How long till we solve chess?

File: 78 KB, 800x741, chess.jpg [View same] [iqdb] [saucenao] [google]

How long till we solve chess? Anonymous Thu Jun 13 15:54:51 2019 No.10722389 [Reply] [Original]

Do you think we will be able to solve chess anytime soon? Next 10 years? 100? 1,000?

I know we have databases of solvable endgames but has this led to any interesting math that can generalize strategies/tactics?

>>	Anonymous Thu Jun 13 15:56:09 2019 No.10722391 >>10722389 How do you know that chess is solvable?

Anonymous Thu Jun 13 15:57:29 2019 No.10722396

>>10722389
Information theorist Claude Shannon argued in 1951 that it is not feasible for any computer to actually solve chess, since it would either need to compare some 10120 possible game variations, or have a "dictionary" denoting an optimal move for each of the about 1043 possible board positions.[4] It is thus theoretically possible to solve chess, but the time frame required (according to Shannon, 1090 years) puts this possibility beyond the limits of any feasible technology.

>>	Anonymous Thu Jun 13 15:58:56 2019 No.10722399 >>10722396 thats 10^120 possible games 10^43 game positions 10^90 estimated years

Anonymous Thu Jun 13 15:59:15 2019 No.10722400

>>10722389

I am 24 years old and have been playing chess for 19 years. I have a FIDE rating close to 2000. Believe me, we will never solve chess. There are more potential positions in this game then there are stars in the universe. Not even Deep Mind can come close to solving chess.

>>	Anonymous Thu Jun 13 16:06:23 2019 No.10722416 >>10722400 >There are more potential positions in this game then there are stars in the universe yes but you only need a subset of those positions to solve it.

>>	Anonymous Thu Jun 13 16:08:16 2019 No.10722426 >>10722416 It won't happen at least not in our lifetime

Anonymous Thu Jun 13 16:11:02 2019 No.10722432

there may be 10^90 positions or whatever, but you can eliminate types of positions in groups. Like you can eliminate positions where your king is diagonal from a pawn or a bishop, etc. So you just have to eliminate types of positions using game theory. The more positions you eliminate, the amount of positions to choose from decreases, until you've eliminated every position but the optimal one.

>>	Anonymous Thu Jun 13 16:15:51 2019 No.10722447 >>10722432 So in other words, to solve chess you wouldn't calculate every position manually, but you would just eliminate groups of positions at a time. Then you could find the optimal move in way less than 10^43 operations.

>>	Anonymous Thu Jun 13 16:18:14 2019 No.10722450 >>10722400 >I have been playing chess for 19 years my condolences

>>	Anonymous Thu Jun 13 16:18:40 2019 No.10722454 >>10722450 It's the best game anon.

>>	Anonymous Thu Jun 13 16:19:46 2019 No.10722459 >>10722389 Maybe someone will discover a general equation for all chess

>>	Anonymous Thu Jun 13 16:24:12 2019 No.10722474 >>10722450 >I have been playing chess for 19 years How can you play chess and stay awake for that long?

Anonymous Thu Jun 13 16:25:52 2019 No.10722478

>>10722447
But that's kind of subjective though, because maybe long term moving that king would be beneficial ...... The way to 'solve' chess would be for a computer to have played every game of chess every possible, IE to factor in every single move that could ever have been made. There's just no way to do that unless you have a giga giga giga giga giga chad computer

>>	Anonymous Thu Jun 13 16:26:59 2019 No.10722482 >>10722454 that would be fortnite actually

>>	Anonymous Thu Jun 13 16:31:42 2019 No.10722497 >>10722478 Not true. That's kind of like saying for there to be an equation for a 1st degree polynomial you have to compute every 1st degree polynomial. You can define the optimal move in terms of an equation that has less than 10^90 terms.

>>	Anonymous Thu Jun 13 16:33:03 2019 No.10722501 >>10722391 https://en.wikipedia.org/wiki/Zermelo%27s_theorem_(game_theory)

Anonymous Thu Jun 13 16:35:45 2019 No.10722509

>Create arbitrary number of neural networks (say, 16).
>Tell them that winning games is good, and losing is bad. Because we don't want a stagnant pile of stalemating AI, we give tie games a slight penalty.
>Have them play against each other.
>Come back in a year or two.
Done.

Anonymous Thu Jun 13 16:37:57 2019 No.10722521

>>10722389
Yeah, it’s not that difficult. A master chess player can generally win every game unless the opponent plays the right moves early on. We don’t have to look through quadrillions of sequences because only a relatively small batch of them are valid, competitive move sequences. For example, the King’s Gambit has already been shown to be objectively bad for white, as black can take the pawn and play d6 as Fischer recommended long ago.

>>	Anonymous Thu Jun 13 16:39:07 2019 No.10722526 >>10722509 That's not solving. We already have extremely successful chess playing AI. solving means something totally different.

>>	Anonymous Thu Jun 13 16:40:10 2019 No.10722528 >>10722521 >Yeah it's not that difficult if you don't do it Why does no one understand what solving means? Obviously you don't need to solve a game to be good at it.

Anonymous Thu Jun 13 16:40:38 2019 No.10722530

>>10722509
Supposedly this already exists in the AlphaZero chess AI. It played itself for 4 hours and beat the best chess engine around. However, only 10 out of 100 games were released, and the time settings seemed to favor AlphaZero since it had essentially memorized a bunch of moves whereas Stockfish had to calculate in-game. It’s a wonder why they haven’t had AlphaZero play itself for months and release its findings

Anonymous Thu Jun 13 16:41:38 2019 No.10722534

>>10722478
>>10722497
Think of it like this. If you can show that moving a piece to a certain position will always result in a loss then you can reduce the number of potential future moves by a factor of 1000 or whatever. Let's say you can show that say 5 moves always result in a loss if the other opponent plays optimally. Then the number of potential future states shrinks by that much, and you can forego searching along those branches. This exponentially reduces the search space. And if you can't eliminate a move, then you can eliminate 2nd turn moves, which also reduce the search space exponentially, and so on.

>>	Anonymous Thu Jun 13 16:43:20 2019 No.10722536 What about the human side of chess though? As in the psychological side?

Anonymous Thu Jun 13 16:46:59 2019 No.10722540

>>10722536
There is, without any doubt, a provably existing but not yet calculated solution for chess. If you have that solution the human / psychological aspect would be rendered 100% obsolete. Would be like trying to use psychology in a tic tac toe game when the right move to make is already known and has nothing to do with psychology.

Anonymous Thu Jun 13 16:47:55 2019 No.10722544 [DELETED]

>>10722389

>>10722396
Most of those positions won't be reachable in practical games. As the most obvious thing, you could truncate moves that lead to a forced mate, as well as transpositions of moves. (there is no difference between 1. e4 e5 2. Nf3 Nf6 and 1. Nf3 e5 2. e4 Nf6)

Depends, it seems unsolvable now, but there may be surprises, like with the 7 piece tablebases, which came unexpectedly early. We may possibly see some openings getting solved in near future.

Anonymous Thu Jun 13 16:49:01 2019 No.10722548

>>10722526
A truly unbeatable AI would, by definition, have solved the game it is playing. There is no possible means to actually beat it, and depending on the ultimate solution, the most one could possibly get is a stalemate. Yes, we have "successful" AI right now, but those are for the most part programmed by hand, and can't alter their strategy. Neural networks can learn, and may improve themselves as long as they continue running.

>>	Anonymous Thu Jun 13 16:50:04 2019 No.10722551 >>10722540 This is patently not true. Checkers has been hard solved and people still play it. A solution for chess would be too complex to memorize, most likely. It would be some theoretical trivia.

>>	Anonymous Thu Jun 13 16:50:44 2019 No.10722555 >>10722548 >A truly unbeatable AI You can't know that unless you actually solve chess. For all you know the trained program could win the first quadrillion games and then lose one. NOT solving.

>>	Anonymous Thu Jun 13 16:51:48 2019 No.10722560 >>10722551 No, it is true. Just because you choose not to generate the ideal checkers move whilr you play doesn't mean human psychology still matters. It just means you're choosing to ignore the real answers.

Anonymous Thu Jun 13 16:54:16 2019 No.10722569

>>10722389 (OP)

>>10722396
Most of those positions won't be reachable in practical games. As the most obvious thing, you could truncate moves that require a player to blunder away a forced mate, as well as transpositions of moves. (there is no difference between 1. e4 e5 2. Nf3 Nf6 and 1. Nf3 e5 2. e4 Nf6)

Depends, it seems unsolvable now, but there may be surprises, like with the 7 piece tablebases, which came unexpectedly early. We may possibly see some openings getting solved in near future.

>>	Anonymous Thu Jun 13 16:55:14 2019 No.10722570 >>10722560 You don't choose, you simply can't see it because your brain is not capable of computing 500 move decision trees. Being nervous/tired/etc will still have an effect on how you play checkers. For chess this is probably exponentially worse.

>>	Anonymous Thu Jun 13 16:57:36 2019 No.10722573 >>10722570 You choose because you can have the answers generated. Whether or not they exist in your brain is irrelevant.

>>	Anonymous Thu Jun 13 17:00:36 2019 No.10722579 >>10722573 I'm not sure if they will let you walk into a chess tournament with your "solution generator".

Anonymous Thu Jun 13 17:02:48 2019 No.10722583

>>10722579
That just means (in the case of checkers, and the hypothetical case of chess) you and others are collectively choosing to ignore reality. You can do that and behave as though psychology matters, but there's still a real answer for any given position and it renders psychology obsolete when used.

>>	Anonymous Thu Jun 13 17:03:05 2019 No.10722584 >>10722579 That's different. This thread is about solving chess, not solving human vs human chess.

Anonymous Thu Jun 13 17:04:01 2019 No.10722586

>>10722555
So we do what they did with checkers: run every possible game against the eventual AI. If it wins every time, great. If it loses, we run it back. Considering AI can play at superhuman speeds, it shouldn't take too long, right?

Of COURSE we'd test it before saying we've solved it, you dumb shit.

Anonymous Thu Jun 13 17:06:25 2019 No.10722591

>>10722583
>>10722584

I mean we already have unbeatable chess AI and I think the same argument largely applies.

Should we just ban human vs human tournament because no human can beat an AI anyway? Just let the AIs continue developing things?

The reason we play stuff is not because they are unsolved. Something being solved also doesn't imply it becoming unplayable or uninteresting. I'm not sure how this implies "ignoring reality".

