The Gauntlet is Thrown!
Tic-Tac-Toe is broken. Checkers was recently broken. Chess is on the verge of being broken.
With these realities in mind, Frank Lantz and I have made a bet: Go will be broken within 50 years. I am betting that it will be broken, while Frank is betting that it can last much longer. The winner will buy the loser the nicest Go set that they can find.
You are all now witnesses. Keep us honest. See you October 19th, 2057.
You guys should post this one on longbets.org. They need some more up-to-date bets that are longer than 10 years. Plus it will become a little more in the public record.
Interesting bet. I think within 50 years, too much is possible. A better question is when do you think chess will officially be broken? Do you think it will happen in the next year, 5 years, 10 years?
I personally think it will be longer than these programs make it seem. There are too many possibilities for it to be completely broken just because a computer program can beat humans. Look at the insane amount of possible board combinations.
Tic-Tac-Toe has so few, i bet its less than 200. Checkers has lots more, but still not as much as chess; after all chess uses twice the amount of squares, and has more than 2 types of pieces. (Checkers has a regular piece and a king, which can be either red or black. Chess has pawns, rooks, knights, bishops, a king and queen(s), all of which can be white or black.) This creates thousands of more possibilities.
Go presents an interesting problem with board combinations, while it has many more squares to place your pieces, it only has 2 types of pieces, white or black. But also, every game is more about the order the pieces get placed, so putting two pieces on the board in a different order can dramatically change the game. So things can change very dramatically with just small changes in the game, but I still think it can be broken. The question becomes, when?
Another question is what does it mean the game is broken? Is it that there is one perfect way to play, and if both players do so, they end in a tie? Or is it that there is one perfect way to play and you always win? This is another interesting question we have to answer.
So lets start taking bets; when do you think chess will be broken? And when do you think go will be broken? And also, what do you consider having it broken? Give the year you think it will be broken and what officially makes it broken, and lets get a pool started.
Ill go now:
Chess- 2017- an always win strategy for the first player to go
Go- 2027- an always win strategy for the first player to go
Are you mixing up two terms? It seems like you mean Checkers has been “broken” because it has been completely solved to the point where one could just use a brainless algorithm to win (or tie) any game. The Chess link, on the other hand, seems to be about a program that’s just capable of beating a grand master player. That’s different. It’s one thing to beat Kasparov, but something quite different to be able to say that you have proof that there’s a pattern of moves that no one could ever beat.
I’d give computers about 30-35 years to reach the point where they could beat any human player at Go. (See this chart from Kurzweil’s book: http://en.wikipedia.org/wiki/Image:PPTExponentialGrowthof_Computing.jpg)
The Checkers article say that there are about 10^20 possible game positions in Checkers. Wikipedia says there are about 2*10^170 possible game positions in 19×19 Go. So Go is about 2*10^150 times as complex. If computing power doubles every eighteen months (Moore’s law), it would take about 750 years to make up that difference. (2^500 = ~3.3*10^150)
So will Go be ‘solved’ within 50 years? Doesn’t seem likely.
Yes, the bet is whether Go will be *solved* by 2058. I’m pretty sure I’m correct, and Charles is seriously underestimating the difficulty of this problem.
Btw, this was part of a conversation about how the meaning of a game as a formal system, a kind of a large-scale math problem, is related to its meaning as an aesthetic object. Once a game is solved, how does this affect our exeperience of it, even if we were never expert-level players to begin with?
And what happens if general computing power and techniques got good enough to solve *all* games, and all possible games? How would that affect their current status as problem spaces that humans collaboratively explore?
Charles maintains that this would (will? is currently?) cause us to focus on the social and expressive power of games, the particular way that *I* play checkers, as an expression of my personality, and the particular psychological and emotional meanings that Shinji Mikami (or whoever) expresses through the game mechanics of a single-player game.
I maintain that 1. this supposed world of total solvability is not at all close and perhaps not even theoretically possible (Captain Kurzweil’s kooky singularity notwithstanding) and 2. even if it were, the design and play of games that weave the recursive and inherently human traits of behavior prediction and psychological second-guessing into the mix (what David Sirlin calls “yomi”) is more than enough to keep this danger at bay. This is one of the reasons I think a game like Poker is currently more interesting than Chess from a theoretical or AI point of view.
I think the social, personality-expressive, “style”, aspect of the enjoyment of multiplayer games is indeed important. But I think this aspect takes place *within* the context of the collaborative problem-solving, not as a replacement for it.
-f
Btw, Charles, you do know that a really good Go board can cost up to $20,000 right? I like shopping. http://www.kiseido.com/goban/hyuga.htm
My point isn’t so much that the ability for computers to best their human opponents or ‘solve’ certain games is changing the way that people play (though I think this is true), but that a greater emphasis on expressive play and mechanics is already happening. What I disagree with is that this is simply a symptom of games becoming ‘interactive stories’ rather than games. What I think is happening is that the importance of games as collaborative problem-solving is diminishing (though not to the point that it’s unimportant) and their potential as expressive and emotional vehicles is finally being explored. There are a lot of catalysts for this trend, and the reality that the highest level of play is no longer the play of human beings is just one.
Will Go be solved in 50 years? I hopes so, because $20,000 is a hell of a lot of money, especially when you consider inflation.