That doesn't show a prediction algorithm can't exists, it shows either one of the following things
-the results of the algorithm can't be accessed or interpreted in this world
-the algorithm can't be limited to only yes-no answers
-The algorithm only exists on a meta level.
-There aren't enough particles in the world to implement a computer to compute the algorithm
That doesn't show a prediction algorithm doesn't exists, it shows either one of the following things
-the results of the algorithm can't be accessed or interpreted in this world
-the algorithm can't be limited to only yes-no answers
-The algorithm only exists on a meta level.
-There aren't enough particles in the world to implement a computer to compute the algorithm
The Website: "When Regulation Is Warranted, John McCain Acts. John McCain does not believe in prescriptive regulation like "net-neutrality," but rather he believes that an open marketplace with a variety of consumer choices is the best deterrent against unfair practices."
Wow that's a point I totally support. Now where can I find John McCain's plan to create such a marketplace?
Why does everybody allways acts like they are being cheated out of their money when caught breaking traffic laws? They are laws, you know them and they improve safety.
It's an interesting idea but I think you're better of spending the computer power on more simulations for the monte carlo algorithm. Those are perfectly suited for parallel computing anyway.
I think monte carlo by itself is enough for the go endgame but I think the best idea is to teach the AI some combinatorial game theory. Making a database of moves with their CGT values would be very interesting.
The method the GP described is called stone counting. While it will give slightly different scores compared to normal counting methods it is in fact equivalent to the ancient Chinese way of counting.
Go is very different from Othello. In Go your goal is to surround empty points on the board. Because your stones stay on the board unless captured every previous move counts. It's hard to actually compare it to other games
here is an interactive introduction to go. Towards the end there are a few simple game examples explained.
The 8-dan was a professional 8-dan who are stronger then amateur 8-dan players. TFA mentioned that that Myungwan Kim estimated MoGo at about as strong as an amateur 2-dan with some moves being 5-dan moves.
I do know that writing a go AI isn't easy, and brute force is definitely not the way to go. But what does that have to do with go being mathematical or not? Go has many beautiful mathematical properties.
What I posted first is just a simplification of course. Here is the algorithm in more detail.
From a given position it will play thousands if not millions of games to determine which move gets the best winning percentage. This is how these games are played
When you are at a node in the game tree and want to decide which move you will make next you first calculate a score for each possible move.
Unexplored moves get a score based on a heuristic.
moves that we have tried before get a score based on the number of times we have tried it and the winning percentage. The higher the winning percentage the higher the score, but the higher the number of times we have tried it the lower the score.
This results in searching down the path with the best winning percentage most of the time. But there is still searching going on in the less good paths.
See here and here for more details.
In chess you also have the advantage of being able to make an endgame database for your program. Chess positions only get simpler as the game progresses because pieces are removed.
In Go this doesn't happen. An endgame database is simply impossible.
MoGo uses a Monte Carlo approach. Basically it keeps playing random games to see which give the best results. Faster computers mean more random games can be played thus given better results.
A GeV is an unit of energy and thus mass. But it's so little that expressing the mass in grams would lead to very small numbers. See wikipedia for more information.
You have a point but it's not exactly acurate. Renting a book from the library isn't copyright infringment. Of course you COULD copy the book you borrowed but by downloading something, you are allready infringing on copyright.
Slashdot Mirror
That doesn't show a prediction algorithm can't exists, it shows either one of the following things
-the results of the algorithm can't be accessed or interpreted in this world
-the algorithm can't be limited to only yes-no answers
-The algorithm only exists on a meta level.
-There aren't enough particles in the world to implement a computer to compute the algorithm
Arrg. I accidentally replied to the wrong topic.
That doesn't show a prediction algorithm doesn't exists, it shows either one of the following things
-the results of the algorithm can't be accessed or interpreted in this world
-the algorithm can't be limited to only yes-no answers
-The algorithm only exists on a meta level.
-There aren't enough particles in the world to implement a computer to compute the algorithm
The article was from 2006. Here's a link to wikipedia for some details.
The Website: "When Regulation Is Warranted, John McCain Acts. John McCain does not believe in prescriptive regulation like "net-neutrality," but rather he believes that an open marketplace with a variety of consumer choices is the best deterrent against unfair practices."
Wow that's a point I totally support. Now where can I find John McCain's plan to create such a marketplace?
Why does everybody allways acts like they are being cheated out of their money when caught breaking traffic laws? They are laws, you know them and they improve safety.
I'm perfectly capable of making spelling mistakes while sober you insensitive clod!
Now what is the actual threath? Shrinky dink or easily duplicated keys?
If this gets overturned it'll probably be written into law in a few months.
It's an interesting idea but I think you're better of spending the computer power on more simulations for the monte carlo algorithm. Those are perfectly suited for parallel computing anyway.
I think monte carlo by itself is enough for the go endgame but I think the best idea is to teach the AI some combinatorial game theory. Making a database of moves with their CGT values would be very interesting.
The method the GP described is called stone counting. While it will give slightly different scores compared to normal counting methods it is in fact equivalent to the ancient Chinese way of counting.
Go is very different from Othello. In Go your goal is to surround empty points on the board. Because your stones stay on the board unless captured every previous move counts. It's hard to actually compare it to other games
here is an interactive introduction to go. Towards the end there are a few simple game examples explained.
The 8-dan was a professional 8-dan who are stronger then amateur 8-dan players. TFA mentioned that that Myungwan Kim estimated MoGo at about as strong as an amateur 2-dan with some moves being 5-dan moves.
It's true that the practical applications are pretty much only in the endgame. But as a mathematician what do I care about practical applications ;)
I wouldn't call UTC monte carlo algorithms brute force. Not to mention that MoGo also uses varies heuristics and expert knowledge databases.
Why wouldn't it be reducible? Just because we don't have enough time and computer power to calculate the entire game tree?
I do know that writing a go AI isn't easy, and brute force is definitely not the way to go. But what does that have to do with go being mathematical or not? Go has many beautiful mathematical properties.
What I posted first is just a simplification of course. Here is the algorithm in more detail. From a given position it will play thousands if not millions of games to determine which move gets the best winning percentage. This is how these games are played When you are at a node in the game tree and want to decide which move you will make next you first calculate a score for each possible move. Unexplored moves get a score based on a heuristic. moves that we have tried before get a score based on the number of times we have tried it and the winning percentage. The higher the winning percentage the higher the score, but the higher the number of times we have tried it the lower the score. This results in searching down the path with the best winning percentage most of the time. But there is still searching going on in the less good paths. See here and here for more details.
In chess you also have the advantage of being able to make an endgame database for your program. Chess positions only get simpler as the game progresses because pieces are removed. In Go this doesn't happen. An endgame database is simply impossible.
It is not really comparable to mathematically reducible games like Mancala, Chess, Backgammon, Draughts/Checkers, etc.,
Huh? What are you talking about? Go is much more 'mathematical' then chess or backgammon. It's one of the best examples of combinatorial game theory .
MoGo uses a Monte Carlo approach. Basically it keeps playing random games to see which give the best results. Faster computers mean more random games can be played thus given better results.
A GeV is an unit of energy and thus mass. But it's so little that expressing the mass in grams would lead to very small numbers. See wikipedia for more information.
You have a point but it's not exactly acurate. Renting a book from the library isn't copyright infringment. Of course you COULD copy the book you borrowed but by downloading something, you are allready infringing on copyright.
You mean like this one ?
Ah yes, you are right. I forgot about that. It's been a while since I worked with this kind of stuff.