Anonymous Thu Jun 13 17:07:25 2019 No.10722598
File: 178 KB, 1190x906, bd8.png [View same] [iqdb] [saucenao] [google]

>>10722586
>run every possible game against the eventual AI
That's retarded and even MORE work then actually solving chess would require since you probably wouldn't get an AI that's perfect on your first try, meaning you're now running through every possible game MULTIPLE TIMES.

Anonymous Thu Jun 13 17:09:49 2019 No.10722606

>>10722591
Dude, the guy you were talking to said that psychology becomes obsolete when the game is solved, not that psychology isn't relevant for human vs human games. Just because something becomes obsolete that doesn't mean people don't still use psychology when they play the game. Unless they memorize the algorithm.

>>	Anonymous Thu Jun 13 17:10:15 2019 No.10722608 >>10722591 >Should we just ban human vs human tournament because no human can beat an AI anyway? No, it's enough just to make fun of them for making shit human moves in a fully solved game.

Anonymous Thu Jun 13 17:11:25 2019 No.10722610

>>10722606

No, he specifically gave the example of tic tac toe where you know the solution and it renders psychology useless because of that. You can't say the same thing of chess and checkers because solutions are too complex for a human to just use at will.

>>	Anonymous Thu Jun 13 17:14:02 2019 No.10722619 >>10722610 I don't think his point was that psychology was useless, but that it's obsolete, in other words if you used the algorithm then psychology wouldn't be relevant.

Anonymous Thu Jun 13 17:23:16 2019 No.10722647

>>10722610
You can do the same thing for tic tac toe you're talking about for another solved game like checkers. Would just require that you get extremely intoxicated or brain damaged. Which is the point. Human psychology for solved games only matters if you intentionally ignore the solutions. That it's easier for a human brain to commit the solution to some games to memory more than others doesn't matter at all. That's a completely arbitrary happenstance which can be altered in either direction with pharmaceuticals and/or mechanical protheses.

Anonymous Thu Jun 13 17:25:04 2019 No.10722651

>>10722598
Except proving a solution chess would require exactly that, fixing every mistake along the way. Which is what we always do.
Don't worry, I know your panties are only in a twist because, unlike glorious learning AI, your shit brain can't go from "doesn't know how to play chess" to "beating world champions" in 24 hours of self training.

>>	Anonymous Thu Jun 13 17:25:48 2019 No.10722654 >>10722651 I'm not anti-AI. I just know what 'solved' means, unlike you.

>>	Anonymous Thu Jun 13 17:33:19 2019 No.10722674 >>10722654 >Attaining a sequence that always leads to the predicted outcome doesn't mean you solved it ...I really don't think you do.

>>	Anonymous Thu Jun 13 17:34:02 2019 No.10722676 >>10722501 >if the game cannot end in a draw >chess games can't end in a draw

Anonymous Thu Jun 13 18:10:02 2019 No.10722764

>>10722534
The difficult part about that is knowing which moves lead to a loss (specially in the early game). Unless we can develop a way of assessing a move without super-exponential growth, current technology will probably not suffice. In my opinion, it will be centuries before we solve chess.

Anonymous Thu Jun 13 18:11:51 2019 No.10722771

>>10722676
Did you miss the next sentence?

"An alternate statement is that for a game meeting all of these conditions except the condition that a draw is not possible, then either the first-player can force a win, or the second-player can force a win, or both players can force a draw"

>>	Anonymous Thu Jun 13 18:48:10 2019 No.10722839 >>10722676 Go look up what solving means.

Anonymous Thu Jun 13 18:50:29 2019 No.10722843

>>10722674
You're talking about using neural networks, the entire point of which is to solve a problem *without* explicitly determining the instructions for how to do so. And then you're saying it's OK because you can just check it by running through every possible game, completely defeating the purpose of using ANN in the first place. Just stop.

>>	Anonymous Thu Jun 13 18:51:00 2019 No.10722844 >>10722674 not him but please shut up you fucking retard you are embarrassing yourself

>>	Anonymous Thu Jun 13 18:59:05 2019 No.10722853 >>10722459 10/10

>>	Anonymous Fri Jun 14 00:02:45 2019 No.10723475 >>10722474 lol

>>	Anonymous Fri Jun 14 02:03:41 2019 No.10723710 >>10722400 >19 years >only 2000 My condolences

>>	Anonymous Fri Jun 14 02:08:23 2019 No.10723713 >>10722400 LOL at this hubris. Hope you suffer for it when all those decades of your life culminate in humiliating defeat.

Anonymous Fri Jun 14 02:13:49 2019 No.10723727

>>10722389

Safe money: barring major catastrophe (nuclear war, the antibiotics all just up-and-stop-working one day), it will DEFINITELY occur during this century.

This same topic was being discussed on this same board like over a year ago, and I held forth a similar/more "optimistic" opinion (more like within the next 10/40 years). I was shocked to read all the disagreements, which either totally missed the point with sentimentality "people will still play chess and enjoy it even if it's solved anon" (yes, of course they will, this is beside the point), or more substantively, by insisting that the computational power+sufficiently efficient algorithm to actually solve chess would not be brought into existence during the above interval, or during the 21st century more broadly. This seems to be to be a ludicrous prediction, an untenable one, given historical trends.

I think that, taken as a sample, /sci/ posters are predisposed to sentimentality about chess.

Anonymous Fri Jun 14 03:52:52 2019 No.10723936

>>10723727
or you like other people in the thread don't know what 'solved' means. it means producing a game tree which is optimized for one/both of the players. 'efficient algorithms' don't reduce the number of states that exist in this game tree. at most people/computers could mathematically prove things that pare the game tree down but this is dissimilar from what people are actually doing in their chess engines

>>	Anonymous Fri Jun 14 03:56:32 2019 No.10723940 >>10722389 >implying it hasn't already practically been solved

Anonymous Fri Jun 14 04:38:23 2019 No.10724004

>>10723940
>implying it’s not cheat code tier
I’ve beaten strong computers doing this by luck and grit. Like move your king right then left then strike your black bishop into his pawn and lose it then strike your rook and lose it but somehow checkmate is now unavoidable for your opponent.
It won’t be the faggy standard intro when it’s solved.

>>	Anonymous Fri Jun 14 04:53:41 2019 No.10724024 >>10723940 Either you think that our strongest chess engines now always chooses the optimal move, or you do not understand what "solved" means.

Anonymous Fri Jun 14 05:31:40 2019 No.10724058

>>10723936

I know what solved games, unfair games, perfect play, etc, are. You self-owned about the caveat to which you'd objected, so I don't know why you bothered writing that. Chess will be capital-S Solved during the 21st century. I'm inclined to think it's a draw.

>>	Anonymous Fri Jun 14 06:01:44 2019 No.10724091 Chess is retarded, try playing actually fun games like Sengoku Rance

>>	Anonymous Fri Jun 14 07:13:51 2019 No.10724168 File: 262 KB, 1280x720, DeepMind-AlphaZero.jpg [View same] [iqdb] [saucenao] [google] >>10722389 Did you heard about Alpha Zero?

>>	Anonymous Fri Jun 14 07:30:01 2019 No.10724181 >>10722389 What is practical application in solving chess? We solved tic-tac-toe, now what? Humans will still play chess even after solving

Anonymous Fri Jun 14 08:09:35 2019 No.10724220

>>10723727
>>10724058
to find an optimal move for 10^43 board positions, even at a rate of solving 10^10 per second, it would take 3x10^22 years
you seem completely oblivious to the size of this problem
even if some algorithm could cut the game in half, it would take longer to solve the game than the lifespan of our sun
then trying to dig through the game solution to apply it to any specific game takes another uncountable amount of eons
solved in the 21st century? lmao please

>>	Anonymous Fri Jun 14 08:11:05 2019 No.10724222 >>10724168 didnt solve it, alpha zero is just a very good player the best koreans havent solved star craft

>>	Anonymous Fri Jun 14 08:18:02 2019 No.10724226 File: 1.27 MB, 1500x2000, Klaudia_Wiśniowska_2013.jpg [View same] [iqdb] [saucenao] [google] tfw no chess playing gf tfw no gf at all

>>	Anonymous Fri Jun 14 09:18:37 2019 No.10724281 >>10722396 Yet within the reach of feasible math.

>>	Anonymous Fri Jun 14 09:25:33 2019 No.10724285 >>10722391 You should go check out the very basics of what you're talking about.

>>	Anonymous Fri Jun 14 09:26:59 2019 No.10724288 >>10722676 When applied to chess, Zermelo's Theorem states "either White can force a win, or Black can force a win, or both sides can force at least a draw".

>>	Anonymous Fri Jun 14 09:27:33 2019 No.10724289 >>10722530 Go read the nature paper they released hundreds of games and it doesn't memorize anything, it also generalizes to new positions that were never seen in training. Basically everything you said is wrong and you are a retard.

>>	Anonymous Fri Jun 14 09:29:52 2019 No.10724293 >>10724220 Go look at the very basics of game theory you can eliminate almost all the positions when you're just looking for optimal play.

>>	Anonymous Fri Jun 14 09:31:01 2019 No.10724295 Weak solve maybe, strong solve obviously no.

>>	Anonymous Fri Jun 14 09:40:57 2019 No.10724304 >>10722586 >run every possible game against the eventual AI. Even if it only took 1 second per game the AI wouldn't be done until Earth was long gone.

Anonymous Fri Jun 14 10:07:15 2019 No.10724343

>>10724293
please clarify, because I would love to hear how game theory is going to simplify the sheer amount of computation time into a feasible problem. Be specific, I know my game theory
constructing a game matrix for every specific game state will simplify nothing
trying to construct a vast strategy matrix is a huge waste of time because it means more computations than the original problem
truncating a decision tree means you have to spend uncountable years just constructing the tree. You cant truncate the branches early until you know the results of every possible consecutive decision, because what can seem like early unwinnable states still need millions of years to find a conclusive answer, and figure out how not to accidently put yourself into a losing strategy
you could spend lifetimes creating a comprehensive strategy guide (which none exists after thousand years of chess) which lets you eliminate the vast majority of unoptimal strategies, and wow you reduced it to a billion years of computations
solving a game means an analysis of every strategy, not just finding the obvious strategies, because especially for a game like chess, some of the best strategies have been discovered to be weak against a strategy nobody discovered until it was needed
Solving the game isnt just finding a handful of strategies that are very good and recallable in a feasible time limit like chess engines do. Giving them better processing speed and more time per move vastly improves their performance, which isnt solving the game. Solving the game means an analysis of every strategy and every counter strategy, which trillions exist of. There is no shortcut to skip that
even the best chess players and engines only plot out possible outcomes like 6 moves ahead, and their game theoretical approaches either are loose qualitative strategies, or valuation methods to approximate advantages because they *can't* solve the game from that point (until about 5 moves until mate)

Anonymous Fri Jun 14 13:17:30 2019 No.10724644

>>10724058
>I know what solved games, unfair games, perfect play, etc, are
lol

the algorithms people use to play chess don't mathematically prove anything about chess you pseud. actually proving things would require completely different programs looking for proofs that we have no reason to believe actually exist.

what's happening here is that you have no actual knowledge about computation and only wikipedia level knowledge of game theory so you lapse into reddit 'fucking love science' tier mysticism about how ai can do anything without understanding anything about the problem being discussed other than it vaguely sounds like playing good chess

>>	Anonymous Fri Jun 14 14:51:35 2019 No.10724830 >>10722396 >beyond the limits of any feasible technology this should go into the wiki article about hubris. how much of today's computing power was imaginable in the eighties, when I went to university? almost none of it.

