[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 6 KB, 320x320, chess.gif [View same] [iqdb] [saucenao] [google]
3133013 No.3133013 [Reply] [Original]

Are computers unbeatable in chess or is there a grandmaster that can defeat them?

>> No.3133019

How can computers be unbeatable? It's not like chess has a formula to win every time.

>> No.3133025

Only if the programmers are unbeatable.

>> No.3133031

>>3133019
I don't know, that's why I'm asking.

>> No.3133034

>How can computers be unbeatable? It's not like chess has a formula to win every time.
We are actually unsure whether chess is "solvable", or more accurately, whether white or black can consistently win (through playing perfectly).

http://en.wikipedia.org/wiki/Solved_game

>> No.3133036

No computers are unbeatable. They're only as good as the team which designed the algorithms which they run on, and there are always ingenious moves which the computer isn't always going to see.

However, that considered, computers regularly win sets of games against grandmasters.

>> No.3133050

Depends on the level.
In the highest level computers are unbeatable.

http://en.wikipedia.org/wiki/Deep_Blue_%28chess_computer%29
After this computers should be able to beat every human player and this happened in 1997

>> No.3133051
File: 11 KB, 320x320, 1.png [View same] [iqdb] [saucenao] [google]
3133051

Your move nigger

>> No.3133056

I'm sure it is possible to determine if you can win 100% of the time.

But you'd have to go through every single combination of moves which no one has probably bothered to do

>> No.3133063

>>3133051
http://lichess.org/robdabmi

>> No.3133074
File: 23 KB, 508x406, comeatmebro.jpg [View same] [iqdb] [saucenao] [google]
3133074

http://lichess.org/5x88fj6h

>> No.3133112
File: 59 KB, 393x393, 1297709924062.jpg [View same] [iqdb] [saucenao] [google]
3133112

http://lichess.org/fjsfi7q3

>> No.3133127
File: 24 KB, 288x374, Trap.jpg [View same] [iqdb] [saucenao] [google]
3133127

>>3133074
knew you couldnt fucking resist going for the knight in the corner.
rook to E8 = checkmate, except you resigned.

>> No.3133131
File: 287 KB, 576x576, 1284896737273.jpg [View same] [iqdb] [saucenao] [google]
3133131

http://lichess.org/friend/6mxi79bby8ul

>> No.3133139

>>3133127
I hate timed games.
Especially when I'm stoned.
:)

>> No.3133144

>>3133139
That being said, I challenge someone to a non-timed game: http://lichess.org/vao1dz1g

>> No.3133155

>mfw I wanna learn chess strategy but also have millions of books on my list, exams to revise and television shows to watch

The last is obviously of least importance, but so much easier

>> No.3133159

>>3133131
y u quit, eh?

>> No.3133163

http://lichess.org/6tidntyh

Non-timed game, not too serious

>> No.3133284

>>3133163
my first loss of the day. well played, dear opponent.

>> No.3133347

>>3133284
I'd like to play

>> No.3133350

>>3133347
then copypaste this link into your address bar.
http://lichess.org/9xwif657

>> No.3133384

>>3133159
lol, needed 2 moves to finally kill that pawn? You suck.

>> No.3133394

Assuming you made a modern day equivalent of deep blue (essentially a brute force machine) you would never be able to beat it, at best you would be able to force a draw.

>> No.3133396

>>3133394
I'm good at forcing draws

>> No.3133416

http://lichess.org/j1fqg1ls

>> No.3133554

>>3133013
Rybka is so good these days, that they challenge GMs to odds matches... and still win.

And now there are free programs (Houdini, Stockfish) that are even better than Rybka.

You heard that there was a program running on a cell phone that won a major tournament, right? A cell phone!

Our time as the smartest beings on the planet is running out fast...

>> No.3133641

http://lichess.org/friend/6bmzfto8rxc2
im up for a round

>> No.3133693

Honestly I think that playing against a computer should be more like a rally than a race. Imagine competing in a foot race with a machine that can output 10x the strength, has infinitely more endurance than you fears no enviormental change, etc.

Obviously the thing is going to win. At least eventually... Sure the first models might have a few kinks here and there but eventually scientists will work out those kinks and create something that is, in all respects, better than you.

I mean it's cute to see a well renowned chess player being beat by a machine, it shouldn't become a standard. A human should be judged by the amount of moves you make and the amount of risks you make the computer take. THEN that score is compared to other HUMAN opponents. "Computer won" should never take precedence over your score.

>> No.3133720

>>3133554
who made the cell phone and the program dickhead

>> No.3133764

>>3133720
I'd tell you if you asked nicely. Unless "dickhead" is a term of affection in your family.

>> No.3133784

>>3133764
fuck you dickhead

>> No.3133800

Computers are only unbeatable at chess if there is an unbeatable strategy in chess.

>> No.3133796

http://lichess.org/vur6h0mc

>> No.3135040
File: 52 KB, 809x536, defeated lichess AI level 1.jpg [View same] [iqdb] [saucenao] [google]
3135040

It was bloody as hell but the joy of watching the computer struggle until the very end and then work inside hopelessness was immeasurable and captivating.

