NYT Story On Go Programs And AI

mykej writes: "The NYT (registration required, blah blah) has a story on Go, the hardest game for computers to play. From the article: 'Programmers working on Go see it as more accurate than chess in reflecting the ineffable ways in which the human mind works. The challenge of programming a computer to mimic that process goes to the core of artificial intelligence, which involves the study of learning and decision-making, strategic thinking, knowledge representation, pattern recognition and, perhaps most intriguingly, intuition.' There are a few throwaway lines about Nash from 'A Beautiful Mind,' although they don't mention the game he invented after getting frustrated with the inconsistencies of go."

Comments

Go is harder or.....?

[jeffersonebell]

My question has always been is Go really that much harder for a computer to play than Chess or is Chess just more popular and more energy has been devoted to developing computers to play it?

Re:Go is harder or.....?

[paradesign]

im sure that is a minute fraction of it but, if you read the article,

A Go-playing computer would take about 30,000 years to look as far ahead as Deep Blue can with chess in three seconds... ... If processing power were all there was to it, the solution would be simply a matter of time, since computers are growing ever faster. But the obstacles go much deeper. Not only do Go programs have trouble evaluating positions quickly, they have trouble evaluating them correctly.

RTFA

Re:Go is harder or.....?

[freuddot]

Go is definitively harder.

Disclaimer : IAAPP ( professional programmer ) and IAAGP ( go player ) ;-)

The trick is not about the branching factor that is quite high in go, and small in chess.

The thing is that in go many local battle are fought on each region of the board. Each of those battle are usually fair. Fighting more for one region will make it yours. However, during that time, the opponent will secure another region.

So far, no problem, use the divide-and-conquer method, solve every region, and then use a sum-of-game technique to play the whole board. However this doesn't work. Every region has many ways to be fought over, and the way you fight in a region will affect all the other region of the board.

Professional players just *know* or *feel* that playing in a certain way will help another region. They have a very informal perception the relationship between the regions. This is something we don't know how to model. Usually people will refer to it as instinct. I tend to believe that it is the years of practice that enable pros to see those pattern.

Also, Go seems to be only a grid with either nothing, a white or a black stone. In fact, much higher-level concept are seen by go players, and as long as we don't model those in a go AI, go AI will suck.

See sensei to get an idea of the high-level concepts we need to model to program a Go AI. BTW, this is a cool wiki board about Go. Great place to learn.

So, when we'll be able to model high-level stuff like that and program AI rather than do brute-force hacks like Deep Blue, we'll have a Go AI. In the meantime, we humans rule.

Re:Go is harder or.....?

[Anonymous Coward]

There are many reasons why go is harder to pogram than chess. Among them:
1. The game space is much larger, perhaps 220 half moves with typically a dozen plausible alternatives per move, compared with, say, 100 half moves and typically fewer plausible alternatives. This alone comes to dozens of orders of magnitude.
2. At various times in the game, quite different strategic issues become most important, sometimes accurate reading, sometimes strategic placement, sometimes choice of board area, sometimes choices between "attack" or "reduce area", etc. Programs tend to be bad at this sort of planning.
3. The large branching factor and the interplay between openings in different corners mean that the opening book, while vast, is much less stereotyped than in chess. A corner opening which is good in one situation can be bad in another, the difference being the presence of a stone or two in remote areas of the board.
4. The fact that the stones, once played, pretty much stay put means that it is relatively easy for humans to visualize long sequences. Even a week-end player may need to visualize the result of "ladder" sequences 50 half-moves long, and can do so in a few seconds. Forcing computers to deal with this search horizon all the time -- or figure out when not to -- makes the searches even more of a bottomless void.
In conclusion, though chess is much higher profile in the west, considerable effort here and in Japan and to some extent China has been put into go. There are many reasons why go is indeed harder to program -- much harder.
The strongest programs in the world are currently a bit better than 10 kyu on the Japanese scale. This is a level which is exceeded by more than 80% of the entrants at the US Congress. (Which will take place in Chicago next week, for anyone interested in seeing real go.)
-- David Erbach

Re:Go is harder or.....?

[mangu]

Computers don't have intuition, therefor they cannot play games like go very well.

What has one thing got to do with the other? "Intuition" is a sort of logic you cannot explain, and is, most likely, wrong.

Re:Go is harder or.....?

[anonymous_wombat]

To put this more succinctly, in addition to the higher branching factor, coming up with a decent static evaluation function is even more difficult. The tree size could be dealt with eventually, but it is extremely unlikely that there is a decent static evaluation function like for Chess. The DeepBlue static evaluation function only evaluates four different aspects of the game. Try that in Go.

Not news at all

[bfwebster]

I took a grad-level AI class in college nearly 30 years ago; our final exam was a round-robin tournament among Go-playing programs that we had to write. (More precisely, we each wrote two routines--one to evaluate the board, one to generated a list of moves--and a minimax framework called our routines.) It was a great introduction as to why AI is hard.

I still play Go occasionally, and though I am a mediocre player at best, I can usually beat any Go-playing programs that I've found. ..bruce..

Kasparov and IBM

[natpoor]

The NYTimes is not exactly correct about the Kasparov/Deep Blue match. The IBM programmers studied Kasparov's playing style intensely, and programmed Deep Blue to not just play chess but more specifically play and beat Kasparov, which is a slightly different thing from "playing chess." (Granted the machine could still beat almost anyone, but maybe not other masters with a different playing style.) Kasparov, on the other hand, was not allowed to study how Deep Blue might play at all. I also recall that Kasparov became a bit unhinged early on. So yes, Deep Blue did beat Kasparov, but the problem for it was not just "play chess" it was "beat Kasparov."

Re:Kasparov and IBM

[LoudMusic]

And for all you Trek fans out there - remember the great Moriarty episodes? "Computer, design a foe good enough to challange Data"

http://www.startrek.com/library/tng_episodes/epi so des_tng_detail_68364.asp

We're in deep shit when this kind of AI programming is readily available.

~LoudMusic

Re:Kasparov and IBM

[God! Awful]

At one point, I was trying to improve my chess game by studying the game archive that comes with ChessMaster. After only a few weeks of practice, I discovered that I could predict each move (in the midgame) with uncanny accuracy (80%). However, my chess game didn't actually improve. All I had done was train my brain to be a fuzzy logic analyzer for predicting Kasparov moves against high quality opponents. The basic strategy fails miserably against amateur players (who tend to be less subtle in their attack). One of the ways Kasparov came back to beat Deep Blue in their original match was to suddenly switch strategies to something the machine was not expecting.