>>	Anonymous Fri Jun 14 15:14:05 2019 No.10724882 >>10724304 So get a bunch of computers and run multiple instances of it. Besides, the AI would, ideally, always be making the best decision, meaning a vast majority of game-states would likely never occur.

Anonymous Fri Jun 14 15:55:25 2019 No.10724968

>"Mmm, chess... You're so BIG."
>"You couldn't POSSIBLY be solved."

Checkers has 500 quintillion game positions (5x10^20), but is still considered to have been solved for over a decade. With artificial intelligence. And because Moore's Law still appears to be in effect, our computers have been getting significantly better ever since.

>>	Anonymous Fri Jun 14 16:02:45 2019 No.10724980 >>>/sp/

>>	Anonymous Fri Jun 14 16:39:41 2019 No.10725053 >>10722586 You are a special kind of retard, aren't you?

>>	Anonymous Fri Jun 14 17:09:10 2019 No.10725107 >>10725053 >Let's just do the exact same thing they did with a similar problem, and we'll probably get similar results. >RETARD You know? I'm still arguing about stupid shit like this on 4chan, so sure. Why not. I'm a retard.

>>	Anonymous Fri Jun 14 17:12:55 2019 No.10725113 >>10725107 Not him but part of the problem is you really don't understand at all how the checkers solution actually worked. They didn't do it by having a neural network train on games. You're confusing a variety of different topics.

Anonymous Fri Jun 14 17:58:31 2019 No.10725192

>>10722400
>I am 24 years old and have been studying chemistry for 19 years. I have a PhD in biochemistry s.c.l. Believe me, we will never predict protein folding. There are more potential sequences of amino acids then there are stars in the universe. Not even Blue Crystal 4 can come close to solving enzymes.

Anonymous Fri Jun 14 18:28:41 2019 No.10725251

>>10724882
>So get a bunch of computers and run multiple instances of it.
like a billion?
>Besides, the AI would, ideally, always be making the best decision, meaning a vast majority of game-states would likely never occur.
we cant verify the AI is making the best moves without solving the game pretty much exhaustively. not solved

>>	Anonymous Fri Jun 14 18:31:06 2019 No.10725256 >>10724968 every piece on checkers is the same with one direction of movement with a basic understanding of chess and combinatorics, you should realize how vastly different the games are >we already solved sudoku, crossword puzzles are just as easy!

Anonymous Fri Jun 14 19:17:35 2019 No.10725340

>>10722396
but with computing power today, isnt it possible to calculate all of the theoretically possible games moves for any given set of game board conditions, and thereby by calculate every possible outcome?
i mean, if we have computers that can make a billion calculation per second or whatever, isn't it just a matter of modelling the gameplay and pressing enter?

>>	Anonymous Fri Jun 14 19:46:04 2019 No.10725387 >>10725340 not in anywhere near our lifetimes dont you think somebody would have instead of investing all this work into chess engines and valuation methods?

>>	Anonymous Fri Jun 14 20:57:51 2019 No.10725543 >>10725387 why cant we write a logic based program with all of the game rules defined and then tell it to start iterating moves based on basic pre-programmed goals?

>>	Anonymous Fri Jun 14 21:15:14 2019 No.10725575 File: 75 KB, 562x422, Hikaru_no_Go_19-750x5631.jpg [View same] [iqdb] [saucenao] [google] >>10722400 >laughinggirls

Anonymous Fri Jun 14 21:22:25 2019 No.10725580

>>10725543
we can, alpha zero is the best chess engine to date, and it is entirely self taught
that would take magnitudes longer to solve chess than creating a catalog of every move and countermoves, which would already take an unfathomable amount of time to do.
solving a game means finding either a completely optimal strategy that cant lose (which we would have found one by now if existed for chess imo), or find the best counter strategy for *any* strategy an opponent would employ, which is such a huge computation problem that humanity will never solve it. We really cant catalog the best counter strategy for every strategy without exhausting every possible strategy to begin with

Anonymous Fri Jun 14 22:26:17 2019 No.10725651

>>10725580
Alpha zero beat SF8 by a decent margin. However, Leelachesszero and SF Dev beat SF8 by an immeasurable margin. NN engines as well as Stockfish have come a long way since then. I suggest you look up TCEC or Chess.com's chess engine tournaments if you want to see the best compete against the best.

>>	Anonymous Sat Jun 15 00:02:03 2019 No.10725766 >>10724181 >t. brainlet There's literally 0 point of two people that understand tic-tac-toe playing, it's catsgame every time

>>	Anonymous Sat Jun 15 00:04:24 2019 No.10725770 >>10723713 t. 750 FIDE

Anonymous Sat Jun 15 00:08:29 2019 No.10725773

>>10725192
>I am 24 years old and have been studying physics for 19 years. I have a PhD in astrophysics. Believe me, we will never understand blackholes. They are harder for us to observe than my penis. Not even the EHT can come close to solving blackholes.

>>	Anonymous Sat Jun 15 00:10:19 2019 No.10725777 >>10722426 For the sake of your self-worth, sure.

>>	Anonymous Sat Jun 15 00:16:03 2019 No.10725788 >>10725773 I am 27 years old and have a PhD in chemical biology with a specialization in synethetic biology

>>	Anonymous Sat Jun 15 00:17:59 2019 No.10725791 >>10725788 Can you grow me a larger penis?

xxXSuiseisekiXxx Sat Jun 15 00:19:01 2019 No.10725793

>>10722399
Hmmm aound 1.25×10^30 terabytes
Hang on I think I got this.

The human genome is about 3,120,000,000 DNA base pairs long, so half of that is in each spermatozoa. That gives us 1,560,000,000 base pairs in a single sperm cell.

Each of those base pairs can be an A-T bond or a G-C bond, and can be aligned in either direction. That means there’s 4 ways it can be aligned, and that can be represented in two bits of data: 00=G, 01=C, 10=A, and 11=T, for example.

Now, the average dude lets out about 50 and 500 million sperm with each go. Rough average says that it’d probably be about 200 million, right? If we take all this information and combine it with the wonders of mathematics, we have 1.560*10^9 * 2 bits * 2.00*10^8. Do all the equational shit, and we have 6.24*10^17 bits transmitted in a single burst.

That’s 78,000 Terabytes, in what amounts to a half-minute-at-most event.

We could fit a digitized copy of the Library of Congress into your spooge. It’d only take about 20 terabytes.

So your dick has a higher bandwidth than any Internet connection that ever existed, and is likely to exist any time soon.

Of course, god knows most of you would only use that power to surf for pr0n.

>>	Anonymous Sat Jun 15 00:52:22 2019 No.10725840 >>10723727 considering that we are getting pretty close to transistor size limits, you can‘t really expect computaional power to keep exploding like it has. When physics says no then that‘s that.

>>	Anonymous Sat Jun 15 01:10:39 2019 No.10725854 >>10724222 train it really hard and analyze the games for hidden rules/strategies that always work?

>>	Anonymous Sat Jun 15 01:14:01 2019 No.10725858 >>10725854 samefagging, how difficult would such a thing be and is it even possible to get some analytical information out of it?

>>	Anonymous Sat Jun 15 09:40:27 2019 No.10726546 >>10725651 thanks, its hard to keep up with all the engines

Anonymous Sat Jun 15 09:47:03 2019 No.10726557

>>10725854
for chess? all the early and midgame strategy is known. strategic chess is pretty much exhausted, its only competitive because people can only memories 15(??) moves of theory for chess openings and after that its a frontier of possibilities that cant be charted out. all the competition for chess engines are about finding marginal increases for better valuation system or more "thinking" within a turn (memory and plotting possible moves)

>>	Anonymous Sat Jun 15 10:52:28 2019 No.10726651 >>10722432 >there may be 10^90 positions or whatever, but you can eliminate types of positions in groups No, you can't. That's already the number of valid positions.

Anonymous Sat Jun 15 10:57:48 2019 No.10726657

>>10722548
This isn't solving the game, anon. A mathematical solution to the game would not only beat your AI in every game, it would beat every entity that existed, exists or will ever exist. That's what solving a game is, anon. Finding the sequence of moves that guarantees a win every time.

>>	Anonymous Sat Jun 15 11:02:43 2019 No.10726665 >>10722674 You should look up what a solved game is

>>	Anonymous Sat Jun 15 13:17:59 2019 No.10726866 >>10726657 or at least guarantee a draw, like checkers chess is likely not always winnable even if you play optimally

Anonymous Sat Jun 15 16:33:53 2019 No.10727286

>>10722569
Mmh, at best those transpositions can be dealt with in the very early game, but certain positions might occur in different ways, but may actually differ in how good they are.
For example, in the London system, Nf6 is rather crucial to be played after black moves his e pawn but not before, even though the positional result might be the same.
Same with tempo.

>>	Anonymous Sat Jun 15 16:37:13 2019 No.10727297 >>10722583 You pretend like people only play chess to solve it. The game is a battle of wits. Whether it is solvable or not only interests mathematicians.

>>	Anonymous Sat Jun 15 17:03:23 2019 No.10727352 Didn't AlphaGo Zero solve chess like 2 or 3 months ago?

>>	Anonymous Sat Jun 15 17:35:53 2019 No.10727415 >>10727297 it is not interesting to mathematicians. it is a finite game and is solvable. it's just a matter of computation.

