Moderators: hgm, Rebel, chrisw
The Mystery of Go, the Ancient Game That Computers Still Can’t Win.
While computer Go resembles computer Chess from 90's, both are/were not "completely mastered by humans", as you expressed it. Computer Chess in 90's on x386, x486, early Pentiums, was reaching 2,300-2,400 human Elo level, more than 99% of humans are below that, and it was "mastered" only by GMs against bots. Similarly, some Go engines now on 20-cores reach 6 dan level, which is higher than 99% of human Go players, and only Go pros and the strongest of amateurs master it against bots.pocopito wrote:I learned to play Go a few months ago and I've kind of wondered about the same question.
It seems the situation of computer Go is similar to the one of computer chess in the 90's. You say that there is not many money to be made on this field, but I'm not that sure about it (for example, the android app for playing on KGS, a famous GO server costs 7.40€. Of course it has nothing to do with building a Go engine, but it can give you an idea that there are people who spend money on it). It seems that in spite of the advances that have been recently done, Go is for the moment completly mastered by humans. For example, AFAIK, actual softwares don't have yet a proper post mortem analysis feature.
Currently Go looks more like a challenge for programmers than chess; the receipt for a strong chess engine is already known (not saying that it's easy to do, but that the technical knowledge is right there for whoever wants/can use it), but for Go it seems there is a long road to go over.
Best
E Diaz
Similar thoughts crossed my mindLaskos wrote:While computer Go resembles computer Chess from 90's, both are/were not "completely mastered by humans", as you expressed it. Computer Chess in 90's on x386, x486, early Pentiums, was reaching 2,300-2,400 human Elo level, more than 99% of humans are below that, and it was "mastered" only by GMs against bots. Similarly, some Go engines now on 20-cores reach 6 dan level, which is higher than 99% of human Go players, and only Go pros and the strongest of amateurs master it against bots.pocopito wrote:I learned to play Go a few months ago and I've kind of wondered about the same question.
It seems the situation of computer Go is similar to the one of computer chess in the 90's. You say that there is not many money to be made on this field, but I'm not that sure about it (for example, the android app for playing on KGS, a famous GO server costs 7.40€. Of course it has nothing to do with building a Go engine, but it can give you an idea that there are people who spend money on it). It seems that in spite of the advances that have been recently done, Go is for the moment completly mastered by humans. For example, AFAIK, actual softwares don't have yet a proper post mortem analysis feature.
It seems only a question of time, maybe 10-20 years, until bots will become stronger in Go, even compared to Go pros.Currently Go looks more like a challenge for programmers than chess; the receipt for a strong chess engine is already known (not saying that it's easy to do, but that the technical knowledge is right there for whoever wants/can use it), but for Go it seems there is a long road to go over.
Best
E Diaz
Kudos to Remi Coulom! His program Crazy Stone is widely mentioned as the pioneer of MCTS/UCT, that revolutionized Computer Go back in 2005. And now it is the strongest, or one of the strongest, with a 6 dan leval on home PC. Very few human players are above 6 dan, and that means that Go programs will soon be on par with the Kasparov's of Go, and eventually be on another level, just like in Chess.Eelco de Groot wrote:I was just wondering, I sometimes see a post here about computer Go, and vaguely remember that people like Vincent Diepeveen seemed to have tried their hand at computer Go programming. Was wondering if there are some strong programs these days using new heuristics. Maybe Vas gone to computer Go? Well, I suppose not, because there is not much money to be made there and he needs to support a family.
I came across a fairly recent article that gives an introduction, it seems relatively easy to follow for us dumb chess players
I see one person there, not unknown to us here in computer chess, who has taken up the challenges of a new game:
Rémi Coulom (left) and his computer program, Crazy Stone, take on grandmaster Norimoto Yoda in the game of Go. Photo: Takashi Osato/WIRED.
I tried to elicit some interest:Eelco de Groot wrote:Thanks everyone for the replies and comments! If we had a way to display Go diagrams, maybe we could make a subforum for it!