-a

Re:Kasparov and IBM

[Paradise Pete]

programmed Deep Blue to not just play chess but more specifically play and beat Kasparov

Deep Blue was prepared for the match the same way any other chess player would have prepared for the match - by studying the opponent's games. This is not some trick, it's standard preparation. This is always done in teams, and during the match players always have seconds (as in assistants) to help them analyze and study the games played so far. And if a game is adjourned for the night the seconds stay up analyzing the position and in the morning go over their findings with the player. That's the way matches are played.

Kasparov's complaint was not that the program was prepared in this way, but rather that he did not have sufficient opportunity to prepare for the program.

History on Go

[RobinH]

Try this site.

It also has instructions on how to teach Go, if you're interested.

Re:History on Go

[ergo98]

this pdf also offers an extremely good tutorial of the game.

Go and movies

[paradesign]

is it just me or does any movie that has to do with number theory/ mathematics have Go in it somewhere.

the two that come to mind as striking examples are "a beautiful mind" and "Pi"

im sure ther're others. theories? other movies? is it just a trend?

Re:Go and movies

[God! Awful]

is it just me or does any movie that has to do with number theory/ mathematics have Go in it somewhere.

I don't remember any references to Go in "Good Will Hunting" or "Cube".

-a

Re:Go and movies

[Queuetue]

In my experience, most people who deal with number theory/abstract mathematics play go... It may be one of the few technical details that Hollywood gets right...

Re:Go and movies

[William Tanksley]

The mathematical facility of the character isn't a/the focus of the story. It is, really, quite incidental. You could have replaced "mathematics" with "woodworking" and the main story would still hold.

Of course. In which case it would be a story about a woodworking genius instead of a mathematical one. Just as A Beautiful Mind could have been about an insane metalworking genius instead of a mathematical one.

Good movies are about people, not professions or technologies. The fact that you can replace mathematics with something else doesn't suggest that GWH wasn't a mathematical movie; it simply means it was about MORE than math.

As every good movie inevitably will be.

-Billy

A tidbit about Go

[Blind Linux]

lie in the way that the decisions are made and the differences in how they affect the playing field. The average game of Go actually lasts longer than the average chess game and is far older...
For starters, Go in its pure form is played on a 19x19 board as supposed to an 8x8 board. Chess's famous plays, games and styles have all been archived, whereas Go's strategies are largely abstract and can only be learned by repeated play. The game only begins to take structure after 30 to 50 moves. According to this site, Go has approximately 10 to the 750th power of possible board positions. This makes it a very hard game for computers to learn.
On the historical side, Go is a complex game that originated in China close to 4000 years ago and has remained constant to its' original form despite being introduced to many southeast Asian countries since.

Re:A tidbit about Go

[flipflapflopflup]

Whilst I'm personally a Go fan, not a chess fan, I don't think I agree with your arguments.

>Go in its pure form is played on a 19x19 board as supposed to an 8x8 board
So? What has the size of a board got to do with it? In chess you can move pieces around, in GO you cannot.

>Chess's famous plays, games and styles have all been archived, whereas Go's strategies are largely abstract and can only be learned by repeated play
Not really true. There are masses of games available as .sgf files that you can study to your hearts content. THere are many clasic moves to make in certain positions, etc.

>The game only begins to take structure after 30 to 50 moves.
Again, not really. THere are masses of standard opening patterns (fuseki), and also many standardised plays (joseki) that can go on during a game. A whole lot goes on in the first 20 - 30 moves to shape the rest of the game.

Go is a great game, it doesn't need imbalanced comparisons with chess to prove it, you just need to play the game a while to realise that. Maybe you should try.

Re:A tidbit about Go

[Triv]

Go in its pure form is played on a 19x19 board as supposed to an 8x8 board

Not to be anal, but you're comparing squares to intersections. Chess is played in the squares of a board 8 squares by 8 squares. Go is played on the intersections of a board 19 points by 19 points.

You can play a fast game of go on the intersections of a standard chess board - your typical lunch break 9X9 games. Saves the trouble of carting two boards around provided you don't mind the colored squares. :)

Triv

Re:A tidbit about Go

[flipflapflopflup]

You are absolutely right. I re-read my post, and realised that it does indeed sound condescending towards the original poster (Blind Linux).

After reading Blind Linux's post, I got the feeling that he had not actually played Go, just read about it, so I thought suggesting he played it would be a good idea. However, things can often come across in the wrong way when typing.

So, sorry Blind Linux.

Re:A tidbit about Go

[flipflapflopflup]

Sorry Blind Linux, if my original post came across badly. I didn't mean to sound rude, I simply meant that if you have a go at Go, you will get a much better feeling for the game than just by reading about it.

Re:A tidbit about Go

[flipflapflopflup]

The best I know of is a cut down version of "The Many Faces of Go" - called igowin, windows only I'm afraid, and you can only play 9x9 games. After that, try GnuGo. You can compile that for anything.

Your best bet is getting a book (I reccomend Teach Yourself Go), and playing 9x9 games until you get the hang of things. Then try one of the online servers: IGS or KGS. Have fun!

Hex

[Salamander]

Actually, Hex was first invented by Piet Hein, who is perhaps better known for the Soma cube. Nash claims to've invented the game independently, but somehow I just find that hard to believe.

Re:Hex

[Salamander]

No comparison. The work for which he was awarded the Nobel prize was verifiably original. Nobody's found a paper containing those exact same ideas published twenty years earlier, at any rate, and yet that's pretty much exactly the case for Hex.

Thanks for playing. Now get back under your bridge.

Re:Hex

[tswinzig]

Actually, Hex was first invented by Piet Hein [ctaz.com], who is perhaps better known for the Soma cube. Nash claims to've invented the game independently, but somehow I just find that hard to believe.

Me too! It's like two people from two completely different parts of the world claiming to have theorized natural selection at the same time.

Preposterous!

Re:Hex

[identity0]

Actually, if you read the book "A beatiful mind", Sylvia Nassar has corroboating stories of Nash's invention in 1949 - after Piet Hein's invention, but before it was marketed by Parker brothers. It was apparently a bit of a fad there at the time.