>>	Anonymous Sat Jun 15 18:57:12 2019 No.10727607 File: 1.72 MB, 1912x1072, meruem.png [View same] [iqdb] [saucenao] [google] Meruem solved chess in like 2 days

Anonymous Sat Jun 15 19:59:07 2019 No.10727731

>>10727297
>You pretend like people only play chess to solve it. The game is a battle of wits.
No. I never suggested anyone plays anything solve it. All I'm saying is if you have a game (like checkers) that's solved then it can only be a "battle of wits" if you and your opponent both ignore the actual known right answers for how to play. The game itself won't be one of wits, only the game as it's played by people who maintain ignorance.
It might be very easy to maintain ignorance e.g. Just by not looking at a machine that could generate the answers for you. But that's still what you're doing. You can't say checkers is a game of strategy and creativity because there's an actual answer and the strategies you and your opponents come up with will only make any sense at all because both of you are playing with some greater than zero distance from the correct solution.

Anonymous Sat Jun 15 20:03:16 2019 No.10727740

>>10727352
>Didn't AlphaGo Zero solve chess like 2 or 3 months ago?
Absolutely not. Why would you think that? Alpha Go Zero is a program that used machine learning to become very good at chess. That has NOTHING to do with solving chess. I don't know why you people keep confusing those two topics. It probably doesn't help that the media had the same bad confusion between the program Jonathan Schaeffer wrote to play checkers against humans and the solving Jonathan Schaeffer did of checkers.

>>	Anonymous Sat Jun 15 20:08:10 2019 No.10727752 like 150 years from now

>>	Anonymous Sat Jun 15 20:17:29 2019 No.10727775 >>10727352 >>10727607 >>10727752 fuck off

>>	Anonymous Sat Jun 15 20:34:01 2019 No.10727812 I think people go about this the wrong way. It is like one of those paper mazes. You don't solve it by starting at the beginning, but by starting at the end. Chess can be solved easier if you only look at end game wins.

>>	Anonymous Sat Jun 15 20:45:33 2019 No.10727832 I solved tic tac toe :)

>>	Anonymous Sat Jun 15 20:46:59 2019 No.10727837 >>10727812 Extremely good point, solve it backwards.

>>	Anonymous Sat Jun 15 22:59:52 2019 No.10728063 >>10727812 >>10727837 That's exactly what people have been doing, it has been solved for upto 7 remaining pieces.

>>	Anonymous Sat Jun 15 23:07:31 2019 No.10728077 >>10724644 This is an example of the sentimentality that I'd referred to earlier.

>>	Anonymous Sun Jun 16 17:37:29 2019 No.10730058 >>10725793 this is delicious pasta which I'm going to save for later

>>	Anonymous Mon Jun 17 05:17:07 2019 No.10731383 >>10725793 based

>>	Anonymous Mon Jun 17 06:01:29 2019 No.10731429 >>10727740 Because you can use the insights of machine learning algorithms to solve problems, retard. It's difficult, but possible. People derive physical laws and conservation laws by analyzing weights of neural networks.

>>	Anonymous Mon Jun 17 08:05:58 2019 No.10731613 >>10722771 >you can win or lose or draw Groundbreaking

>>	Anonymous Mon Jun 17 09:22:23 2019 No.10731715 >>10722548 >A truly unbeatable AI would, by definition, have solved the game it is playing It, BY DEFINITION, would not have solve it, because that is not the definition of solving it.

>>	Anonymous Mon Jun 17 09:44:04 2019 No.10731745 Half people in this thread are really fucking stupid and ignorant about AI, computational complexity and game theory and they should all be banned from the board.

>>	Anonymous Mon Jun 17 14:15:53 2019 No.10732187 >>10731715 It's basically the definition of solving it. The problem is that we haven't made an unbeatable AI. Current AIs can be beaten if they train the AI longer.

>>	Anonymous Mon Jun 17 14:27:38 2019 No.10732227 >>10731429 Nobody solved chess you absolute mongoloid. Also you have no idea what solving means.

>>	Anonymous Mon Jun 17 14:41:32 2019 No.10732265 >>10722396 What if we try with a smaller version of it? Like a 4x4 grid with 4 pieces each, see if that can be solved, then use similar tactics to scale up?

>>	Anonymous Mon Jun 17 14:55:47 2019 No.10732301 >>10722391 Because there's a finite amount of games possible. The amount of moves is finite. Knowing every game ever you can know all possibilities and always choose the one that has the most winning moves for you.

>>	Anonymous Mon Jun 17 15:23:47 2019 No.10732369 >>10732227 Tell me what you mean by solving then you freaking imbecile.

>>	Anonymous Mon Jun 17 15:28:16 2019 No.10732375 >All this salt about what "solved" means. A solved game is a game whose outcome can be correctly predicted from any position, assuming that both players play perfectly. Prove me wrong.

>>	Anonymous Mon Jun 17 16:57:25 2019 No.10732575 >>10732369 >Thinks chess was solved >Calling ANYONE else an imbecile Never post here again.

Anonymous Mon Jun 17 17:04:12 2019 No.10732588

>>10732375
Then chess will never be solved with that definition, but that isn't even the definition of a solved game.
People in this thread don't understand that there are more valid moves (meaning, good and not stupid and leading to powerful positions) in chess than atoms in the god damn universe so it is not possible to solve regardless of computational resources or level of AI or any of the absolutely retarded shit that the people ITT are saying.
>"Just search the game tree"
>"just work backwords"
>"look at the advancements in AI over the past 10 years! AI will solve it in 50 years!"
This fucking thread man

Anonymous Mon Jun 17 18:30:21 2019 No.10732716

>>10732588
>argued in 1951 that it is not feasible for any computer to actually solve chess, since it would either need to compare some 10120 possible game variations, or have a "dictionary" denoting an optimal move for
teh number is great but not that great, it would be impossible by current standards.

but if you turn an entire planet into a computer you could easily brute force chess

>>	Anonymous Mon Jun 17 18:48:32 2019 No.10732748 >>10732575 Wow I've never met such a fucking idiot. Where have I claimed it was solved? Spoiler: I didn't. If you could learn to suck your own dick you'd finally suck at everything. Go for it.

>>	Anonymous Mon Jun 17 18:51:05 2019 No.10732753 >>10732588 You have no clue about computing whatsoever. A quantum state on a register of 150 qubits (or so) spans a computational space that's larger than there are atoms in the universe. Is this magic to you?

>>	Anonymous Mon Jun 17 18:53:14 2019 No.10732757 >>10732748 Don't reply to other people's conversations then you stupid fuck. This is what we were discussing: >>10727352

>>	Anonymous Mon Jun 17 18:55:40 2019 No.10732761 >>10732757 How is that the same as saying "chess was solved"? You're literally only looking for reasons to lash out at people. Get your shit together.

>>	Anonymous Mon Jun 17 18:59:00 2019 No.10732766 >>10732753 Cool. Now show me the algorithm that maps the solution to chess into those qubits, and do so in an amount of time that doesn't exceed the supposed heat death of the universe. I'll wait.

>>	Anonymous Mon Jun 17 18:59:56 2019 No.10732769 All these people that think you have to brute force chess to solve it.. wtf. You don't have to brute force something to solve it; we don't brute force the equation for ax+b to fit it to two different points.

>>	Anonymous Mon Jun 17 19:02:47 2019 No.10732776 >>10732769 >not brute forcing linear equations just because you can

Anonymous Mon Jun 17 19:04:15 2019 No.10732783

>>10732769
not every problem has an analytic solution
even if trim losing strategies, it will take a long time to prove the criteria for a losing strategy then reduce identical games to them, then solve the rest of possible stratgies
still uncountable lifetimes of computation

>>	Anonymous Mon Jun 17 19:06:10 2019 No.10732791 >>10732783 Most problems have analytic solutions, chess has an analytic solution. Chess isn't a complicated game. The analytic solution could probably be computed on a PC. We just don't know what it is.

>>	Anonymous Mon Jun 17 19:06:18 2019 No.10732793 >>10732588 >That's not the definition of solved >Offers no alternative That is the definition of a solved game though. Tic Tac Toe is a solved game for exactly that reason

>>	Anonymous Mon Jun 17 19:12:11 2019 No.10732810 >>10732783 Even if there's a possibility that chess doesn't have an analytic solution, nobody knows that it doesn't have one, so it's incorrect for people to assume that it would be impossible to compute.

>>	Anonymous Mon Jun 17 19:12:41 2019 No.10732811 >>10722474 he slept for 5 years you fucking retard

>>	Anonymous Mon Jun 17 19:13:58 2019 No.10732815 >>10732766 >doesn't understand the argument k den, good luck trolling

>>	Anonymous Mon Jun 17 19:14:35 2019 No.10732816 >>10732811 Okay, so he learned how to play chess in his sleep

>>	Anonymous Mon Jun 17 19:16:26 2019 No.10732825 >>10732791 did you ever think someone over the last thousand years of strategic and mathematical analysis would have found real evidence for that to be true? you have no clue what youre talking about

>>	Anonymous Mon Jun 17 19:20:27 2019 No.10732834 >>10732825 Dude, game theory isn't an old field. I'm not stupid.

>>	Anonymous Mon Jun 17 19:35:30 2019 No.10732863 >>10732834 see>>10724343 please tell me in detail how to eliminate like 99% of computation time needed to solve the game in anybody's lifetime sounds like a cop-out from someone who still doesnt know what he is talking about

Anonymous Mon Jun 17 19:42:39 2019 No.10732881

>>10732863
Not a cop-out. You're not the only person here who knows anything about game theory by far.

Aside from the ludicrous personal attack, I'm pretty sure there are methods to do it. Chess is an extremely simple game. And game theory has only been around for like 50 years, it would be utterly unreasonable and ridiculous to think that we've reached the highest analytical capabilities that anyone could reach for fully observable deterministic games.

>>	Anonymous Mon Jun 17 19:56:42 2019 No.10732905 >>10732301 What would happen if both players knew how to solve chess.

>>	Anonymous Mon Jun 17 20:00:27 2019 No.10732915 >>10732905 Either a draw every time or the first/seconds player always wins.

Anonymous Mon Jun 17 20:11:53 2019 No.10732937

>>10732881
game theory as a field has kind of dried up as far as theory progress, not that a big breakthrough can't happen
the problem is game theory is very good at being a rigorous framework for understanding and solving problems, but rarely does it provide a way to vastly simplify the problem without first cataloging all the necessary information, which is the biggest hurdle to begin with. The problem with solving chess means we need to prove the best strategy to avoid a loss for any game state, which there are too numerous of. Even to show any game state is reducable or equivalent to another solved game state, we are still stuck with the bulk of the enourmous exhaustive computation process. Even if we can solve a billion states per second, and reduce the game to half, we're stuck with billions of years of computation. That isnt even including an analysis of every possible strategy, counter strategy, and layers of prudential strategies, becuase plenty of "optimal" strategies can equal with different exploitation strategies

Anonymous Mon Jun 17 20:27:01 2019 No.10732961

>>10732937
The simple fact is that game theory can't be expected to know the solution. Game theory doesn't know if there's an analytical solution to all or only some fully observable deterministic games. It's not a very advanced field. So just because game theory doesn't know if there's an analytical solution that isn't very much evidence that one doesn't exist.

