What they should do is to put Alpha- Zero against itself and see how it goes and keep lowering the time until it gets interesting.
Ummm...you do understand how Alpha Zero learned Chess, right? By playing itself millions of games...
It is fairly clear from the foregoing that most board games have been effectively 'solved' by computer programs.
Nonsense. Only the simpler ones have (checkers is one of the most complex).
That is to say that if a human no matter how good cannot beat a program at that game then we have to admit defeat in effect.
That is nothing like what "solved" means.
[the rest made sense]
However, what this means for AI and the consequences for us humans isn't quite as clear. Solving a very (to us Humans) complex, but strictly defined problem even if the program is capable of self-learning within those parameters so that for example, it could also teach itself to play 360 chess or some other variant as well as standard chess does not mean we are on the verge of self-aware computers.
Sure, machines can put vehicles together and mow lawns and there is no real reason except for cost that such machines should not also be able to play a board game as well. This does not imply that they have a sense of self, or that they can ponder in any meaningful way on the big questions of philosophy. Why are we here? From where did we come? How do we know we are free? And so on.
This morning I had to feed my dogs, chop wood for the fire, interact with a few clients online, go to the grocery store, and now restore some fiber arts equipment. When a machine can make those sorts of decisions then it is time to talk.
AlphaZero looked at far fewer variations than stockfish. "Rather than look at the whole tree--it worked out which branch to look at."
This is interesting as this is the same way Correspondence GrandMasters play using chess engines. They force their chess engine to analyze the most promising variation [or variations] and discard the rest.
This is one reason a strong human player with a strong machine can usually beat a strong machine.
.. Just noticed, the following ..
.. I have to admit ; I'm fairly impressed !
AlphaZero looked at far fewer variations than stockfish. "Rather than look at the whole tree--it worked out which branch to look at."
This is interesting as this is the same way Correspondence GrandMasters play using chess engines. They force their chess engine to analyze the most promising variation [or variations] and discard the rest.
This is one reason a strong human player with a strong machine can usually beat a strong machine.
I don't think that correspondence players are allowed to use engines. That would defeat the purpose of playing correspondence chess. It's supposed to be about the art and beauty of the game. I love looking deep into a position (5-10 moves deep, into every candidate move variation) to find the best move. You're only cheating yourself out of that experience if you use an engine.
AlphaZero looked at far fewer variations than stockfish. "Rather than look at the whole tree--it worked out which branch to look at."
This is interesting as this is the same way Correspondence GrandMasters play using chess engines. They force their chess engine to analyze the most promising variation [or variations] and discard the rest.
This is one reason a strong human player with a strong machine can usually beat a strong machine.
I don't think that correspondence players are allowed to use engines. That would defeat the purpose of playing correspondence chess. It's supposed to be about the art and beauty of the game. I love looking deep into a position (5-10 moves deep, into every candidate move variation) to find the best move. You're only cheating yourself out of that experience if you use an engine.
It depends on the CC league and tournament. A lot more are allowing it as it is impossible to enforce a rule like that. For instance, I believe ICCF allows their use.
"Centaur Chess" should be commonly understood.
Some of the posters above haven't read that memo, unfortunately.
I don't think that correspondence players are allowed to use engines. That would defeat the purpose of playing correspondence chess. It's supposed to be about the art and beauty of the game. I love looking deep into a position (5-10 moves deep, into every candidate move variation) to find the best move. You're only cheating yourself out of that experience if you use an engine.
It depends on the CC league and tournament. A lot more are allowing it as it is impossible to enforce a rule like that. For instance, I believe ICCF allows their use.
That sucks. I could become a legitimate correspondence Master if people didn't cheat with computers.
Will there ever be a computer strong enough to solve chess to the point where white uses its half tempo advantage to always beat black no matter what moves black plays (in otherwords the same computer can never win with black even after a thousand random games against itself)
I beleive one day there will be a computer so strong and so big that it will solve chess completely but perhaps that is 50 or 100 years off, its possible to solve it but we may never see it even in a 100 years
Check out the book reviewed here - https://www.chess.com/blog/smurfo/the-secret-of-chess
It might help with your query.
I don't think that correspondence players are allowed to use engines. That would defeat the purpose of playing correspondence chess. It's supposed to be about the art and beauty of the game. I love looking deep into a position (5-10 moves deep, into every candidate move variation) to find the best move. You're only cheating yourself out of that experience if you use an engine.
It depends on the CC league and tournament. A lot more are allowing it as it is impossible to enforce a rule like that. For instance, I believe ICCF allows their use.
That sucks. I could become a legitimate correspondence Master if people didn't cheat with computers.
Play here: computer use is banned, and this is enforced quite efficiently. I have had many opponents banned for cheating, (and won quite a few games because of this).
Rsava is right that the great thing about turn-based chess is the deep exploration of possibilities.
No, "solved" has a very straightforward definition, which goes beyond merely winning all the time.
Checkers and tic-tac-toe are solved, that doesn't stop people from playing them.
.. More, 'grist' for the mill. -- Well, actually ; 'Rolled oats'; For those, who were hoping for 'corn'.
http://www.dailymail.co.uk/sciencetech/article-5271343/The-machine-read-better-you.html
which computer
It wouldn't be that hard to figure out if a computer solved chess, just put into the program what the definition of 'solved' is -winning all games- then having it answer yes or no if it did it.
Going back to an earlier post, I agree that 'solved' in strict terms has a definite meaning with regard to board games. My point earlier was that if and when programs can reliably beat unaided humans in every game, as black or white, then the game is for all intents and purposes 'solved'.
It is fairly clear from the foregoing that most board games have been effectively 'solved' by computer programs. That is to say that if a human no matter how good cannot beat a program at that game then we have to admit defeat in effect.
However, what this means for AI and the consequences for us humans isn't quite as clear. Solving a very (to us Humans) complex, but strictly defined problem even if the program is capable of self-learning within those parameters so that for example, it could also teach itself to play 360 chess or some other variant as well as standard chess does not mean we are on the verge of self-aware computers.
Sure, machines can put vehicles together and mow lawns and there is no real reason except for cost that such machines should not also be able to play a board game as well. This does not imply that they have a sense of self, or that they can ponder in any meaningful way on the big questions of philosophy. Why are we here? From where did we come? How do we know we are free? And so on.
This morning I had to feed my dogs, chop wood for the fire, interact with a few clients online, go to the grocery store, and now restore some fiber arts equipment. When a machine can make those sorts of decisions then it is time to talk.