I can solve tictactoe for $10. May not be so impressive, but it is surely a bargain. Also, I will not bodge the job like @tygxc.
Chess will never be solved, here's why
Sort:
optimissed I would rather try to employ @ tygxc. Because he is coming up with theories rather than just proving others wrong like you are
And then just being able to pop in and judge everyone Conversing
The point @tygxc is trying to get across is that it would take a lot of time, money, and effort keep it simple just leave it at that
But I do also understand optimissed
Guess we just can't all be as smart as you
*Sarcasm*
Listen i'm not super smart about some things but I am good about telling other people what somebody is trying to communicate to them
bye
mary?
Dec 16, 2022
0
#5073
magipi wrote:
tygxc wrote:
@6683
"How much does it cost"
++ I have estimated before:
3 grandmasters * 100,000 $/year/grandmaster * 5 years = 1.5 million $
3 cloud engines = 3000 desktops = 3000 * 500 $ = 1.5 million $
____________________________________________________________________________
Total 3 million $
Try running those resources not for 5 years, but 5 billion years. It's a more reasonable estimate, though probably still many-many magnitudes too low.
Not really, because after 5 years @tygxc would probably be sitting beneath palm trees on a sunny beach and impossible to contact.
I agree
Dec 16, 2022
0
#5075
Optimissed wrote:
tygxc wrote:
@6681
"Does the 5 years start every time you say it again"
++ No, the 5 years start after funding to hire the 3 grandmasters and rent the 3 cloud engines.
I know you never respond to me but it looks very much like you've got your sums wrong again.
Post #6655
If I had to solve chess, I wouldn't employ tygxc. That's a given. I'd rather employ Elroch because he tends towards caution. But if it's capable of being solved, it wouldn't be by someone who tends too far towards caution, either and someone who finds he cannot claim that he knows
1 e4 e5 2 B a6 loses tends far too far towards caution.
So I'd employ my son, who's managing a team of data scientists at the moment, to set it up. To tempt him to do so, I suppose I'd have to pay him £140,000 a year. And then maybe a team of five data scientists, at, say £40,000 to £60,000 a go, depending on expertise. At least one grandmaster. One might be got for £70,000 year. Let's see: that seems to be £460,000 per year for five years, which is 2.3 million, which translated into dollars is about $2.9 million in five years in wages alone.
How much does tygxc think it will cost for the entire thing, again?
If you are interested in commissioning me, I will be designing a quantum computer to do a strong solution.
Seriously, that is currently beyond the capabilities of humans (still more me) but at least it makes sense conceptually.
@6677
"We do know. There are only 10^44 legal positions, the vast majority of them stupid.
Now 10^44 = 2^146. So even if there are only 2 choices per move, 146 moves exhaust the number of legal positions. Checkmates with more than 146 moves must contain a string of forced only moves for the defending side. "
But you don't know that, and that is the whole reason why it would take a quantum computer to solve chess on a 8x8 board and thats what it means that 8x8 Chess is an EXP decision problem. It can easily be assumed that through pruning the four "known best moves" will lead to the best outcome but you literally can't say that because its not a guarantee that White will win if both sides play perfectly. You are talking about machine learning and through pruning you can reach 99% accuracy but unless you go through the entire decision tree of possible moves you could never know. Those four best known moves you are talking about are only based on the data we have now. Also I am a bit confused what you mean by "146 moves exhausts all the moves", https://www.chess.com/forum/view/endgames/who-wins-from-this-puzzle , right there is a thread of a game in the 7-move tablebase that takes 546 moves to play perfectly.
To put it right I think, 8x8 chess is in EXP time and generalized chess are in EXP-Complete time, both are solvable but they require an exponential amount of time to do it. Thus it would take a quantum computer to do it because the numbers are so vast, but the game is definitely "solvable". I think looking at your post that we agree that it is solvable, but definitely not through your method of pruning the data, because in essence all the data you have to "prune" is our training data. We do not know the whole thing, I think Claude Shannon, a renowned computer scientist can put it better than I can.
"These two facts imply (von Neumann and Morgenstern, 1944) that any given position of the chess pieces must be either: -
(1)A won position for White. That is, White can force a win, however Black defends.
(2)A draw position. White can force at least a draw, however Black plays, and
likewise Black can force at least a draw, however White plays. If both sides play
correctly the game will end in a draw.
(3)A won position for Black. Black can force a win, however White plays.
This is, for practical purposes, of the nature of an existence theorem. No practical method
is known for determining to which of the three categories a general position belongs. If
there were chess would lose most of its interest as a game. One could determine whether
the initial position is a won, drawn, or lost for White and the outcome of a game between
opponents knowing the method would be fully determined at the choice of the first move.
Supposing the initial position a draw (as suggested by empirical evidence from master
games [1]) every game would end in a draw."
tygxc wrote:
@6673
"it will take a supercomputer to solve chess."
++ It even takes 3 supercomputers during 5 years to weakly solve chess.
"positions that have been known to be draws that end up being mate in 546 moves"
++ Yes, but these long forced mates cannot be forced from the initial position.