If you formulate chess as a minimax game then you can reduce the state space exponentially. That shows that you can reduce the state space without brute force. And that's only with alpha beta pruning, not other heuristics that exist but haven't been discovered.

So the argument that you have to brute force the game is out the window. We already know we don't have to brute force the game. Alpha beta pruning can solve it in exponentially less time. And that's only a minimax optimization. Chess isn't adequately described with minimax, because the pieces don't have the utility value that the minimax algorithm assumes.

Anonymous Mon Jun 17 21:14:46 2019 No.10733038

>>10732961
doesnt really matter how exponentially you can solve the game, one one hundreth of the 10^47 legal positions is still 10^45, a computation that would outlast entropy death of the universe barring some future sci fi quantum computation. I dont know what kind of heuristics that would exist that could cut it down to any feasible number of computations, at a feasible rate of computation, with any feasible amount of computers.
At the same time, chess is strategically exhausted too. There isnt some framework that lets us perfectly compute the winner or loser (or provable draw) from any position. At some point you have to accept even a simple game can't be analytically solved, even ignoring whether anybody will bother to do so, dedicate the storage space to a ginormous solution, or the feasibility of ever using a giant solution codex

>>	Anonymous Mon Jun 17 21:16:02 2019 No.10733043 >>10732961 alpha beta pruning also falls to the horizon effect in chess exactly. people already use it, but only for approximate valuation methods, no true solutions

Anonymous Mon Jun 17 22:17:38 2019 No.10733143

>>10733038
>doesnt really matter how exponentially you can solve the game

By definition that's false, if you solve it exponentially enough then you can iterate through the remaining positions computationally. There is no valid argument that we will never find a pruning method that reduces the amount of positions by the required amount since game theory is such a young field, so one wouldn't have adequate grounds to say that it would be impossible to compute a solution to chess.

Anonymous Mon Jun 17 22:24:29 2019 No.10733155

>>10733038
Okay, let me put it this way. Here's the summary of my argument: All you have to do is find enough methods that reduce the search space until it's small enough to iterate through. Game theory isn't an old field, so it hasn't found all the methods that you could use. Since we haven't found all the methods, there are more, meaning we can further reduce the search space, potentially to the point where we can solve chess computationally. Since we could potentially reduce the search space to that point, we could potentially compute a solution to chess and it would be invalid to say that it's impossible to compute the solution to chess.

Anonymous Mon Jun 17 22:29:05 2019 No.10733159

Also, you assume that every pruning operation we have and will find only reduces the positions by a factor of 100. Pruning could potentially reduce the search space by more orders of magnitude. When we discover more types of pruning, that would make it more feasible to compute a solution. >>10733038

Anonymous Mon Jun 17 22:40:47 2019 No.10733169

>>10722389
What do you mean by solved? Meaning every single move is mapped out and every single "best" choice is found? Computers maybe 15-20 years. Humans? Never. There's too many options, and there's no particular strategy that wins, it's all on a positional basis.

Anonymous Mon Jun 17 23:34:51 2019 No.10733267

>>10733143
>>10733155
my arguement is there is just too many positions and moves. you would have to prune way too many to make it feasibly computable. Too many strategies have mistakes that can be made mid and late game to completely turn a game around. There are too many outcomes for early decisions to definitively prune any early losing moves. There are early game moves that don't manifest a sizable disadvantage until way late game, like 20 moves later. Computers and players can both figure that out effectively, but solving it requires fully exploring every losable mistake to make from that position, which there are just way too many of.
Any single state except for like 8 remaining pieces has at least billions of possible outcomes to work through. Pruning is just not an effective method of solving chess, it would barely simplify the game whatsoever.
>>10733159
not at all, see>>10724220
if we could somehow solve it an equivalent time of exhaustively proving a solution for 10^10 game states per second, it would still outlast our sun in computation time. The time needed to find every possible pruning method or heuristic solution to make it a computationally feasible problem would take billions of years, even IF those methods would exist. Even if you just want to find a method for white to always force a win or draw from start regardless of trying to explore every possible game state, you still have to explore every possible exploitation and counter strategy for whatever strategy you employ. Where do you even start with that? People have been trying for a thousand years, the game is strategically exhausted as far as we can go except just plotting out every game.

Anonymous Mon Jun 17 23:41:50 2019 No.10733278

>>10733143
>>10733155
>>10733159
There are many famous gambits that show how even large piece advantages fall short to deep positional advantages mid and late game that just arent feasible to plot out because the sheer number of possible outcomes. They are famous because openings and early game is plotted out pretty well, and the experience of chess theorists get them published, but it doesnt take long to realize the number of gambits in mid game are many many orders of magnitude larger, and their outcomes need to be explored for any strategy to not fall for them. Pruning is not an effective method of chess solving because it still would take trillions of years to find all the mid game gambits and how to handle them correctly
you might know plenty about game theory and computation, but I think if you knew more about chess and the strategy you'd realize how we would need an entirely new game theoretical framework to analytically solve the game or make it computationally feasible, IF such framework even exists, and thats a very big IF

>>	Anonymous Mon Jun 17 23:43:45 2019 No.10733281 >>10733169 >single move is mapped out and every single "best" choice is found? read the thread, you clearly don't know how big the game of chess is

Anonymous Tue Jun 18 00:06:07 2019 No.10733326

>>10733278
My point isn't that minimax pruning is how you solve chess. Reducing the number of potential moves is how you solve chess. You can call it whatever you want. I call it pruning. My bet is we haven't nearly found all the reduction techniques we can. Game theory is too new to have found all the techniques. Once we find enough techniques, we will be able to analytically solve more complex games. I think the techniques are out there for solving chess analytically, we just haven't found them. But think you want. It will probably require a new game theory framework, I'm just saying there's no way we can't significantly reduce the number of moves further. And that doesn't cover quantum computers, if a civilization knew how to use them for combinatoric problems.

Just saying, it's chess. Why in the world would a simple game like chess be impossible to solve?.. That's just silly.

>>	Anonymous Tue Jun 18 00:26:54 2019 No.10733364 >>10732761 >How is that the same as saying "chess was solved"? Are you having a stroke? Please explain to me how: >>10727352 >Didn't AlphaGo Zero solve chess like 2 or 3 months ago? Doesn't mean you believe "chess was solved."

>>	Anonymous Tue Jun 18 00:51:06 2019 No.10733398 File: 78 KB, 720x720, MFW hifan.jpg [View same] [iqdb] [saucenao] [google] >>10724226 <--- huge upgrade

>>	Anonymous Tue Jun 18 02:38:45 2019 No.10733572 If a model was trained which effectively won 100% of the time against the best players and models when it played the same color/turn, would that constitute solving chess?

Anonymous Tue Jun 18 02:41:59 2019 No.10733579

>>10733572
I haven't seen a specific proof for Chess, but with certain simple games there is a way to prove the game is finite, and then it's deterministic who wins from the initial set-up. I think if someone ever "closed" chess, then you could say it's solved. I.e. From starting Chess, one player should always win, or the game ends in a draw.

>>	Anonymous Tue Jun 18 02:46:19 2019 No.10733588 >>10733572 No, it needs an actual proof. That would just be an approximation at best.

Anonymous Tue Jun 18 03:05:47 2019 No.10733605

>>10733572
>effectively won 100% of the time against the best players and models
>would that constitute solving chess?
No, that's not solving. And "effectively [winning] 100%" isn't something you could ever know regardless. What if your program wins the first quadrillion games it plays and then finally loses a game in the distant future? I don't get why people keep on asking about ML for this. The whole point of solving is calculating the actual solution, not approximating. It's like the difference between symbolic differentiation vs. numerical differentiation. One is the actual right answer, the other is an estimate.

>>	Anonymous Tue Jun 18 03:11:08 2019 No.10733609 >>10733398 really not.

>>	Anonymous Tue Jun 18 03:37:26 2019 No.10733626 >>10733572 no, models are probabilistic. to solve chess, you'd have to compute all possible remaining positions given any board state.

>>	Anonymous Tue Jun 18 03:48:04 2019 No.10733636 >>10733626 If a solution (forced-win tree) exists it is possible (likely, even) we will stumble upon it eventually, even if cannot prove it

>>	Anonymous Tue Jun 18 03:54:23 2019 No.10733639 >>10733636 >even if cannot prove it If you can't prove it you didn't solve it. >likely, even No, it's not likely at all likely a program will "stumble upon" the one absolutely perfect strategy. Where the fuck are you getting that idea from?

Anonymous Tue Jun 18 08:32:36 2019 No.10733915

>>10732753
You obviously have no clue about quantum computing or you're a disingenuous asshole. Quantum computing will make no difference to solving chess it has no relevance to the problem and just throwing the "muh quantum computing can solve anything" shows you don't know what you're talking about.

>>	Anonymous Tue Jun 18 09:30:52 2019 No.10734005 File: 25 KB, 300x291, B0612F5C-23E0-4973-99B3-AC8A0DF0C565.jpg [View same] [iqdb] [saucenao] [google] >>10722389 Get on my lvl

Anonymous Tue Jun 18 10:02:05 2019 No.10734052

>>10733326
you just dont understand the size of the problem or the strategy of chess. Pruning is how you solve the game, but it just doesnt make the game feasible to solve.
>Why in the world would a simple game like chess be impossible to solve?
why cant we analytically solve the stock market, or perfectly simulate the universe? Even random and chaotic variables aside, we would need too much information than is feasible to do so. There are too many damn possibilities, that you would need to prune the game after the first 15 moves to make it feasible, and chess strategy has demonstrated how impossible that is.
Maybe you could watch some chess games like Fischer's queen sacrifice and understand how little pruning would ever simplify chess, and try to estimate the size of the game with some combinatoric methods to understand how much pruning is necessary to make it computable

Anonymous Tue Jun 18 10:07:56 2019 No.10734056
File: 36 KB, 318x400, A134DDF6-74B3-43D9-9BCF-3B1ADEB0E207.jpg [View same] [iqdb] [saucenao] [google]

>>10734052
This is retarded, chess is fixed and a game of prefect information.
The stock market and the universe are not fixed and we don’t have prefect information
>>10733278
> it still would take trillions of years to find all the mid game gambits and how to handle them correctly
Extreme doubt.