Parallelism required?

[mikewas]

The trend towards Massively Parallel Computers, such as the STARAN developed in the 50s/60s at Goodyear Aerospace Corporation by Ken Batcher were discarded for the most part. Pipelined machines were easier to design, cheaper to build, and easier to program (i.e. could use existing languages).

It would seem that a Massively Parallel Processor would be ideal for this applications, especially a STARAN with it's large Content Addressable Memory. Or do I, as a former STARAN user & developer of similar machines, just see this as a nail since I have the hammer in my posession?

Re:Parallelism required?

[laertes]

I think that massively parallel processing would be fairly maladapted as a platform for a Go AI. The basic problem in Go for deciding what move to make right now is this: given the enourmous number of possible moves that can be made right now, limit the list to a small number of moves, and pick the best one.

While generating the "short list" can be made parallelized to some extent, the much harder problem is deciding among those. Because you can't apply traditional game theory methods (where you would decide directly what move would lead to a better end game), you have to analyze each of these alternatives by comparing the goal they are trying to achieve, and the likelyhood of doing so. This, at least, is how I feel humans accomplish choosing a move. Deciding the most worthy goals seems to me to be a serialized task, largely inappropriate to MPP.

In my personal experience playing Go, I almost always see the best potential moves right away (at least, the best ones I will ever see). Then, it takes me quite a while to decide among those moves. There are a few situations (Joseki) where I attempt to construct a mental tree of moves, but computers are already okay at those anyway.

Still, there's nothing you can do on a single processor machine that cannot be done on a MPP, and vice versa. Since we can't figure out how to do the "goal juggling" I mentioned at all, it doesn't matter what architecture we use, not even if we use Blue Gene.

Re:Parallelism required?

[mikewas]

You say that "...I almost always see the best potential moves right away..." that sounds like parrallel processes happenned, even though your subsequent analysis is a serial process. the serial processes are more explicit & you're more concious of them.

A later poster mentions that he "is good at tactics" but can't manage the entire board at one time. This also sounds like fertile ground for parrallelism.

Just my opinion, it's not like I'm actually an experienced go player nor have I tried to program the game.

rithmomachia

[dustmote]

I was never a big fan of hex. My favorite game right now is Rithmomachia (or rythmomachia), but it's not good for AI stuff, since it's kinda based on simple number theory. Apparently, it competed with chess for a couple of hundred years as the big intellectual board game before essentially falling into obscurity. Rules are at
http://www.gamecabinet.com/rules/Rithmomachia.html

Go programs

[Janitorial_One]

For an example of how elementary a regular Go AI is, visit kgs.kiseido.com and click "Play Go Against Your Computer". You'll notice how sometimes the computer passes for no reason, or continutally defaults the game in certain stone patterns.

OT: Nash's game, Hex

[droid_rage]

Hex was actually first created in 1942 by a Danish mathematician, Piet Hine. It was then discovered independantly later on by Nash in 1947. It is another game which has only been solved on small boards. A good beginner's game (written in java) with 7 hex to a side is available here and a better one with more info can be found here. There's also a games site where you can play this against other people, but I'm at work and can't find it now. Sadly, there is seldom anyone else there :-(.

Good article about Go and servers

[Antity]

There was a really good article about Go on kuro5hin maybe three weeks ago. In fact, it caused me to start playing again and it still is much fun. :-)

Just try it. There are lots of free Go servers online. I prefer the KGS server. All you need is to download the client or just play it online in your browser with others (Java required). There are usually ~100 people online in the English room (yes, chat included).

It's a wonderful game.

Re:Good article about Go and servers

[guttentag]

I like this quote from the K5 article:

"The Master said, 'Hard is it to deal with him who will stuff himself with food the whole day without applying his mind to anything good. Are there not gamesters and go players? To be one of these would still be better than doing nothing at all.'"
Confucius (who had better things to do)

Which of the following seems most probable:

• Confucius actually spoke/wrote the English words "Hard is it to deal with him who will stuff himself with food the whole day without applying his mind to anything good."
• Confucius's quote cannot be translated into proper English (i.e. "It is hard to deal with he who will...").
• The error was manufactured to give the quote more authenticity/credibility. People will be more likely to revere a statement they doesn't sound right to them if it is attributed to someone like Confucius ("It doesn't sound right to me, but it must be right because Confucius said it! Golly, he's smart, that Confucius.").
• The person quoting Confucius is a big Star Wars fan and imagines that Yoda really was Confucius in another lifetime.

Re:Good article about Go and servers

[RedWizzard]

The person quoting Confucius is a big Star Wars fan and imagines that Yoda really was Confucius in another lifetime.

Or rather that Confucius was Yoda in another lifetime (a long time ago...).

It will take a general-purpose AI to play go

[Bob Hearn]

I love go (I'm a 2 kyu player), and I'm an AI researcher. But I don't work on go-playing programs. Much as I'd like to, I don't think it would be a productive activity for me.

I think that the minute you start to write a game-playing program, you're trapped by the very natural structures you have to use to make the program even play a legal game. You can't help but start to use minimax search. With go, you add modules for life & death evaluation, influence generation functions, the list goes on and on.

But all these things are just hard-coded approximations of some of the ways people think about go when they play, ripped out of their essential representational context. Real people have rich conceptual networks linking all of these skills together, which multiplies their power enormously. Give a beginning human player a perfect black-box life and death evaluator, like go programs ideally have, and he will never become a strong player. Only by solving life and death problems yourself (to take just one example) can you integrate that kind of knowledge into your total go knowledge. I maintain that this integration is essential.

Will computers ever beat people at go? Sure. But I'll bet the first program to do so will be a general-purpose near-human level AI, that thinks of board positions in terms of physical metaphors. It will have a rich mental landscape.

Bob Hearn

Re:It will take a general-purpose AI to play go

[Kintanon]

The most powerful computer in the world won't be able to run a system like this in a reasonable time frame. And if at some point a computer DOES become as good/better than the masters at 19x19 go, we'll just switch to 21x21 and then you'll need another few orders of magnitude more powerful computers to win. Seriously though, there are already Go AIs that can beat mediocre amatuer players with no handicap, the most popular one is probably GnuGo which I would rate around 4 or 5 kyu. But the processing power required to beat a pro player is massive enough that it will still be several years before it is available. Then, someone with an understanding of the game that rivals a pro will have to program it, because the 'Best' move isn't always the same in the short run and the long run, and when Pro players are reading 200+ moves ahead into the game, you REALLY have to be able to analyze well to stand a chance.
I don't think it's impossible for a computer to beat pro players, but I think it will take 5-10 more years before any computer is powerful enough to do it, and probably 5-10 more years after that before anyone develops an AI that has a complete enough grasp of the game to do so.

