Ugh, articles written by people who don't know about either game, for people who don't know about either game.
Anyway, I wiki-ed this "professional" and "champion" it beat, Fan Hui. He is only 2 dan. This is like a chess master. Not at all comparable to Kasparov and Deep Blue.
But it is interesting, thanks for the link.
It mentions how there are more legal moves per turn in Go when compared to Chess. Surely this means that Chess is no problem for computers at this point?
Thoughts?
Well, not entirely.
It is true that there are many many many more positions in Go than in chess, but both games have so many possibilities that a hundred thousand computers running 200k ply each second would never come close to calculating all the different moves, even if you let them run for millions of years. They both offer practically (but not literally!) infinite possibilities.
I'll try find the link that explained it better, but one important difference is that it's much easier in chess to assign numerical values to different aspects of the position.
Still very interesting :)
Alpha Go may not beat the World's strongest human player yet... but it is "learning".
In any case, the true test of "computer intelligence" would be to decode the Voynich manuscript. 
go is not more complex than chess. It may have more legal moves, but the difference between them is also less relevant. This simply makes it less suitable for brute for calculations that computers(and some humans) are good at...and it explains why high level go players really arent calculating much, but it doesnt make it fundamentally more complex.
Your ignorance continues to reveal it has few boundaries.
In addition, Go has not been mastered by computers. Only real advancement that a computer can now beat the best Go players.
Go has not been solved.
What did FoS say which is incorrect?
It mentions how there are more legal moves per turn in Go when compared to Chess. Surely this means that Chess is no problem for computers at this point?
Thoughts?
Well, not entirely.
It is true that there are many many many more positions in Go than in chess, but both games have so many possibilities that a hundred thousand computers running 200k ply each second would never come close to calculating all the different moves, even if you let them run for millions of years. They both offer practically (but not literally!) infinite possibilities.
I'll try find the link that explained it better, but one important difference is that it's much easier in chess to assign numerical values to different aspects of the position.
Still very interesting :)
Chess is complex but not as much as being infinite.
The fact is, while in any given position there are like 20 moves available, the moves that make some sense are much less. Same goes for the positions.
For instance, if you have the Lucena position (rook ending) you can have it in a ton of different combinatinations depending on where the two rooks are; but the key aspect of the position is the same (own king in front of passed pawn at the second to last row).
Another example: you can just teach a computer that R+K vs K is a win. So when it calculates a previous position that liquidates to it, it won't have to calculate to mate but it will just stop at knowing that it's a forced win - just like human players do.
As far as things like symmetries and rotations (identical positions, just different orientations) I'm sure go has many many more than chess.
Go is more complex than chess, that's a given. Even the best go players makes mistakes every game, and you can slip several times and come back in go, but you cannot do that in chess because chess players' play are much closer to optimal play than go players' play. Why, because go is much harder to master. You can have multiple battles going on at the same time, where as in chess you typically have one or two battles going on at any given moment.
Alpha go is going to play one of the top players in the world, I would not say he is the best right now. Lee was the best player in the past, but not at the moment.
Go is more complex than chess, that's a given. Even the best go players makes mistakes every game, and you can slip several times and come back in go, but you cannot do that in chess because chess players' play are much closer to optimal play than go players' play. Why, because go is much harder to master. You can have multiple battles going on at the same time, where as in chess you typically have one or two battles going on at any given moment.
Alpha go is going to play one of the top players in the world, I would not say he is the best right now. Lee was the best player in the past, but not at the moment.
Hmm, even the best chess players make mistakes in all their games (ok, not if it's a short draw or something). Could you be more specific for how chess mistakes and go mistakes are different?
Go is more complex than chess, that's a given. Even the best go players makes mistakes every game, and you can slip several times and come back in go, but you cannot do that in chess because chess players' play are much closer to optimal play than go players' play. Why, because go is much harder to master. You can have multiple battles going on at the same time, where as in chess you typically have one or two battles going on at any given moment.
Alpha go is going to play one of the top players in the world, I would not say he is the best right now. Lee was the best player in the past, but not at the moment.
Hmm, even the best chess players make mistakes in all their games (ok, not if it's a short draw or something). Could you be more specific what you mean?
What i mean is you could play and lose a battle or two early and still come back and win the game, in fact this happens very often, in go. This would be like chess player come back after dropping several ps or even a piece without compensation in chess, which is very hard at the GM level. The reason is the board is so big 19 by 19, it is hard to know what is the optimal move, in fact there are no such thing as optimal moves most of the time during the middle games in go.
Hmm, but blundering a pawn and blundering a small group of stones doesn't seem like a good comparison. It seems more comparable to choosing a sub-optimal piece placement or move order resulting in losing tempo. It makes your position worse, but you're not losing.
But maybe you're telling me small mistakes like this are more common in go, because the players can't predict what will happen as easily?
Hmm, but blundering a pawn and blundering a small group of stones doesn't seem like a good comparison. It seems more comparable to choosing a sub-optimal piece placement or move order resulting in losing tempo. It makes your position worse, but you're not losing.