>>	Anonymous Tue Jun 18 10:16:53 2019 No.10734064 File: 15 KB, 576x256, regimental.gif [View same] [iqdb] [saucenao] [google] >>10734005 >Playing a game AI already dominates Step up to regimental chess.

>>	Anonymous Tue Jun 18 10:24:03 2019 No.10734072 >>10734064 Ehh, shuffle chess is the GOAT prefect information game. But honestly even bridgehead more depth because you have to work with probabilities. Autists like chess because they don’t like probability

>>	Anonymous Tue Jun 18 10:25:02 2019 No.10734073 >>10734005 What level dan are you?

>>	Anonymous Tue Jun 18 10:27:23 2019 No.10734078 File: 17 KB, 268x371, Magic_the_gathering-card_back.jpg [View same] [iqdb] [saucenao] [google] >>10734064 >>10734005 you are like little babys watch this

>>	Anonymous Tue Jun 18 10:34:16 2019 No.10734080 >>10734078 >pay to win

Anonymous Tue Jun 18 10:39:36 2019 No.10734087

>>10733278
> it still would take trillions of years to find all the mid game gambits and how to handle them correctly
>>10734056
>Extreme doubt.
Not him, but each chess game has an average of 40 moves and for each move a player chooses between an average 30 possible moves. That means the LOWER bound for chess game-tree complexity is 10^120 (1,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000) possible games.
Number of atoms in the observable universe is only 10^80 in contrast.

Anonymous Tue Jun 18 10:44:26 2019 No.10734092

Everyone is confused about the word "solve". Here is the simplest definition.

Given any board position in a game of chess, and assuming both players play optimally, correctly output "Black wins", "White wins" or "Draw".

I also think there's a reasonable chance that optimal play will always lead to a Draw. If this is the case, I suspect it will be computationally easier than if white/black win with optimal play.

>>	Anonymous Tue Jun 18 12:02:14 2019 No.10734223 >>10734056 one thing is shared between the examples, and that is insurmountable computation necessary. why is it so hard to accept that some problems are too big to solve?

Anonymous Tue Jun 18 13:32:01 2019 No.10734439

>>10733639
>No, it's not likely at all likely a program will "stumble upon" the one absolutely perfect strategy. Where the fuck are you getting that idea from?
If models continue to improve in one direction to may end up with a set of weights which is near optimal. From those wights you could build out the decision tree and see if improvements could be made.

>>	Anonymous Tue Jun 18 13:39:37 2019 No.10734468 How do I into chess? I barely ever see two moves ahead and I never have an "endgame" strategy. Normally I hate the queen because half the game is spent saving her

>>	Anonymous Tue Jun 18 13:49:12 2019 No.10734489 >>10734223 Problems are big when you don't have the right algorithm. AI techniques can reduce many problems from exponential to polynomial time.

Anonymous Tue Jun 18 13:56:48 2019 No.10734507 [DELETED]

>>10734223
Problems are big when you don't have the right algorithm. AI techniques can reduce many problems from exponential to polynomial time. The number of possible positions isn't really relevant. How many times do we have to repeat that? There are more possible combinations in a 1000 element list than there are in chess, but that doesn't mean we have to try each order to sort it.

Anonymous Tue Jun 18 13:56:59 2019 No.10734508

>>10733364
Are you retarded?
You are retarded.

Same thing. You'll still deny the difference just so you can feel smug about whatever it is that's driving you. Good luck with that.

You could have simply said "no, it just..." or something instead of this monkey-speak.

Anonymous Tue Jun 18 14:00:32 2019 No.10734517
File: 47 KB, 720x720, 0F16853A9D67420CB1A39F99BD7780B2.jpg [View same] [iqdb] [saucenao] [google]

>>10733915
Fuck off you actual fucking retard. His argument was "it's impossible because there are more states than atoms in the universe" to which my reply was "states in QC require more than that on ~150 qubits registers to be completely classified is that magic to you?"
I have never said it would solve chess or help doing so you pretentious asshole.

I hope you'll have a hair in your mouth that you can't find.

Anonymous Tue Jun 18 14:12:55 2019 No.10734541
File: 5 KB, 250x140, 1515881711935s.jpg [View same] [iqdb] [saucenao] [google]

>>10734508
Asking "didn't AlphaGo Zero solve chess like 2 or 3 months ago?" means you believed it did. Which makes you completely retarded given how solving chess is one of the most famous examples of an absurdly intractable task e.g.:
>>10734087
>Each chess game has an average of 40 moves and for each move a player chooses between an average 30 possible moves. That means the LOWER bound for chess game-tree complexity is 10^120 (1,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000) possible games.
>Number of atoms in the observable universe is only 10^80 in contrast.
You also could've just looked up what solving chess means instead of shitting up the thread with your retarded idea it was already solved, something I still have no idea how you even came about belieiving in the first place. Who told you that? Where did you get that idea from?

Anonymous Tue Jun 18 14:14:05 2019 No.10734544

>>10734223
Problems are big when you don't have the right algorithm. AI techniques can reduce many problems from exponential to polynomial time. The number of possible positions isn't really relevant. How many times do we have to repeat that? There are many solved problems that have more configurations than chess.

>>	Anonymous Tue Jun 18 14:15:33 2019 No.10734550 File: 8 KB, 277x271, e8e.png [View same] [iqdb] [saucenao] [google] >>10734544 >There are many solved problems that have more configurations than chess.

>>	Anonymous Tue Jun 18 14:18:57 2019 No.10734560 >>10722389 *until

Anonymous Tue Jun 18 14:23:36 2019 No.10734569

>>10734541
Look, I'll help your underdeveloped brain understand this:
I've heard a news story about alpha zero and chess. Since chess bots already outperformed humans I thought it might have been a new breakthrough. That's why I was asking. Sorry for trying to stir up a conversation. I can tell you're not interested in that at all.
Also, as several people pointed out, game-tree complexity is not a reliable quantifier for impossibility of solving something.
I have never stated it was solved, however many times you may repeat that nonsense.

>>	Anonymous Tue Jun 18 14:25:54 2019 No.10734578 >>10734550 Lists have more permutations than chess. You can easily sort a 1000 element list, the number of permutations is much more than chess.

>>	Anonymous Tue Jun 18 14:28:17 2019 No.10734581 >>10734468 Don't move your queen out early. If you want download lichess. Doing puzzles will show you a great amount of tactics to use in game. After that learn a simple opening move for both white and black. Then have fun!

>>	Anonymous Tue Jun 18 14:39:35 2019 No.10734606 >>10734550 Lists have more permutations than chess

>>	Anonymous Tue Jun 18 14:45:27 2019 No.10734619 >>10734550 Pathfinding, list sorting, etc, etc.

Anonymous Tue Jun 18 15:22:13 2019 No.10734728

>>10722586
>calls anon dumb shit
>doesnt realize that in order to test the AI to the point where you could say you solved it, you would have to solve it
>in fact, doesnt seem to understand what proof means in mathematics

You are why democracy was a mistake.

>>	Anonymous Tue Jun 18 15:29:16 2019 No.10734758 >>10722396 Isaac Asmiov in the 50s though that genetic sequencing would be too expensive for the average person 20,000 years in the future. Today you can have it done for $1000.

Anonymous Tue Jun 18 15:30:25 2019 No.10734765

>>10734569
>but I didnt mention solving!
Except you did see
>>10732369
>>10731429
>>10727352
>>10722586
>>10722548

Your options are now play the semantic card of "I NEVER SAID VERBATIM CHESS IS SOLVED" or accept you are retarded and take the L on this one.

>>	Anonymous Tue Jun 18 15:34:49 2019 No.10734777 >>10734468 watch game plays on youtube especially chessnetwork on youtube

>>	Anonymous Tue Jun 18 15:36:47 2019 No.10734785 >>10734544 not every problem is reducable enough to solve like that though. play some damn chess and you might realize why we havent been able to reduce it any further analytically

>>	Anonymous Tue Jun 18 15:40:20 2019 No.10734801 Can I actually get better at chess or should I just give up if I'm not born Russian or Chinese?

>>	Anonymous Tue Jun 18 15:51:07 2019 No.10734839 >>10734785 Why wouldn't there be an analytical solution for a fully observable deterministic game? Doesn't make sense. It doesn't sound like you understand AI and algorithms.

Anonymous Tue Jun 18 15:58:40 2019 No.10734865
File: 118 KB, 236x219, 1486507218984.png [View same] [iqdb] [saucenao] [google]

>>10734765
>tell me what you mean by solving
Does not mean "chess is solved".
>you can use insights from machine learning to solve problems
Does not mean "chess is solved".
>didn't AlphaGo Zero solve chess some months ago?
Does not mean "chess is solved".

The other two aren't me.

Thanks for making an idiot out of yourself. I'm getting entertained by this.

>>	Anonymous Tue Jun 18 16:30:07 2019 No.10734941 >>10734865 This is the definition of “solved” that only math autists care about. For everyone else a statistical solution is sufficient.

>>	Anonymous Tue Jun 18 17:01:21 2019 No.10735003 >>10734941 No you idiot. There isn't a separate version of solving for laypeople that makes your wrong ideas right. Writing a program that's good at chess isn't solving by any definition. If you don't know what a word means don't use it.

>>	Anonymous Tue Jun 18 17:03:02 2019 No.10735006 >>10734839 methods exist to solve it if thats what you mean, but there is no reason a way to reduce it analytically to feasibly solve it necessarily exists. The game is too big and it is very unlikely we can simplify it enough

>>	Anonymous Tue Jun 18 17:04:01 2019 No.10735010 >>10734765 He's already outed himself as a brainlet who heard about Alpha Zero winning at chess and somehow thought this was solving. There's nothing else worth saying to him.

>>	Anonymous Tue Jun 18 17:08:17 2019 No.10735018 >>10735006 It just seems simplistic and egocentric to think that a game (an extremely simple game, too) we can't solve is intrinsically impossible to solve, no offense. The chances that there is no analytical solution are low (zero, in my mind).

>>	Anonymous Tue Jun 18 17:14:41 2019 No.10735033 >>10735003 >t. seething math incel every other science and engineering is satisfied with confidence intervals. only math autists waste their lives on closed solutions

>>	Anonymous Tue Jun 18 17:18:16 2019 No.10735039 >>10735033 There is nothing wrong with statistics. Your problem is you're using "solving" to refer to something that has nothing to do with solving.