Kintanon

Re:It will take a general-purpose AI to play go

[iabervon]

I think that a good AI go player will need to have a rich mental landscape, but that it doesn't need to think at all like humans think. Humans have a certain set of metaphors which are very natural simply because we've evolved the hardware optimized for them, so that we can move through the world reflexively. A computer go player definitely needs metaphors, but they do not need to be physical metaphors; humans play go with physical metaphors because that's what we have, but there may be better metaphors for go which we can't understand or use as well.

I'm fairly certain that a master go AI will have a definite intelligence, but it will be a very alien intelligence. It will have grown up in a world where turning by anything other than a right angle makes no sense, where the granularity of the world is noticeable, where the world is bounded on all sides.

Re:It will take a general-purpose AI to play go

[Bob Hearn]

From the replies, I guess I came across as pessimistic on computer go.

On the contrary, I'm optimistic that go is in fact interesting, that this one simple problem does somehow have something useful to say about the nature of intelligence.

I'm also optimistic about AI. I think that the average PC has enough computing power to play world-class go, given the right program, and these will be forthcoming within the next couple of decades.

My point was that trying to tackle AI through go is dangerous, because it's so easy to be led astray into inappropriate representations and algorithms. I think the general problem will be easier to solve, because the false paths will not look so inviting.

Re:It will take a general-purpose AI to play go

[b1t r0t]

There's one other problem with 21x21. That is the point at which the center starts having more point value than the edges. If you make two squares by filling the third line with black stones and the fourth line with white stones, black wins at 19 and white wins at 21.

Re:It will take a general-purpose AI to play go

[jafac]

My impression with GnuGo - very slow. On my 450 MHz G4. It's just a matter of CPU horsepower? So if I get a Dual 1GHz, it'll be able to beat me in my lifetime?

Re:It will take a general-purpose AI to play go

[Kintanon]

I didn't say GnuGo as the best, only the most popular. I think your analysis of the game is faulty if you don't believe that the complexity of the game increases by several orders of magnitude with the increase to 21x21.

With a 19x19 board you have 361 places you can make your first move, of which approximately 20 are acknowledged as the standard 'best' moves, the second move has 360 possible moves, and depending on your strategy and your strength there are around 100 moves that could be considered good playable moves. The ability to see more than 4 or 5 moves deep becomes pretty difficult for a computer, on my 850 mhz Athlon GnuGo seems to see around 11 moves in, I see around 8 moves in. GnuGo can beat me almost every time. A good friend of mine can trash GnuGo every time, easily, and I can beat my friend about half the time. Professional players see around 200 moves deep into a game, planning on a wide scale.

Increasing the board to 21x21 gives us 441 moves for the first one, and increases the number of accepted Good opening moves to approximately 60, abd with 440 possible moves on the second move, around 200 of them are considered good playable moves. AS the game develops, you'll see that the addition of those 2 rows to the game will lead to an incredibly complex new set of possible board positions. I don't doubt that a 1st or 2nd dan Go Program will come about in our lifetime, but I doubt we'll see one in the reasonable future that can beat a 9dan player. And I don't think conventional computers will ever be able to effectively play on a 21x21 board against a 9dan player.

Kintanon

Re:It will take a general-purpose AI to play go

[Kintanon]

If you load the system with the accepted opens you will be handicapping it against extremely poor and extremely good players. The openings you listed are the ones which are accepted as the most reasonable opening moves because they give the best control of the immediate surrounding territory. But play can change when someone voluntarily opens by sacrificing a corner to build life elsewhere 160 moves into the game. Unless your AI can connect the importance of a move made NOW with that moves importance 150+ moves down dozens of different possible paths then it will be defeated by master level players.

Kintanon

Hikaru no Go

[Spacelem]

I've just started playing Go recently with my flatmates and a friend. It's all because of this amazing anime series "Hikaru no Go" about a boy who meets the spirit of a thousand year old Go master from the Heian period, who teaches and encourages him to start learning the game. From there his own love of the game develops, and he heads towards being a pro.

HNG was sponsored by the Japanese Go society as a way of making Go more popular, and Japanese Go schools are currently being swamped by new players. It's up to episode 38 already, so you'll have some catching up, but the fansubs are great. This link http://www.toriyamaworld.com/hikago/ has some of the original manga if you're interested.

Go and find out more about Go!!!

Re:Hikaru no Go

[Dr. Smoe]

There's actually an article about this on the BBC News web site today, though it doesn't say a whole lot more than the original post. It spends it's time emphasizing how this is causing a revival of Go in Japan.

Dr. Smoe

Re:Hikaru no Go

[MtViewGuy]

In fact, there's an article from MSNBC.com in the Newsweek section specifically about the success of Hikaru no Go.

It has single-handedly turned Japan into a nation of Go devotees, something that has really surprised a lot of Japanese because many in Japan recently thought Go was only played by elderly Japanese. It has caused something you'd never thought would happen in Japan: children are putting down videogames and Pokémon cards and taking up Go in a big way.

The MSNBC article can be read here: http://www.msnbc.com/news/780055.asp

Re:Hikaru no Go

[Kintanon]

WOohoo! This is the reason I started playing Go as well. A friend of mine from Korea showed it to me and I got addicted. I now eagerly await each new fansub. The latest Manga is absolutely suspensful though, driving me crazy waiting for 109!! >:)
The HNG fans from Toriyama's world also have a weekly Go night on Yahoo, but I can't remember which room or what the times are.

Kintanon

Re:Hikaru no Go

[b1t r0t]

children are putting down videogames and Pokémon cards and taking up Go in a big way.

This is a good thing, because the best time to learn to play is as a child. I think this is perhaps because it lets aptitude for the game develop in the language centers of the brain, which are in their primary development years then.

P.S. It was the Hikaru fansubs that got me started too. I had always known I would be interested someday, and then I made a point of learning the rules so I could understand what was going on in the series.

