The Electronic Holy War: Computer Go in the New Yorker

[Mike Novack]

Tim, are you saying that a go playing computer program would fail the "Turing Test"?


Interesting question. However, I think limiting the test to a server that does not allow for chat is a bit like cheating. I generally don't play games where I have no chat option. Maybe that says something about my personality, and what I'm looking for in games. Nonetheless, I'll try to answer in the spirit in which the question was asked...

Well ...... let's say that "chatting" was allowed. The current state of the art (actually going back decades in this regard) is that an AI program might be able to fake passing the Turing Test in chatting provided the realm of discourse (what you are chatting about) were limited. By "fake" I mean a sort of cheating where the program uses tricks to get out of situations where it can't figure out what you "said" (typed). Ranges from "Doctor" asking about your mother, "Parry" getting aggressive, etc. Here I think some set of "let's concentrate on the game", "let's stick to talking about go", etc. might do the trick to fool a human.

But more to the point, I think you missed the final part of what I was saying. Yes I understand, based upon what you "found out" later your feelings about the game, the value of the game, etc. would change. However note that this was the result only of your belief, not about any reality << I did include the possibility that you might have been misinformed >>

Yes, told that it was a team that beat you you feel "cheated". But that's not because it was a team against you but because you were told a team against you and you believed that.

[xed_over]

Yeah, the article makes it sound like computers are already at pro level, it's very misleading IMO.

what I find misleading is that the article is about a game that was played a year ago (its buried in there somewhere that the game was played last March).


March of this year, in the 2nd Denseisen, Crazy Stone won again, this time against Yoda -- and also on 4 stones.

And, like last year, Zen also played our pros on 4 stones -- but lost.
As far as I can tell, in both the 7th UEC (this year) and the 6th (last March), Zen won overall over Crazy Stone.

[moyoaji]

While we were worrying about computers coming onto the scene, apparently the chickens managed to cross the road into go solving...

Researchers Solve 9x9 Go using Chickens

Humanity had a good run, but I guess we need to let the superior intelligences of the world have their way with our game.

[Bantari]

A computer doesn't feel pain when you kill them. Isn't that part of the joy of playing or am I unusually sadistic?

Heh... if you enjoy other's pain, then yes, this is pretty much the definition of "sadistic".
Also, it begs the question - how do you feel about playing masochists who enjoy pain? No fun?

[Tim C Koppang]

But more to the point, I think you missed the final part of what I was saying. Yes I understand, based upon what you "found out" later your feelings about the game, the value of the game, etc. would change. However note that this was the result only of your belief, not about any reality << I did include the possibility that you might have been misinformed >>

Yes, told that it was a team that beat you you feel "cheated". But that's not because it was a team against you but because you were told a team against you and you believed that.

I'm sorry, Mike, but I don't think I'm following you. If I understand correctly, you're making a distinction between my belief about who or what I'm playing and the objective reality of the situation. However, I'm not sure how those two things are different in the example you're giving.

I could believe that I'm playing a computer, or I could believe that I'm playing a human. Either of those things could be objectively true given that the game is happening over the internet. I'm saying that, yes, my attitude towards the game would change based on who I believed I was playing -- but, I'm going to base that belief (regarding who I'm playing) on what I'm told before the game begins. If I visit KGS, for example, and get into a game, I'm going to know ahead of time if my opponent claims to be a bot or a human.

Are you asking me how I would feel if I started playing a game and knew nothing about what my opponent was ahead of time? If so, I have no idea. If I were to guess, I suppose I would assume human, and then feel cheated if I found out afterwards that it was actually a computer. I'll reserve my explanation for why that is until after you respond. I've made a lot of assumptions, and I want to hear from you first.

[happysocks]

Lol'ed at this bit from the article:

“the status of one group can affect that of its neighbors—like a cowboy who points a revolver at another cowboy, only to find himself covered by a rifleman on a roof.”

Just insert own name at the end.

[John Fairbairn]

March of this year, in the 2nd Denseisen, Crazy Stone won again, this time against Yoda -- and also on 4 stones.

And, like last year, Zen also played our pros on 4 stones -- but lost.

I was just catching up with this event (the Electronic Holy War) for GoGoD, and I came across a significant piece of information that will alter perceptions. In both games there was komi of 6.5, so the difference was half a stone more or less than stated, which is quite a lot at this level.

The trouble is, I don't know who gave komi. The Japanese says "senban" which means "first to play", but it does not state who is first to play. Obviously the first numbered stone is White 1, but it can be argued that Black's placement of handicap stones counts as first move, and senban usually does refer to Black.

The report also says, incidentally, that Japanese rules were used. I wonder how the games finished. Unfortunately the report only gives the first 100 moves for the CS game (CS won by 2.5) and it gives 229 moves for the Zen game, which appears to be a complete record, but the result caption says "Black's resignation was unavoidable".

Can anyone cut through the murk?

[hyperpape]

Hiroshi Yamashita posted a link with what he says are official SGFs. http://computer-go.org/pipermail/comput ... 06588.html

[xed_over]

Hiroshi Yamashita posted a link with what he says are official SGFs. http://computer-go.org/pipermail/comput ... 06588.html

once you get them, you can add "CA[Shift-JIS]" to the end of the first line (without the quotes, of course) in order to properly render the name of the game title: 無題 (which means "untitled")

[John Fairbairn]

Hiroshi Yamashita posted a link with what he says are official SGFs. http://computer-go.org/pipermail/comput ... 06588.html

Thank you. These give the komi as 0.5, which accords with normal computer practice of course, but also fits the result of the one game that was scored (and given in full in the link). Looks like Gekkan Go World has major mistakes, of the type that brought back memories of T Mark's gnashing teeth!

Among the comments I found interesting were that in the CS game Black was ahead by 20 points after 100 moves. This suggests he has already lost 20 or more points through strategic ineptitude, and almost all of the remaining 20 points were apparently lost through a tendency for the program to play inside its own territory when it thought it was winning.

It was also observed that CS has a tendency to emphasise the centre, which can work easily with large handicaps but might be a lot trickier in an even game.

Zen, also a member of the centrist party, appeared to do much better than CS in that it had maintained more of the 4-stone advantage after 100 moves, but lost (by the narrowest of margins) through an endgame mistake.

Yoda said CS made better moves than he expected, and in fact was totally different from what he was expecting. It seems he was expecting lots of stupid fighting. I think we may reasonably infer that he must have based this on typical human behaviour. There's a lesson in there!