>> No.3135047

>>3133034
Chess should be solvable.
Whether we can find an algorithm which does so is another matter completely.

>> No.3135067

play

http://lichess.org/xiag1so0

>> No.3135075

>>3133554
What? Chess AI is not intelligent, and I don't here people saying Kasparov is the smartest man in the world.

Computers don't play perfect, there is a GM (or IM) that regulary beats computers with the aid of a chess engine.

And here's Nakamura humiliating Rybka
http://www.chessgames.com/perl/chessgame?gid=1497429

>> No.3135118

>>3135075
This game is completely silly. Neat ending though.

>> No.3135126
File: 39 KB, 448x293, 1302136952005.jpg [View same] [iqdb] [saucenao] [google]
3135126

>>3135040
> A.I. level 1 (of 8)
> watching computer struggle

>> No.3135205

What computers do is to simulate alot of steps forward and decide based upon that which move to take. It takes alot of memory to make a deep tree (alot of steps) but RAM is cheap nowadays. They could look ahead alot futher than any human with every piece.

So (possibly) yes, to OP's question. But no if you only have say one or two seconds to make your move. Then the computer could make mistakes because it does not have time to do every possible move.

>> No.3135236

>>3135126
I will encourage you to take on the challenge and share your success by posting the results.

>> No.3135245

Come at me

http://lichess.org/z2eytw2j

>> No.3135253

There's only a finite amount of moves in chess. Thus, someday scientists will have calculated all possible combinations of moves.

From that, a computer powerful enough can just pick any move from a pool of moves that has the highest possibility of leading to a victory/not losing a figure/leading the game to counter a move from its opponent.

So yes, computers will be pretty much unbeatable soon.

>> No.3135267
File: 59 KB, 500x306, jugblob.jpg [View same] [iqdb] [saucenao] [google]
3135267

>>3135253
what happens when it plays another unbeatable computer?

>> No.3135275

>>3135267
Tic tac toe on a grand scale

>> No.3135281
File: 103 KB, 410x400, 1299702393512.jpg [View same] [iqdb] [saucenao] [google]
3135281

>>3135253
> calculated all possible combinations of moves
> orders of magnitude more possibilities in chess than atoms in the universe

>> No.3135303

>>3135267
Solved non-luck based games have only one result if both play perfectly.

It could be that chess is a game where white wins every time. This means that if both play perfectly the game ends up in white winning every time.
Now this ultimate solution is still unkown, it might be that perfect game ends up un draw or black winning.

The thing is still the same. There is only one reasonable game to be taken once the game is solved.

>> No.3135305

>>3135267

That depends, if chess is solved and it turns out white can always win, then the computer playing white, and the same for black can always win or can always stalemate.

>> No.3135333

>>3135236
Argh, it really isn't that simple.

That, and I suck.

>> No.3135362

>>3135303
>>3135305
http://en.wikipedia.org/wiki/Game_theory#Perfect_information_and_imperfect_information

So basically you'd need to solve for there being any nash equilibriums which very often end up with there being irrational things and outcomes i.e. if the computer was aware that would not be able to achieve anything greater than a stalemate it might not play the game in order to do something which better utilizes it's time and processing power, aside from that particular outcome computers do not subscribe to reason.

>> No.3135381

>>3135253
Except there are too many combination. I think a few end games have be completely solved up to 6 pieces. That means if you have 5 pieces left, the computer can see every single solution possibles, so if the computer can win, it will win. But beyond a few pieces, the calculations are too large to be don't.

>> No.3135382

>>3135253
>Thus, someday scientists will have calculated all possible combinations of moves.
oh shit nigger what are you doing

>> No.3135397

>>3135253
>Thus, someday scientists will have calculated all possible combinations of moves.
We know pi to less digits than number of moves possible in a game of chess, and that's had a few thousand years head start
TL;DR We will never calculate all possible combinations

>> No.3135420

>>3135397
>never

hurr

>> No.3135428

What we need is for some random computer scientist or mathematician to come up with a completeness proof for chess (much like the one we have for checkers). Only then will we be able to make machines that will constantly win.

On an unrelated note, there are chess programs which are able to beat grandmasters from time to time, but usually only when you pit them against each other for a series of games. This comes from the simple fact that a grandmaster gets fatigued while the computer doesn't (in fact there is a good chance the computer only gets better and better the longer you play if it uses learning methods)

>> No.3135444

>>3135420
Well there are about 10^(10^50) different combinations of moves in a game under 40 moves, so i'd say that's probably never

>> No.3135453

>>3135444
Eliminate the useless by going in the other direction? Start with the lowest number of pieces => move upwards.

>> No.3135454

>>3135075
That's a blitz game, very different from the kind of match Kasparov played. It's almost impossible for a human player to win a regular match against a modern dedicated chess computer.

>> No.3135500

>>3135454
OK, now that I've actually watched the game.. apparently the program doesn't take into account that even in the face of such a big material advantage, accepting the draw can be the best option. Well played by Nakamura to exploit this flaw, but this should be fixable.

>> No.3135517

anyone fancy a go?

http://lichess.org/xmpy0kn5

>> No.3135533

>>3133554

Bitches don't know bout my pattern recognition.