Re:Hikaru no Go

[MtViewGuy]

I think this is going to be good for Japanese culture because to really play Go well, you have to have to be very good at sportsmanship and patience, things Japanese culture revere strongly. It's small wonder why a senior member of the Japan Go Association is a technical advisor for both the manga and anime versions of Hikaru no Go.

It's a good game for people who want to think carefully during play.

Try your hand at go

[MjDascombe]

It's still far more fun (for now) to play other people online - try your hand at http://kgs.kiseido.com/

GnuGo

[Peyna]

Surprised I haven't seen this yet, but there is a GnuGo project. I've played with it a little bit, it's okay in terms of AI, but definitely needs a lot of work. I played a palm adaptation of it, and the scoring was done incorrectly, and if you figured out the quirks of the AI you could beat it everytime.

Re:GnuGo

[nagora]

I played a palm adaptation of it, and the scoring was done incorrectly,

There are two major scoring systems for Go, are you sure it wasn't using one you didn't know?

TWW

Re:GnuGo

[Peyna]

Yeah, it was real messed up, plus you had to basically score it on your own. Got linkage to the two?

Re:GnuGo

[b1t r0t]

I played a palm adaptation of it, and the scoring was done incorrectly

What was it called? Was it PilotGo, by any chance? When I first started learning, that's the first program I tried. Within a week I realized how pathetic it was. I've heard it was based on GnuGo 1.2 or so. GnuGo has significantly improved its play in the 3.x versions.

A slightly less pathetic program is the one included with the Zaurus PDA. Just throw sacrifice stones in its area, and it will waste its time capturing them while you build up your position somewhere else.

Yes, go is harder

[jo-do-cus]

In fact, there are quite a large number of reasons why Go is harder for computers than chess.

First, there is the board size and the fact that you can play (almost) anywhere on the board, which accounts for the large branching factor (number of possible moves in each position) for the search tree.

Next, there is the fact that games take more moves to finish (about 300 ply, i think, for about 80 for a chess game), which makes the search tree even more staggeringly big. Many many millions of times bigger than that of chess, even when you do a shallow search.

Then there is the difficulty of deciding when the game is over. In go, this happens when both players pass, so this means you have to know when there are no sensible moves anymore. This turns out to be a major problem, whereas in chess the end of a game is more clearly defined.

In fact, it is even very difficult to determine the score for a game when both players have passed. Especially in human expert games, end positions require a great amount of understanding of the game to determine the score.

These, and many other reasons, make Go a very difficult game for a computer. Many (brute force) search/evaluation methods we use in chess and checkers are really not up to the task of playing Go. So we try and figure out some more 'intelligent' methods...

BTW, I have not read the NYT article, but i really doubt they can say anything sensible about 'intuition'. We don't know what intuition is, and even if we would, I think the strenghts of computers lie elsewhere. Let people do what they are good at (intuition, fuzzyness), let computers do what they are good at (count really really fast)...

Re:Yes, go is harder

[Latent IT]

Well, that's kind of the whole thing. If the game is only going to have 80 moves, looking 12 ahead will mean you're looking at a large, meaningful part of the game. If a game is 300 moves, looking at the next twelve moves ahead is a lot less effective. In fact, it probably scales exponentially, but what the heck. Math ain't my thing.

Short version - it's about percentage. ;p

Chess is not dead yet

[Anonymous God]

I was at Game 2 of the 6 game Deep Blue-Kasparov series and felt that perhaps Kasparov was making moves to throw off DB. Maybe this cost him the game eventually.

It was a great day for AI but in retrospect a sad day for chess. However it's going to a long time before Chess engines play chess with attitude, emotion and an individual style. If Chess engines just used minmax, alphaBeta pruning, the quality of play would not be very high. Storing opening lines and end game rules makes it a much tougher opponent.

However when these programs are able to store every combination of every possible game and then based the outcomes move up the tree to decide what move to make Then chess will truly be dead.

That will take more than a few EMC boxes! - More like (~35 options per move (starting with 20), ~100 moves per game) = 35^100 = ~2.55e+154 positions. Roughly assuming each position uses 64 bits = 1.85e+143 TB (> a googol Tera bytes!) When you have this 'database' 'populated' you can tell what moves ensure sucess by looking at the end node outcomes.

But for now Go play chess !

Automated theorem proving

[ortholattice]

Another area that has been elusive to computers is automated proofs of theorems in mathematics. Automated theorem provers such as Otter and Isabelle typically cannot prove "deep" theorems that human mathematicians prove on a routine basis.

Only occasionally does a computer prove a theorem previously unsolved by humans, such as Robbins algebras are Boolean, but these tend to be problems (like this one) involving simple algebraic manipulations. Something like Fermat's Last Theorem, forget it; Wiles' proof has not even been verified by computer, much less automatically proved. The correctness of Wiles proof is at this point based on a consensus of human mathematicians, who may or may not (hopefully not) have overlooked some subtle flaw in its incredibly deep proof.

BTW don't confuse theorem provers with symbolic algebra systems such as Mathematica, Maple, or the GPL'ed Maxima. While indispensable for complicated calculus problems etc. beyond what a human can practically do, AFAIK they cannot prove even a simple abstract result such as the irrationality of the square root of 2.

Nash and hex

[gatekeep]

Actually, hex was invented in 1942 by Piet Hien. Nash independently invented virtually the same game in 1948, but he was a little late. Kind of like Thomas Edison and what's his name's race with the telephone. Funny how in one case the second inventor is virtually unknown while in the other, it's Nash who gets the credit.

Look here for documentation of this.

Re:Nash and hex

[TheShadow]

Umm... I think most people know that the telephone was invented by Alexander Graham Bell. Hence Ma Bell and all the Baby Bells.

Re:Nash and hex

[gatekeep]

D'oh! I knew that, just had a lapse in thinking.. haha.. Ohh well, my point is still valid even if the facts are obviously messed up.

I guess I'll probably get modded down on that one.. *shrug*

Why Go Programming is Difficult

[mechner]

A more in-depth article on go programming, from the point of view of a programmer and a player, originally published in The Sciences: http://mechner.com/david/compgo/ Click on "All Systems Go".

Chess Will Never Die

[Transient0]

Estimated number of Atoms in the Universe: 1.00E+81

Soo.... If you show me:

1. A way to make a single atom store 10 to the power of 62 terabytes of information

2. A government-stamped letter of permission to turn the known universe into a chess computer

then I'll admit that chess is dead.