>>	Anonymous Tue Jun 18 17:18:53 2019 No.10735041 >>10735033 If confidence intervals are so great then why not have confidence intervals for addition and multiplication, instead of using the closed form?

Anonymous Tue Jun 18 17:20:04 2019 No.10735044

>>10735018
Nobody said chess is "intrinsically impossible" to solve. There's already a well known proof games like chess definitely do have solutions. The issue is how much computational power do you need to get there, and for chess the game is far from simple.

Anonymous Tue Jun 18 17:21:17 2019 No.10735045

>>10735018
it can be solved, absolutely, but there is no reason to believe we will ever have the computation necessary to do so. I just don't think you know a lot about chess. We can set up all of our computers and wait for the universe to die before we analyze every game state, strategy, and counter strategy, or we can try to dedicate our civilization to trying to solve it and realize its a big waste of time and entropy will kill us first. We collectively know everything there is to know about the game, all that is left is exploring more examples of strategic situations or building more efficient computers to play the game for us. Possibly some supercomputer with enough power to simulate a universe could come along, or we could find some secret analysis that solves the game entirely with an infinitesimal comparison of computation time, but at some point I've accepted that some problems are too big to exhaustively solve, and it wouldn't even be worth trying.

>>	Anonymous Tue Jun 18 17:23:05 2019 No.10735046 >>10735044 Chess isn't complicated. It only has 6 pieces. The argument would be maybe valid if chess had tons of pieces. And if the actions for each piece varied based on the game state. Chess is extremely simple, however.

Anonymous Tue Jun 18 17:24:35 2019 No.10735050

>>10735046
>Chess isn't complicated. It only has 6 pieces. The argument would be maybe valid if chess had tons of pieces. And if the actions for each piece varied based on the game state. Chess is extremely simple, however.
Please don't pretend to be an idiot like that. It isn't funny. Get a real hobby, like maybe pick up a musical instrument or start hiking on nature trails.

Anonymous Tue Jun 18 17:24:51 2019 No.10735052

>>10735045
>We can set up all of our computers and wait for the universe to die before we analyze every game state, strategy, and counter strategy
I've gone over this many times with you. I don't think you can really understand the problem. You don't have to calculate every game state to solve chess. This conversation is a waste of time for me if you can't understand that.

>>	Anonymous Tue Jun 18 17:25:51 2019 No.10735058 >>10735050 Nothing idiotic about that. I'm sorry that you can't see that simple games have simple solutions.

>>	Anonymous Tue Jun 18 17:29:58 2019 No.10735062 >>10735010 Again, that's not what I said. Perpetuate it all you want.

>>	Anonymous Tue Jun 18 17:37:38 2019 No.10735079 >>10735052 How many game states do you need to analyze? How do you know which branches can be pruned?

>>	Anonymous Tue Jun 18 17:38:19 2019 No.10735083 File: 55 KB, 540x960, 1559049016670.jpg [View same] [iqdb] [saucenao] [google] >this thread /sci/ is literally autism

>>	Anonymous Tue Jun 18 17:41:12 2019 No.10735097 >>10735079 You analyze each game state, but that doesn't mean you iterate through each one. You can analyze a large set of game states in a small amount of operations using logic and analysis.

Anonymous Tue Jun 18 17:45:37 2019 No.10735104

>>10735097
>You analyze each game state
Have you even tried calculating how large that is? Then consider how many branches there are into and out of any game state. The complexity is enormous. You wouldn’t even be able to fit in the memory of any single computer, let alone compute it.
If finding a closed solution were practical we would have it by now.

>>	Anonymous Tue Jun 18 17:51:26 2019 No.10735115 >>10735104 This is how I know you don't have a large amount of experience with AI and algorithms. The size of the state space isn't the limiting factor for whether an algorithm can find an optimal state.

>>	Anonymous Tue Jun 18 17:52:28 2019 No.10735117 >>10735104 >>10735115 For example, even something as simple as sorting a list has a larger state space.

>>	Anonymous Tue Jun 18 17:54:38 2019 No.10735123 >>10735115 >size isn’t a limiting factor Cool fairy tale, constants matter in the real world though

>>	Anonymous Tue Jun 18 17:55:27 2019 No.10735126 >>10735104 >>10735115 The problems you're referencing are present in other algorithmic problems which have been solved, so those objections are meaningless.

>>	Anonymous Tue Jun 18 17:57:25 2019 No.10735130 >>10735123 The size isn't always a limiting factor.. You can sort an arbitrarily large list using only like a few integers of memory.

Anonymous Tue Jun 18 17:58:23 2019 No.10735132

>>10735115
Yes anon, he and everyone else who have dwelled on this ridiculously famous computationally intractable problem just don't know that the game-tree complexity doesn't matter. Also Shannon's an idiot and we should retroactively remove his number from the mathematical lexicon.
https://en.wikipedia.org/wiki/Shannon_number
Clearly you have a mature and deep understanding of this topic and aren't just being a retard despite how much it might appear the opposite is true. Carry on.

Anonymous Tue Jun 18 18:02:34 2019 No.10735139

>>10735130
Assuming you can fit an arbitrary number of integers into contiguous memory. Which you cannot, there is a real upper bound. Performance also suffers dramatically once you lose locality and costs unrelated to the algorithm (disk reads, network accesses) will outweigh the complexity of the algorithm.

>>10735126
Then why hasn’t it been done? Why haven’t you done it? As you say it should be trivial to just write a solver and let it sit there for a few decades until it spits out an answer.

>>	Anonymous Tue Jun 18 18:03:14 2019 No.10735142 >>10735132 >ignores obvious counter examples to his ramblings >thinks the existence of said counter example somehow implies the game-tree size is wrong How autistic are you?

Anonymous Tue Jun 18 18:03:18 2019 No.10735143

>>10735132
>he and everyone else who have dwelled on this ridiculously famous computationally intractable problem just don't know that the game-tree complexity doesn't matter.

Game theory is so undeveloped that nobody should even be saying that any games are impossible to solve with a straight face.

>>	Anonymous Tue Jun 18 18:04:55 2019 No.10735146 >>10735139 You're talking about upper bounds that are in a different category than the number of game positions.

Anonymous Tue Jun 18 18:07:08 2019 No.10735149

>>10735046
do some fucking research before you spout your stupid opinions
>>10735052
>>10735115
do you think you can just make some algorithm to solve the traveling salesman problems for every star in the universe within a millenium? do you not realize there are some problems we cant just reduce into a feasible problem? you think you can just prune branches of a game tree without realizing you cant effectively do that with chess because you know jack shit about chess strategy. every "bad" move however you practically define it has winnable outcomes that need to be explored before you prune the branch, you cant just expect some magical method to exist that solves it in feasible time to fix your cognitive dissonance

>>	Anonymous Tue Jun 18 18:08:19 2019 No.10735151 >>10732915 Commonly accepted that if both players play perfectly White wins.

>>	Anonymous Tue Jun 18 18:09:50 2019 No.10735155 >>10735143 sure who knows, we can't prove future math won't include some mx+b method to solve any problem ever in a small amount of time maybe we can have FTL and teleporting and live forever too

>>	Anonymous Tue Jun 18 18:10:00 2019 No.10735156 >>10735149 That argument is only based on the premise that chess is NP-hard. I don't think it's been shown that it's NP-hard.

>>	Anonymous Tue Jun 18 18:10:54 2019 No.10735157 >>10735151 the majority of high level games by both master and computers end in draw to my knowledge. source?

>>	Anonymous Tue Jun 18 18:13:28 2019 No.10735162 >>10735149 Also, chess is nowhere near as complicated as a traveling salesman problem for every star in the universe.

Anonymous Tue Jun 18 18:13:44 2019 No.10735163

>>10735156
chess has been strategically studied by professionals, mathematicians, and computers over the last thousand years. we have come as close to showing it is NP-hard as we can without proving it. even proving it might be just as hard as solving it
people who study this and not stray /sci/ anons have just accepted a solution to the game is impractical

>>	Anonymous Tue Jun 18 18:14:46 2019 No.10735165 >>10735162 really? why not?

>>	Anonymous Tue Jun 18 18:16:32 2019 No.10735170 >>10735155 Yeah but chess isn't complex, so it makes sense that it has a simple solution. If chess is NP hard then maybe it can only be solved in exponential time, but even those problems can be solved if they're small enough.

Anonymous Tue Jun 18 18:17:25 2019 No.10735172

>>10735146
The number of total configurations is 10^120, which is greater than the number of estimated atoms in the universe 10^78.
The lower bound is estimated to be 10^50 unique boards, which is an outrageous number that is orders of magnitude larger than what you could possibly address in any general computer.

>>	Anonymous Tue Jun 18 18:17:43 2019 No.10735174 >>10735163 >uses rigorous definition when it supports his ramblings >disallows rigorous definitions when it supports his ramblings Amusing.

Anonymous Tue Jun 18 18:21:07 2019 No.10735178

>>10722400
Sure, with today's equipment chess is unsolvable. But just wait until computation tools like quantum processors become widespread. It has already been preditcted that modern day RSA encryption will be broken by this equipment, infact methods like Shor's algorithm has already been proven to work, so I think the same will happen to chess. I suppose we'll just have to wait and see though.

>>	Anonymous Tue Jun 18 18:21:29 2019 No.10735180 >>10735170 why isn't it complex? do you know any chess strategy? are you skeptical that it could be NP-hard?

>>	Anonymous Tue Jun 18 18:21:35 2019 No.10735181 >>10735165 All else equal, there are more stars than there are chess pieces. I guess you could say that one chess piece equals a billion stars but that would only be speculation.

>>	Anonymous Tue Jun 18 18:22:55 2019 No.10735183 File: 115 KB, 1285x1015, 1464994547318.jpg [View same] [iqdb] [saucenao] [google] >>10735170 >chess isn't complex

>>	Anonymous Tue Jun 18 18:23:34 2019 No.10735185 >>10735174 I'd love to hear whatever rigorous definition I'm using or ignoring

>>	Anonymous Tue Jun 18 18:25:58 2019 No.10735187 >>10735181 bruh... there are 10^120 legal board positions and an estimated 10^20 stars in the universe thats ignoring strategy and connecting different board states together

>>	Anonymous Tue Jun 18 18:26:59 2019 No.10735192 >>10735172 >Implying you would iterate through each position

>>	Anonymous Tue Jun 18 18:31:11 2019 No.10735200 File: 115 KB, 681x384, 80EE042D-AAEE-47B3-AF7D-95E601CEABB7.jpg [View same] [iqdb] [saucenao] [google] >>10735192 I solve math problems through divine revelation too

Anonymous Tue Jun 18 18:32:05 2019 No.10735201
File: 49 KB, 498x573, helper.jpg [View same] [iqdb] [saucenao] [google]

>>10735187
I like how he keeps on insisting board states aren't reliable to go off of for determining complexity yet somehow the number of distinct pieces is.
>Chess is only 6 pieces, easy problem ;)

>>	Anonymous Tue Jun 18 18:32:26 2019 No.10735202 >>10735200 You don't solve math problems by iterating, you do it with algorithms. This conversation is dumb.

