The Electronic Holy War: Computer Go in the New Yorker

[oren]

Well, seen from afar it seems like there is no pro 9x9 play.

I don't think there are any professional games where the victory in 9x9 board is important. I've seen a few games historically, but they're not particularly significant to the players.

Arguably, the denou-sen held recently was the first time 9x9 was important in a while, and the players did seem quite nervous about it.

[Mef]

As someone who just read the article...

Peter Norvig, a director at Google Research and one of the founders of modern A.I., told me that, even if Google or I.B.M. hired a cadre of experts, invested “one hundred times more hardware than anyone else had ever applied to the problem,” and was “very clever about the system-design architecture, the exact machine learning algorithms, and the insights from neuroscience,” he doesn’t know if this would be enough to make the equivalent of Deep Blue for Go.

I think this is putting it lightly...we're quite a ways out from computers challenging top professionals...

[Sampi]

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

I was always curious about such statements, both in chess and in go.
To me, when you play an opponent that can beat you, give you good game, and make you sweat to stay afloat - in other words, one who makes very good moves, what does it matter if it is a computer and a human? Sure, there is a difference in how computer evaluates positions and chooses moves, but such differences exist between human players as well. I would even go as far as to say that there is a lot of value precisely because of that difference.

I mostly hear this kind of stuff from chess, in go it is a relatively new "issue". And while people actively look to watch and play strong player (who us us would pass a chance at a game with a pro?) - they seem to look with disdain when they learn it is a computer who makes such good moves and plays strong enough to beat the human players. Suddenly, such games become uninteresting and such opponents undesirable. Why? Good moves are good moves, and if they are better than your moves, you can certainly learn from them, regardless who makes them, no?

The problem with playing computers is that they make stupid moves as well as good moves, which annoys most humans quite a bit. A strong player might want to be nice and not win by 100 points, and thus make a sub-par move, but a human will never take away it's own points in a ridiculous way or use up ko threats like a computer does. So when they end up winning by 0.5 points you feel as though you were being made fun of by a computer program, some people find this annoying.

[gowan]

The issue of whether/when computer programs will be able to defeat the best human players is really only of serious interest to computer programmers. Caring about saving the honor of the human race simply means too much self-worth is invested in winning. I suspect that this may be why go players become less able to win titles as they reach middle age. By that time most of them have seen what really matters in life and it is not winning tournaments. For us (weak) amateurs we might already have accepted the fact that so many people are better players than we are and we focus on "personal bests" and learning rather than beating everyone else. The same behavior can continue when/if computers are stronger players than all humans.

[Mike Novack]

As someone who just read the article...

Peter Norvig, a director at Google Research and one of the founders of modern A.I., told me that, even if Google or I.B.M. hired a cadre of experts, invested “one hundred times more hardware than anyone else had ever applied to the problem,” and was “very clever about the system-design architecture, the exact machine learning algorithms, and the insights from neuroscience,” he doesn’t know if this would be enough to make the equivalent of Deep Blue for Go.

I think this is putting it lightly...we're quite a ways out from computers challenging top professionals...

I would be rather surprised if none of the teams working on go playing programs using MCTS for evaluation do not already know the limits of their program given essentially unlimited time << time per move a couple orders of magnitude more than normal >> In a case like this where the overwhelming amount of the "crunch" is in one process easily run in parallel << the playing out of a game from a given position to the end and reporting the winner >> the "a hundred times more hardware" and the "hundred times more time" are closely equivalent.

[badukJr]

You could tell the writer wasn't a proper go player. He thought four stones was a small handicap. (Ishida lost by 5 points)

Although the march of the machines is unstoppable, I found it interesting that the pros who are now specifically studying 9x9 in response to the 9x9 programs appear to have improved their results (see the series in Gekkan Go World).

There is, I think, another important point these articles tend to ignore (although the one above hints at it), and that is value to human learners. Chess programs think differently from humans but there is enough overlap, especially on the tactical side, that programs can "explain" their moves to a degree that humans find useful. In go, there is so far no overlap at all and so, apart from satisfying the initial curiosity, playing a computer is less fun than playing thwack-a-mole.

Come to think of it, though, next to nothing is as satisfying as thwack-a-mole.

All relative, Lee Sedol could give other pros 4 stones (like AGA pros), so its not a massive gulf by any means.

Also want to point out margin of victory is deceptive in Go, especially with MCTS programs. They choose the highest win probably, they don't try to maximize score at all.

[EdLee]

Lee Sedol could give other pros 4 stones (like AGA pros)

Maybe 4 is too much; probably 3 -- the AGA guys got beat down to 2 stones,
and were just about to drop to 3 when the tourney was suddenly terminated.

A former insei from China estimates Lee Sedol could give him 2 stones. That's already immense.

[Tim C Koppang]

Good moves are good moves, and if they are better than your moves, you can certainly learn from them, regardless who makes them, no?

I think what you're saying is true to a point, especially when you are studying the game. However, many invest heavily in the idea that a game is meaningful because it is between two people, a conversation if you will. You can't get that human connection with a computer. So a good move by a computer just doesn't carry the same weight. That's not to say playing against a strong computer is useless, but it is a different sort of thing.

[Bantari]

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

I was always curious about such statements, both in chess and in go.
To me, when you play an opponent that can beat you, give you good game, and make you sweat to stay afloat - in other words, one who makes very good moves, what does it matter if it is a computer and a human? Sure, there is a difference in how computer evaluates positions and chooses moves, but such differences exist between human players as well. I would even go as far as to say that there is a lot of value precisely because of that difference.

