feat, but "human intelligence" won't be any more defeated than it was the first time, with chess. Now, if a computer can beat a human by evaluating fewer moves, ok, maybe then we've been bested.
- jonnathanson 15y ago
  If I'm following the line of reasoning correctly, it's that chess players will migrate to a game like Go because it can't be reduced as easily to positional values and min/maxing of said values. While it's true that chess has an absurd amount of possible moves, combinations, and games, computers are showing us that chess strategy and theory are essentially meaningless in the face of brute calculational ability. From a poetic point of view, you could argue that computers are sucking the mysticism and romance out of the game. From a practical point of view, you could argue that computers are steadily making the game pointless for humans. Either way, it's a less exciting game now that it's been reduced. Kasparov seems to be an optimist, for what it's worth, and in several of his disquisitions on this topic, he argues that a computer program will never "solve" chess the way a program solved checkers a few years ago. There are too many numbers to crunch. Furthermore, there are too many weaknesses in AI's ability to form strategies or think creatively. Unfortunately, this line of reasoning assumes that creativity is necessary for success in chess. It might be in a human-vs-human game, but in absolute terms, evidently it's not.

BasDirks 15y ago

Games by Rybka[1] and Houdini[2] have often been amazingly beautiful. 1. http://www.rybkachess.com/ http://www.rybkachess.com/ 2. http://www.cruxis.com/chess/houdini.htm http://www.cruxis.com/chess/houdini.htm