Note 1: Deep Blue is no longer the most powerful chess computer, that honour has passed to Deep Fritz which is capable of running on an i86 architecture(unlike th proprietary machine that ran deep blue). Also, Deep Fritz was not designed to beat anyone in particular, yet it has succeeded at defeating both Deep Blue and Kramnik(the current world chess champion).

Note 2: I know that my 10E-81 figure does not include free subatomic particles(photons, free electrons, mesons, etc.), but you get the idea.

Re:Chess Will Never Die

[Reality Master 101]

and Kramnik(the current world chess champion).

Unless you know something I don't, the match between Kramnik and Fritz isn't until this October.

Other information about Go - Links-a-plenty

[gatekeep]

For those of you interested in learning more about Go, here's some links to resources I've found helpful since starting to play 3 weeks ago.

k5 had an article about go which is what initially piqued my interest and got me started in the game.

Kiseido Go Server is my favorite place to play online, and very newbie friendly.

Some great introductions are available from Kiseido The Interactive Way to Go and Tel's Go Notes

Uligo and Goproblems.com are great places for learning how to play in common situations.

If you prefer a phyiscal board and stones check out Samarkand and Kiseido

Also, anyone in the Chicago area should check out the Evanston Go Club

A word of caution, if you decide to learn go, expect to lose most of your first 50-100 games. It's a long road, but once you start making progress, you'll grow quickly. I know I sure have. Anyone who's up for a game look for 'jjarmoc' on KGS.

Re:Versions of Go

[Anonymous Coward]

Nonsense - there are NO major variations of the game of GO (baduk, weiqi). There ARE other games that use a similar board (grid) and stones - these are NOT Go. Othello is NOT Go. These other games are NOT Go.

One of them is Go Moku - a children's game mostly - played on the same board, otherwise known as "five in a row."

There are minor rule variations in modern Go, but these do nothing to change the basis and play of the game except to a very, very minor extent.

Perhaps these other games are interesting, but don't confuse them by putting them all in the same category. They're not.

I wrote the first commercial Go program

[MarkWatson]

I am fairly sure that I wrote the first commercial Go playing program, Honninbo Warrior for the Apple II.

It really did not play a very good game, but it was fairly well reviewed in the Apple II magazines because at least it could play (and perhaps because of the money I spent on advertising in those magazines!)

Anyway, I agree that Go is a great platform for AI research (probably only real time robot soccer is better in my opinion).

-Mark

This isn't so much about AI

[anthony_dipierro]

All are very strong Go players, and it takes a strong Go player to write even a weak Go program.

I think that line captures the problem I have with the article. The purpose of AI is to produce a computer where the programmer doesn't even have to know that the game of Go exists. Yes, an intelligent computer can play Go half-decently, but a computer which can play Go half-decently isn't necessarily showing any intelligence.

So yes, writing a good Go program is challenging, but I wouldn't exactly call it research, unless you're using a method completely different from all the successful ones out there (basically a pruned tree search).

Go on the Palm Pilot

[Mr_Silver]

I fancy learning Go and I've got a Palm Vx. Does anyone know of any Palm Pilot versions of Go that have some sort of AI and configurable board sizes (since I'm told beginners should start with 9x9 and work upwards).

I looked on PalmGear but they only have apps that give you a Go board which can record and replay games. I need someone to play against!

The AI doesn't have to be that great as I'm only a beginner.

Any ideas?

Re:Go on the Palm Pilot

[crisco]

GnuGo is available for the Palm. It is a relatively weak AI, even I learned to beat it as a beginner.

AIGO is a shareware Go game for the Palm. It has a stronger AI.

telephone invention

[Transient0]

this is off-topic, but i need to say it.

thomas edison did not invent the telphone.

he had nothing to do with it,

It was Alexander Graham Bell, a Scotsman living in Canada, who invented the telephone.

There WAS a race for the patent, because at least four different researchers were indpendently working on the same project at the same time.

Nice to see Go getting some press

[BlackMesaResearchFac]

I've been a big Go fan for several years now, ever since I started playing it online in the early 90's on The Sierra Network (TSN) aka ImagiNation Network (INN).

I had not even heard of it before but the mind power involved in playing drew me to it. Anyone elses brain hurt the first few times they played? Another example of an interest of mine that I didn't have in common with anyone else I knew.

I bought "Many Faces of Go" by Ishi Press, which played pretty well for someone like myself (14kyu) who only played maybe 20 games over a year.

The nice thing about a game like that for beginners is you can make a mistake (bad move) and realize it right away and go back a few moves and try a different move, in order to see what would have been best. Being able to save and go back is a very good learning tool.

I would agree with most posters here in that once you're to a certain level it's best to play against real people anyway. Not only because they're better, but they're also much more unpredictable. Many Faces of Go was somewhat unpredictable, but nothing can beat a human opponent when it comes to that area.

Moreover, being able to watch players a few notches above your playing level is IMO more useful to improving your game than even playing yourself.

Standing on the shoulders of giants

[dsr9996]

Hello Bob,

I respect what you are saying here and understand your reasons for not working on a go-playing program yourself, but I would challenge you with this: Even though you will probably not be the person to write the go ai program that is "near-human level", the person who does eventually write it (X number of years/decades in the future) will most definitely only be able to do it because he learned from people who came before him and attempted the endeavor. In short, it takes Newton to formulate the basic laws for physics and the calculus before Einstein can go further and discover relativity and quantum physics. And as Newton said, the only reason he could accomplish what he did was because he "stood on the shoulders of giants" that had come before him.

I am sure that this is not a new idea to you, but I present it again because I think it is very valid. We are at a very primitive state when it comes to computer ai, as anyone who has done any ai knows, both because of our lack of understanding of how our own intelligence/consciousness works, and because of our lack of good programming tools that allow us to work at a high abstraction of thought (i.e. most of the code we write is very tedious, and even though it is necesssary for our ai programs, it has little to do with actual ai). It is similar to knowing that you need a modern race-car when oil refineries, engines, and smelting have not even been invented yet. It is up to us to create those go-carts, pardon the pun, and start exploring how we might create a smelter, looking forward to the day when the infrastructure will be in place for others to continue the progress.