I mostly hear this kind of stuff from chess, in go it is a relatively new "issue". And while people actively look to watch and play strong player (who us us would pass a chance at a game with a pro?) - they seem to look with disdain when they learn it is a computer who makes such good moves and plays strong enough to beat the human players. Suddenly, such games become uninteresting and such opponents undesirable. Why? Good moves are good moves, and if they are better than your moves, you can certainly learn from them, regardless who makes them, no?

The problem with playing computers is that they make stupid moves as well as good moves, which annoys most humans quite a bit. A strong player might want to be nice and not win by 100 points, and thus make a sub-par move, but a human will never take away it's own points in a ridiculous way or use up ko threats like a computer does. So when they end up winning by 0.5 points you feel as though you were being made fun of by a computer program, some people find this annoying.

Yes, I understand that. As a matter of fact, this was one of my observation when I played some on-line bots, so I share your overall sentiment.

This is why, for the sake of this discussion, I stated that a computer shall make good moves.
In other words - *if* plays well, why not learn from it. Since there are some programs who claim to be high dans and some who can beat pros on 4h, I assume they must make some pretty good moves, on average.

So I guess my question is:
Is it worthless and boring to play against a computer because it is a computer, regardless of how well it plays, or is that attitude attached to the poor state of today's computers and will change when computers get stronger?

[Bantari]

Good moves are good moves, and if they are better than your moves, you can certainly learn from them, regardless who makes them, no?

I think what you're saying is true to a point, especially when you are studying the game. However, many invest heavily in the idea that a game is meaningful because it is between two people, a conversation if you will. You can't get that human connection with a computer. So a good move by a computer just doesn't carry the same weight. That's not to say playing against a strong computer is useless, but it is a different sort of thing.

This is an interesting approach. Thanks.

However, in the context of the game-as-conversation, what you say is that you dismiss a good argument because it comes from a source you dislike. Or something like that.

[Mike Novack]

However, many invest heavily in the idea that a game is meaningful because it is between two people, a conversation if you will. You can't get that human connection with a computer. So a good move by a computer just doesn't carry the same weight. That's not to say playing against a strong computer is useless, but it is a different sort of thing.

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

Assume that you are playing go against an opponent on a server, one that does not allow for side conversations, so the only interactions you have with your opponent are the moves made.

What is the value of this experience to you? Does this value change after the game if somebody tells you "your opponent in this game was a computer" or "your opponent was a computer"? <<you can't tell from the game itself; that's the "Turing Test", tell human from computer>> Having made the assignment of value to the game based on whether you were told human or computer does this value change once you consuder that perhaps you were lied to?

[Tim C Koppang]

However, in the context of the game-as-conversation, what you say is that you dismiss a good argument because it comes from a source you dislike. Or something like that.

I don't think that's what I'm saying at all. What I'm trying to say is that there is intrinsic value in playing a game with another human because the game then transforms into a type of conversation with that person. The game allows us to connect as two human beings. It's sort of like when I read a book, I can connect with the author on some level. If that book was written by a computer, who would I be connecting with?

I'm not saying that playing against a computer is valueless. As a practiced Backgammon player, I have often used a strong computer program to evaluate matches, and to try to learn where I may have gone wrong. Even playing a match against the computer is a learning experience, and one that has value. But that experience is akin to studying, not playing against another person.

From a mathematical point of view, it's perhaps easy to dismiss this point of view. After all, Go can be represented as a math problem. That's interesting, but I don't think that's why everyone plays games. I want to interact with another person, to match my wits against his or hers. Matching my skill against a computer is almost like playing against a room full of pros who can work in perfect collaboration. That's hardly the experience of one mind against one mind.

[Tim C Koppang]

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

Assume that you are playing go against an opponent on a server, one that does not allow for side conversations, so the only interactions you have with your opponent are the moves made.

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...

I don't think I'm skilled enough to tell a human player from a computer player, at least not if either were playing at a sufficiently high skill level. Perhaps I could tell a human apart from a computer if both were roughly at or below my own skill level. But a computer that was slightly above me? -- I think it would pass the Turing Test assuming no other interaction but the moves on the board.

If I played a game on a server, and only learned after that fact that my opponent was a computer program, I'm not sure how I'd feel. I suppose I would at first feel cheated because it wasn't the experience I was hoping for. Later, after thinking about the game, I I would feel -- maybe, interested is the right word? The game would transform from one type of experience to another. Both would be potentially valuable, but for different reasons.

Let me pose another hypothetical that I think falls into the same category. What if you were playing a similar game, but later found out that your "opponent" was really a team of players working together, discussing moves, and otherwise collaborating? How would that change the way you felt about the game? As a potential learning experience, or as a novelty, I'm guessing you'd be interested. However, wouldn't you also feel like the other side cheated because they changed the rules without telling you ahead of time?

[Uberdude]

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

[daal]

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

I don't think so, as long as you appreciate that part of a good fight is taking some licks yourself (otherwise, you're just a bully). To me, go is less like a conversation and more like a boxing match. I play less to enjoy someones company and more to enjoy fighting with them. It's a nice side benefit if they happen to be pleasant people, but I do see that as a side benefit - albeit an important one.

As such, I prefer playing with a human opponent because it feels like a real fight, but a computer can be a good sparring partner too. It doesn't feel pain - but only because it can't. If it could act like it did, I wouldn't mind watching it squirm.