"we really don’t know (yet) after 30000000000 moves"
++ We do know. There are only 10^44 legal positions, the vast majority of them stupid.
Now 10^44 = 2^146. So even if there are only 2 choices per move, 146 moves exhaust the number of legal positions. Checkmates with more than 146 moves must contain a string of forced only moves for the defending side.
"the computers we have today could solve it"
++ Yes, it takes 3 of these 5 years to exhaust the 10^17 relevant positions.
"We do know. There are only 10^44 legal positions, the vast majority of them stupid.
Now 10^44 = 2^146. So even if there are only 2 choices per move, 146 moves exhaust the number of legal positions. Checkmates with more than 146 moves must contain a string of forced only moves for the defending side. "
But you don't know that, and that is the whole reason why it would take a quantum computer to solve chess on a 8x8 board and thats what it means that 8x8 Chess is an EXP decision problem. It can easily be assumed that through pruning the four "known best moves" will lead to the best outcome but you literally can't say that because its not a guarantee that White will win if both sides play perfectly. You are talking about machine learning and through pruning you can reach 99% accuracy but unless you go through the entire decision tree of possible moves you could never know. Those four best known moves you are talking about are only based on the data we have now. Also I am a bit confused what you mean by "146 moves exhausts all the moves", https://www.chess.com/forum/view/endgames/who-wins-from-this-puzzle , right there is a thread of a game in the 7-move tablebase that takes 546 moves to play perfectly.
To put it right I think, 8x8 chess is in EXP time and generalized chess are in EXP-Complete time, both are solvable but they require an exponential amount of time to do it. Thus it would take a quantum computer to do it because the numbers are so vast, but the game is definitely "solvable". I think looking at your post that we agree that it is solvable, but definitely not through your method of pruning the data, because in essence all the data you have to "prune" is our training data. We do not know the whole thing, I think Claude Shannon, a renowned computer scientist can put it better than I can.
"These two facts imply (von Neumann and Morgenstern, 1944) that any given position of the chess pieces must be either: -
(1)A won position for White. That is, White can force a win, however Black defends.
(2)A draw position. White can force at least a draw, however Black plays, and
likewise Black can force at least a draw, however White plays. If both sides play
correctly the game will end in a draw.
(3)A won position for Black. Black can force a win, however White plays.
This is, for practical purposes, of the nature of an existence theorem. No practical method
is known for determining to which of the three categories a general position belongs. If
there were chess would lose most of its interest as a game. One could determine whether
the initial position is a won, drawn, or lost for White and the outcome of a game between
opponents knowing the method would be fully determined at the choice of the first move.
Supposing the initial position a draw (as suggested by empirical evidence from master
games [1]) every game would end in a draw."
@6673
^^ And the reason why his quote his important is because like you said, if you knew a position was a given for white (what you think is e4, nf3, c4), the decision tree would be way smaller to exhaust through, but you DONT
Optimissed wrote:
asher48 wrote:
The point @tygxc is trying to get across is that it would take a lot of time, money, and effort keep it simple just leave it at that
No, wrong. He's saying it would take virtually no money ($3 mil) and virtually no time (5 years). He hasn't a clue about the reality and neither, unfortunately, have you. It's better to understand who's talking sense and not to guess. Elroch is also talking sense, as are others.
I agree. He's not saying it would take a lot of time, money and effort. He's saying the opposite. He's saying it would take very, very little time, money, and effort.
He's the only person I've ever seen that says solving chess is very, very easy. Very little time, very little money, and very little effort. I think if it were as easy as he says, it would have been done by now. Several times over.
lfPatriotGames wrote:
Optimissed wrote:
asher48 wrote:
The point @tygxc is trying to get across is that it would take a lot of time, money, and effort keep it simple just leave it at that
No, wrong. He's saying it would take virtually no money ($3 mil) and virtually no time (5 years). He hasn't a clue about the reality and neither, unfortunately, have you. It's better to understand who's talking sense and not to guess. Elroch is also talking sense, as are others.
I agree. He's not saying it would take a lot of time, money and effort. He's saying the opposite. He's saying it would take very, very little time, money, and effort.
He's the only person I've ever seen that says solving chess is very, very easy. Very little time, very little money, and very little effort. I think if it were as easy as he says, it would have been done by now. Several times over.
its because hes talking about machine learning methods. You train those on data and then use them to make accurate predictions, you can never be 100% sure of anything this way because you never have all the data necessary or its a chaotic system (in this case we are missing data) and once you do you wouldn't need a machine anymore cause everything could be logged in a table
@6683
"How much does it cost"
++ I have estimated before:
3 grandmasters * 100,000 $/year/grandmaster * 5 years = 1.5 million $
3 cloud engines = 3000 desktops = 3000 * 500 $ = 1.5 million $
____________________________________________________________________________
Total 3 million $
Try running those resources not for 5 years, but 5 billion years. It's a more reasonable estimate, though probably still many-many magnitudes too low.
Imagine how far 3 million dollars will go then. Probably barely enough to buy a used chess clock, circa 2022.