I know what you mean when you say that when you begin work on an ai problem like go, you are immediately trapped into things you have been taught, common procedures that you know others have used for similar types of problems, etc. However, this does not mean that you cannot be the one to think up the next innovation with respect to ai, taking the next step in creating a "rich mental landscape" that will lead to the integration you believe is essential to true ai.

I am quite positive that you are more knowledgeable about ai than me, since I have only dabbled in it here and there, but I hope you take my encouragement in the spirit I have intended to give it.

Peace to you,
Devin

Download an intro to Go program for Windows

[big_debacle]

Here's a neat little (9x9) free version of The Many Faces of Go. Download Go

It provides a nice taste of go for noobs.

My only thought is that if computers are no good at Go, then I must really stink since I can't even beat the free intro version most of the time.

Britain Go Association and the BBC

[hether]

Slightly off topic, but Britain has a Go Association. I imagine they are pretty pleased to see the game coming back in style. They have lots of info about the game, downloads, etc. available.

The BBC had a story in their CBBC kids section about Go just yesterday or the day before.
http://news.bbc.co.uk/cbbcnews/hi/world/newsid_216 3000/2163584.stm

I'd never heard of the game before I saw this story. I've obvsiously never played it, but they way they described it, it seemed like games could take a very long time when you use the larger board. I could see how this could be very complicated.

IGS is also good

[nyri]

I prefer IGS server. Client is way worser than with KGS, but it easier to get to play in IGS. Also pros hang out in the IGS.

-- nyri

US has one too

[pls2917]

American Go Association

It wasn't even Bell

[crisco]

It was an Italian named Antonio Meucci. He first demonstrated a system in 1860 but couldn't raise the money to file a patent. Supposedly Bell even shared a lab with him.

First referece I found was this. You can google from there.

Re:It wasn't even Bell

[rmohr02]

I know Bell beat someone else to the Patent office by half an hour, perhaps that was the guy he beat.

Massachuetts Go Association

[limekiller4]

If any Slashdotters are interested in Go and reside in the Boston area, there is a 24-hour Go club (only two exist in the US) in Somerville (Davis Sq. stop off the Red Line T, about 5km or so away from Harvard) is the Massachuetts Go Association. While I'm not a member any more (moved to Rhode Island), they're a great club and very much interested in helping new players become familiar with the game. Their book selection in the library is quite impressive as well, covering all the obvious topics (fuseki, joseki) as well as strategy, tactics, theory, etc.

Generally, a newcomer would start off on board with a 9x9 grid and progress from there to a 13x13 and then a 19x19. The great thing about go is that since each piece (stone) is precisely the same as any other stone, it is quite straightforward to handicap a strong player so that an equitable game can be found vs. almost anbody simply by giving the weaker player position on the board before the game begins.

It's a fantastic game, and without being too bombastic (I hope), it can teach the player quite a number of things entirely outside the scope of the game itself.

Topic Drift Funnies

[crisco]

While searching for the above, I came across this grade school report on Bell.

To be a kid again...

Re:Octi has similar properties.

[b1t r0t]

Very interesting. It reminds me of the old 3M Bookshelf game Ploy, except you add the movement directions to the pieces as play progresses.

Data does a Microsoft

[Tablizer]

(* And for all you Trek fans out there - remember the great Moriarty episodes? "Computer, design a foe good enough to challange Data" *)

Another interesting one is where a grandmaster of a certain game (don't remember name, had air-finger connectors) came on board and was giddy and pompous about beating the famous Data.

Data eventually realized that he could never beat the grandmaster, at least not by "winning" the game. Instead he learned ways in which to increasingly postpone his loss. Eventually the grandmaster would physically tire, which androids allegedly don't do, and thus win by attrition.

The frustrated grandmaster stormed out of the room in a Kasperov-like fit, and Data was cheered by the crew for rattling his chain.

(In reality, the rules would probably eventually be altered such that the discovered attrition pattern does not work any more. Or, some way to get a score after a set period of time.)

Anyhow, it was one of my favorite episodes.

Re:Data does a Microsoft

[LoudMusic]

Oh yes, one of my favorites as well. How could I have forgotten this episode?

I've used the 'win by attrition' method a number of times in StarCraft, Red Alert, and the like. When you know you're beaten, curl up in your shell and don't die.

~LoudMusic

Star (related game)

[jbennetto]

Star is a game related to Hex but much more intricate. It was created by Craige Schensted (a physicist) and popularized in Games Magazine back in '83.

Star is similar to Go and Hex in the simplicity and consistancy of the rules. The play and tactics are those of Hex (connect stones on a hexagonal board), but there's an additional strategic element involving groups of connected stones that touch the edges at as many points as possible.

a fun, difficult game of my own invention

[option8]

while everyone is mentioning go-like and chess-like games of recent invention, i thought i'd throw a plug out there of a game i came up with (though it's similar to several) a couple years ago, called the Aggravator.

of course, i failed to check my spelling at the time, so the game is acually called "aggrivator" which only adds to my aggravation.

anyhoo, it's a fun little game on a 5x5 board. you can play it online (requires shockwave plugin) here: http://www.niftee-tron.com/shock/aggrivator.html.

i've managed to beat it - once. the rules are fairly simple, but the goal is to get all the squares on the board to look the same, which is fiendishly difficult.

the sequel, aggrivator 2.0 is also online, http://www.niftee-tron.com/shock/aggrivator2.html. i've never beaten it, and i don't know anyone who has. it has several levels, each with their own "goal". i tend to think that some of them are impossible to attain, but a math major friend of mine tells me that, using the progression rules in the game, it's possible to attain any board state, but then, he's just a grad student, so what can he know?

anyhoo, maybe some of the distributed might of the slashdot crowd can solve A2.

Re:a fun, difficult game of my own invention

[option8]

just as i thought. it seems a little countuerintuitive, but i won't argue with you.

i'll have to pass this on to my mathematician friend, who came to the opposite conclusion. of course, he had a convoluted way of coming to his conclusion. i'll see if i can find the email he sent me...

he and i were never able to come up with a "pattern" to beat A1 (and never really bothered with A2 :), despite our best efforts, and we each could only stumble on a solution after some frenzied clicking and much "aggrivation"

as for contacting me, my email is plastered all over the site with the aggrivator on it. or, you could put my slashdot id together with the URL and have a pretty good idea...

IGS rocks a**

[rodentia]

Competition is far superior on IGS. The ratings on IGS were recently adjusted by four stones and are still a stone or two stronger than AGA ratings. IGS is good for your game; the others are for kicking the weak around the board.