But maybe you're telling me small mistakes like this are more common in go, because the players can't predict what will happen as easily?
Yes, the chance of turning a good position into a losing one in go happens so much more than in chess that it is not even comparable.
I guess that makes sense. The games are much longer too. Longer games on a bigger board should mean more chance to error.
I suppose I'll live long enough that computers can beat all humans at go. It will be interesting to see if professional go players will use engines to analyze like chess players.
In chess you can explore a line for a few moves to get a sense for what the engine is aiming for. In go I imagine seeing the engine's point would be much harder.
That article is riddled with errors, as usual for lay people trying to explain anything they don't actually understand. Just like any mainstream article you ever read about string theory or the like.
"Everything is relative, man...that's Einstein".
Google is just pushing this for publicity. Go is not "solved", and chess is nowhere near being solved, either.
I guess that makes sense. The games are much longer too. Longer games on a bigger board should mean more chance to error.
I suppose I'll live long enough that computers can beat all humans at go. It will be interesting to see if professional go players will use engines to analyze like chess players.
In chess you can explore a line for a few moves to get a sense for what the engine is aiming for. In go I imagine seeing the engine's point would be much harder.
Think of it this way, the go computer just cracked 2300 or maybe 2400, like chess computers did in the 90s. It won't take them long now to catch up, the hard part is getting to the fm im level in chess, after that, it became much easier. I don't think Alpha go will beat Lee, one of the top go players in the world in March, but you never know, supposedly this machines plays over a million games each day and is learning from those games, not sure when it will reach its peak, but google is known for super fast searches, so maybe it will win a game or two against Lee.
I guess that makes sense. The games are much longer too. Longer games on a bigger board should mean more chance to error.
I suppose I'll live long enough that computers can beat all humans at go. It will be interesting to see if professional go players will use engines to analyze like chess players.
In chess you can explore a line for a few moves to get a sense for what the engine is aiming for. In go I imagine seeing the engine's point would be much harder.
Think of it this way, the go computer just cracked 2300 or maybe 2400, like chess computers did in the 90s. It won't take them long now to catch up, the hard part is getting to the fm im level in chess, after that, it became much easier. I don't think Alpha go will beat Lee, one of the top go players in the world in March, but you never know, supposed this machines plays over a million games each day and is learning from those games, not sure when it will reach its peak, but google is know for super fast searches, so maybe it will win a game or two against the human soon.
For all practical purpose, chess is "solved" for humans, no human will ever beat the strongest chess computer again unless human evolve again or become cyborgs.
Think of it this way, the go computer just cracked 2300 or maybe 2400, like chess computers did in the 90s. It won't take them long now to catch up, the hard part is getting to the fm im level in chess, after that, it became much easier. I don't think Alpha go will beat Lee, one of the top go players in the world in March, but you never know, supposed this machines plays over a million games each day and is learning from those games, not sure when it will reach its peak, but google is know for super fast searches, so maybe it will win a game or two against the human soon.
Interesting.
I remember reading a few years ago some professional go players saying a computer will never beat a professional player, and I had to laugh. I remember some (very old obviously) quotes from Kaspaorv and other GMs who said the same thing. Meanwhile the computer scientists always said it's only a matter of time.
For all practical purpose, chess is "solved" for humans, no human will ever beat the strongest chess computer again unless human evolve again or become cyborgs.
That's not the definition of a game being solved:
https://en.wikipedia.org/wiki/Solved_game
For all practical purpose, chess is "solved" for humans, no human will ever beat the strongest chess computer again unless human evolve again or become cyborgs.
That's not the definition of a game being solved:
I said for practical purpose for humans, of course we do not know what is the best move each turn.
Think of it this way, the go computer just cracked 2300 or maybe 2400, like chess computers did in the 90s. It won't take them long now to catch up, the hard part is getting to the fm im level in chess, after that, it became much easier. I don't think Alpha go will beat Lee, one of the top go players in the world in March, but you never know, supposed this machines plays over a million games each day and is learning from those games, not sure when it will reach its peak, but google is know for super fast searches, so maybe it will win a game or two against the human soon.
Interesting.
I remember reading a few years ago some professional go players saying a computer will never beat a professional player, and I had to laugh. I remember some (very old obviously) quotes from Kaspaorv and other GMs who said the same thing. Meanwhile the computer scientists always said it's only a matter of time.
That's exactly the point, computer beat a low level pro go player 5-0, and it totally destroyed all the previous go computers, winning at 80-90% rate after giving away four stones, that like giving a rook and winning at 80-90%. Alpha go is the first truly good go playing engine.
This article pretty much says it all:
http://www.theguardian.com/technology/2016/jan/27/google-hits-ai-milestone-as-computer-beats-go-grandmaster
It mentions how there are more legal moves per turn in Go when compared to Chess. Surely this means that Chess is no problem for computers at this point?
Thoughts?