>>	Anonymous Tue Jun 18 18:32:59 2019 No.10735204 >>10735187 even still there are about 10^40 "good" board states with no proof that the set includes the result of all optimal play >>10735192 >just come up with some BS to ignore about 10^25 of the 10^40 good board states too easy

>>	Anonymous Tue Jun 18 18:34:20 2019 No.10735207 >>10735202 What algorithm? Hand-wave traversal?

>>	Anonymous Tue Jun 18 18:35:33 2019 No.10735211 >>10735201 just PRUNE the GAME TREE its all so EASY >>10735202 just use an ALGORITHM. it HAS TO EXIST or I'd lose an internet arguement! its OUT THERE we just don't know WHERE! game theory is a YOUNG FIELD

>>	Anonymous Tue Jun 18 18:37:15 2019 No.10735214 >>10735207 >All math problems are solved by counting up from 1 Top kek

>>	Anonymous Tue Jun 18 18:38:22 2019 No.10735216 >>10735211 >Game theory isn't a young field

>>	Anonymous Tue Jun 18 18:47:11 2019 No.10735238 >>10735216 being young doesnt guarantee any future progress will or could be made, nor it could apply to this specific game

>>	Anonymous Tue Jun 18 18:50:22 2019 No.10735246 >>10735238 If no progress could be made then AlphaZero wouldn't have been better than any other systems. Anyway, I'll just stop debating this.

>>	Anonymous Tue Jun 18 18:53:34 2019 No.10735251 >>10735246 >I can’t admit I’m wrong so I’ll run away AlphaZero isn’t solving anything, it’s producing a model which plays the game very well but it makes not attempt to solve anything. It is inherently statistical.

>>	Anonymous Tue Jun 18 18:55:23 2019 No.10735255 >>10735246 lmao alpha zero? is this dumb bait?

>>	Anonymous Tue Jun 18 18:57:54 2019 No.10735262 >>10735185 >we have come as close to showing it is NP-hard as we can without proving it

Anonymous Tue Jun 18 18:58:47 2019 No.10735265

>>10735251
I admit that I'm maybe wrong, but you guys are the worst to debate with. It's obviously intellectually dishonest to say that game theory couldn't be improved if AlphaZero is better than the systems created by game theory. If you really can't see that then maybe you're just retarded. Idk.

>>	Anonymous Tue Jun 18 19:02:00 2019 No.10735277 >>10735265 >systems created by game theory lmao you are way out of your element

>>	Anonymous Tue Jun 18 19:03:02 2019 No.10735280 >>10735262 yep

>>	Anonymous Tue Jun 18 19:03:28 2019 No.10735281 >>10735265 This. It's not worth wasting time with these brainlets.

>>	Anonymous Tue Jun 18 19:07:20 2019 No.10735289 This thread is autistic chess fans trying to cope with being btfo'd. The state of /sci/. Probably just samefagging even.

Anonymous Tue Jun 18 19:09:47 2019 No.10735296

>>10735265
chess strategy is too complex for game theory to be effective and doesn't simplify the problem of solving it
>>10735281
feel free to find outside source that agrees with your magic pruning algorithm method
>>10735289
its so easy even though nobody else thought of it

>>	Anonymous Tue Jun 18 19:14:06 2019 No.10735309 >>10735296 Game theory is effective to some extent and AlphaZero is more effective. All I said was that Game theory could be improved, not that game theory could solve chess.

>>	Anonymous Tue Jun 18 19:17:41 2019 No.10735315 >>10735309 see>>10735238 game theory CAN solve chess, but the computation is too large, and we don't know that it can be simplified

>>	Anonymous Tue Jun 18 19:20:28 2019 No.10735321 >>10735315 That isn't relevant to what I said. I didn't say "AlphaZero is better than game theory, therefore game theory can solve chess."

>>	Anonymous Tue Jun 18 19:33:44 2019 No.10735337 >>10735321 He doesn't understand logically structured sentences. I suggest abandoning and letting this die.

>>	Anonymous Tue Jun 18 19:35:58 2019 No.10735339 >>10735296 That's not me. There are several people calling you out in deliberately misunderstanding things and starting them incorrectly. You might want to work on your attitude problem. You'll find friends I'm sure

>>	Anonymous Tue Jun 18 21:18:31 2019 No.10735551 seething tards ran out of arguements guess whatever you read in a textbook doesn't automatically apply to things without critical thinking

>>	Anonymous Tue Jun 18 21:39:05 2019 No.10735604 >>10733281 It's like 10^40 or something. If you think computing power won't be 10^40 within the next 20 years, maybe you should look at computers from 1999.

Anonymous Tue Jun 18 21:47:24 2019 No.10735626

>>10735604
to get through 10^40 in a ten year period, we would have to have the compatibility of 3x10^31 per second. 10^20 is roughly the amount of stars in the universe
call me skeptical
one estimate I read of quantum computation is 10^8 times faster than current machines, not even scratching the surface of the speed needed

>>	Anonymous Tue Jun 18 22:02:03 2019 No.10735645 >>10735626 We currently have 10^46 in a desktop computer, that can't be right.

>>	Anonymous Tue Jun 18 22:05:28 2019 No.10735648 >>10735645 10x46 what? Even if it were atoms you'd be wrong

>>	Anonymous Tue Jun 18 22:14:16 2019 No.10735663 >>10735604 Moore's Law is dead now and there is a hard limit on how small and fast you can make modern processors. >>10735626 Quantum computers aren't a given either because it is still unclear if they will scale to the sizes necessary to be useful.

Anonymous Tue Jun 18 22:15:22 2019 No.10735666

>>10735604
Even Moore doesn't think Moore's Law is going to continue applying for much longer.
https://en.wikipedia.org/wiki/Moore%27s_law
>Most semiconductor industry forecasters, including Gordon Moore,[96] expect Moore's law will end by around 2025.[97][98][99]

Anonymous Wed Jun 19 04:14:27 2019 No.10736298 [DELETED]

A thing people don't seem to realize that reducing the number of moves that need to besearched also reduces the search space exponentially.

the Shannon number comes from assuming 1000 pairs of moves (cca 31.6 per ply) and 40 moves, giving the total complexity of 10^120. If we could remove just move on average for each player, the complexity would drop to. 7.65*10^118.

So even relatively insignificant advances in chess theory could result in dramatic decresates in the number of calculations needed.

Anonymous Wed Jun 19 04:15:44 2019 No.10736299

A thing people don't seem to realize that reducing the number of moves that need to besearched also reduces the search space exponentially.

the Shannon number comes from assuming 1000 pairs of moves (cca 31.6 per ply) and 40 moves, giving the total complexity of 10^120. If we could remove just move on average for each player, the complexity would drop to. 7.65*10^118, reducing the number of calculations more than tenfold.

So even relatively insignificant advances in chess theory could result in dramatic decreases in the number of calculations needed.

>>	Anonymous Wed Jun 19 04:59:08 2019 No.10736403 >>10735626 Quantum computers aren't faster. Also, your stupidity level is 10^50 while there are only 10^20 stars in the universe so how can you even exist?

>>	Anonymous Wed Jun 19 05:05:11 2019 No.10736424 >>10722586 >why yes i am a black man how could you tell

>>	Anonymous Wed Jun 19 05:39:07 2019 No.10736476 >>10734865 Imagine being so dumb you do exactly what I predicted in advance.

>>	Anonymous Wed Jun 19 09:26:18 2019 No.10736960 >>10736476 Lie as often as you wish. Your dignity cannot be restored.

Anonymous Wed Jun 19 09:34:45 2019 No.10736985

>>10736299
we have an estimated 10^40 "good" moves without good criteria of what seperates them from "bad" moves, and even that is too large of a number to work with. It would be great to further reduce the number of moves, but chess theory is pretty much exhausted already, and we really can't come up with a clear way to ignore certain moves because you could only ignore them if they strictly lose against optimal play, which we haven't found to begin with. Chess theory is not rigorous enough, its only a collection of things that generally work very very well, and the fact that some engines ignore some of it shows we can't even rely on the advances we already have.

Anonymous Wed Jun 19 12:15:03 2019 No.10737346

>>10722536
It exists and is at this point irrelevant. If you know what the best moves are, it doesnt matter if you are playing against a neurotic depressive emo kid or against the buddha. We have developed algorithms for finding out the right moves in an exceedingly efficient manner.

Anonymous Wed Jun 19 13:07:49 2019 No.10737489

>>10735018
Very simple rulesets or axioms or initial premises can explode in complexity. There is nothing about the simplicity of chess, or any problem, that makes it simple to solve.
The majority of mathematical conjectures are incredibly simple to state formally, but very very difficult to solve.

>>	Anonymous Wed Jun 19 13:08:54 2019 No.10737492 >>10735033 And thats why scientists and engineers are fucking retards, you seething physishit incel cuck retard.

>>	Anonymous Wed Jun 19 13:10:22 2019 No.10737495 >>10735058 Because simple games DONT have simple solutions you fucking pseudo intellectual retard

>>	Anonymous Wed Jun 19 13:27:24 2019 No.10737527 >>10737492 Only seething here is in your comments. >>10737346 I'm convinced that it must be possible to extract some information about a general solution out of efficient approximations.

Advanced search
Text to find
Subject [?]Search by post subject. Leave empty for any.
Username [?]Search for user name. Leave empty for any user name.
Tripcode [?]Search for tripcode. Leave empty for any.
Email [?]Search by email. Leave empty for any.
Filename [?]Search by image filename. Leave empty for any.
From Date [?]Enter what date to start searching from. Format is YYYY-MM-DD
To Date [?]Enter what date to start searching until. Format is YYYY-MM-DD
Image hash
Search in	All Posts OPs Only
Deleted posts	Show all posts Show only deleted posts Only show non-deleted posts
Internal posts	Show all posts Show only internal posts Show only archived posts
Order	New posts first Old posts first
Capcode	All Posts Only by Users Only by Mods Only by Admins Only by Developers
Results	Posts Threads
Action	[ Simple ]

/sci/ - Science & Math