Re:AI book recommendations anyone?

[kallisti]

Game AI at this point is still finite state machines and A* pathfinding for the most part. Pre-canned scripts, such as placing move-to points on the map are also used heavily. Every once in a while, someone writes something fancier like Genetic Algorithms or Neural Nets, but those generally don't do as well. Examples include Black & White and Galapagos.

Check out The Game AI page for lots of good sources of information. The series "Game Programming Gems" have some really good articles, they recently spun off an AI-specific "AI Programming Gems" which I haven't seen yet.

How to build a Go program

[anonymous_wombat]

I have been considering starting an open-source Go project in a few months (after I get laid off and have some time). Here are some of my thoughts.

Writing a Go program may not be exceptionally difficult. It is just a very large problem. Most previous attempts were done by one or two people, in an attempt to have a commercial success. A large team of people would be required. Probably, this could only be done within an open-source framework.

There might be some things that can be learned from Chess programs, for example how to solve Life and Death problems, but a completely different approach will be needed.

I believe that people have attempted to solve the wrong problem. The problem is not

How can I find the next move on an arbitrary board with n stones?

but,

How can I find the next move from a board with n stones which I have already analyzed?

If you are a Go player, you know that in the middle game, you can usually come up with a decent move within a few seconds, but if you walk over to a game being played by other people in the middle game, it would probably take you several minutes of intense thought to analyze the board, figure out what is alive and what is dead, and who is ahead. Therefore, the program needs to save its state between moves. Figuring out the next move should then be MUCH easier.

Most open-source programs are designed by one or a few people, and then outsiders either just fix bugs, or add additional functionality within the current design. For a Go program to have a decent chance of achieving dan level play, not only does it have to be open-source, but open-design.

Assume that a framework exists, and that modules which perform discrete tasks can be plugged into this framework. A standard interface is defined for each module. There would be modules for fuseki, joseki, life and death, etc. The overall framework which would perform multi-threading, saving of state, communication between modules and the client could be designed and coded by one person. But each of the Go specific modules could be designed and coded by a different person. Modules such as Life and Death are probably straightforward, but what about a module that determines good and bad shape? Many different people would have different ideas of how to do this. The framework should allow multiple competing shape modules to be written. They should be interchangeable, so that any of them can easily be plugged into the framework. In this way, many people can take part in the design of the system. There would be no political fights over how a shape module should be implemented. Different people can write their own, and the one that performs the best within the framework would be used.

Using this set of ideas, I believe that a small subset of the many people who both play Go and are software developers could develop software far better than anything now existing.

I intend to have such a framework implemented by the spring of 2003. I have not posted anything online yet. If you might be interested in participating, feel free to email me at dweiss51@yahoo.com.

Re:How to build a Go program

[anonymous_wombat]

I have used GnuGo, and examined the source code. Although it does not play a bad game, it is certainly not an easy program to modify. It would not be an easy task for a casual developer to come along and add additional functionality. It is certainly not designed to allow the same functionality to be coded in different ways, or to allow different building blocks to be assembled differently. My main reason for rejecting GnuGo is that, unlike most other problem domains, people don't really know how to build a good Go program. The ability to try different techniques and experiment will be vital. I do not believe that GnuGo provides this, but I do not mean to disparage that project in any way, because they have done extremely high quality work given the limits of coding in C.

Minute fraction? Sure...

[freeBill]

...if you define "minute fraction" as "less than 99 percent, but greater 0 percent."

The truth is we won't know whether it was easier to solve or harder until either:

• somebody solves it and writes a program which beats the best human Go players
• we have spent the same amount of (and quality of) effort trying to play Go that we did to get to Deep Blue and failed to get close to a solution

It has been my observation that this kind of article is written about each hard game as programmers try to get a grip on the strategy. I can remember 25 years ago when similar things were said about chess: too many combinations for brute force; too much of great play was intuitive; chess players cannot describe just what they do.

Ten years ago we heard similar things about bridge. Now, it's Go. As in most journeys, we won't know what it's like until we get there.

As far as Reading The Fine Article, the poster may well have read the article, WHICH DID NOT DEAL WITH HIS POINT to any significant degree. Chess proved to be a very hard problem, which required decades of concentrated effort by some of the best programmers in the world. It was a classic problem which attracted huge amounts of effort. While Bridge and Go afficiados may love to think of their games as more difficult that Chess, neither has had a fraction of the theoretical effort lavished on it that Chess had.

The article was a very superficial treatment. Not only did it fail to deal adequately with the question of the relative amount of effort put into Chess and Go, but many of the facts it presented are either wrong or misleading.

Take the combinatorial questions, for instance: At the beginning of a game of Go, there are 19^2(=319) possible legal moves. If a programmer uses brute force, the combinations go up very quickly (319 x 318 x 317 ...). However, a Go master only considers four of those legal moves for his first stone.

At the beginning of a chess game, there are 20 possible moves, but a good chess player must have a thoroughgoing understanding of at least six of them, with another six being in the realm of possibility. So, over half of the moves must be considered in any brute-force algorithm. Most chess-playing programs do not consider all branches of all trees to the same depth. The ratio of likely-to-be-considered-moves-to-legal-moves is much higher in Chess than Go. Indeed, in Go there are many situations when only one possibility need be considered as the consequences of failure to respond to a particular attack can be so catastrophic.

Another difference between Go and Chess is the degree to which moves open up new possibilities. Most moves in Chess create the possibility of new moves on the following turn which were not legal on the present turn. Indeed, one could argue that one of the key strategies in Chess is opening up as many possible moves as possible. Most moves in Go reduce the number of legal moves (for both players).

Go players are much more likely than Chess players to follow out long sequences of moves without considering alternatives at each step along the way. Chess players find they must consider branches as they think ahead in most such situations unless responses are forced.

I find it interesting that Danny Hillis thinks Go is more intuitive than Chess. But we may be talking about an intuition about intuition. Hillis' intuitions about how computers can be taught intuitive processes are probably worth considering. Other than that, the article offered very little accurate information about programming Go.

Download a Go program for OS X

[b1t r0t]

The program Goban is a Go client for OS X with GnuGo 3.2 packaged into it. It now also supports two internet Go servers, IGS and NNGS, so you can play against more than just the computer.