Will computers ever solve chess?

Garry_Fish52

That's not what I mean here; Brute force means Stockfish already knows every single possible position that can ever come up in chess.

ilovesmetuna

how do you know ? and are its evaluations faultless ? like myself ?

IM_Serious

FACT

SantaCruz2017

i am sure AlphaZero can help with chess being solved!

wars64

Total starting positions: 5,040 x 6,561 = 33,067,440

Good luck solving that.

LOL

https://wars64.blogspot.com/

null

Garry_Fish52

Computers can analyze over 8 million positions/second . happy.png

Garry_Fish52

(we can only analyze up to 5 positions/second)

s23bog

How does it analyze them?  Is it the correct way, or just a way good enough to compete?

Garry_Fish52

It searches for all the elements that make a position winning (material, seeing moves ahead, position, etc.) in every single position, then it determines the best possible way to get to the best possible position.

s23bog

Why do you speak in absolutes like every and best?  You know those are exaggerations, at best.

s23bog

They sound more like flat out lies, created to boast.

wars64

When you are actually writing code for an engine it is both scary and laughable when you read what the "experts" have to say about it.

Garry_Fish52

They are facts.

s23bog

No.  No, they aren't.  They are advertising claims.  Which is odd because Stockfish is supposed to be free.  But I guess someone has to pay the bills to keep the lights on.

wars64

Not that I actually read any of this, I just come here occasionally, quick skim then a good laugh.

SantaCruz2017
wars64 wrote:

Total starting positions: 5,040 x 6,561 = 33,067,440

Good luck solving that.

LOL

https://wars64.blogspot.com/

 

You can solve each individual opening by name under the ECO codes.

wars64

You all talk about alpha whatever quite a lot, but do you understand?

Take a traditional chess engine, lets say stockfish.

It is always minimax, no matter the fancy words: negamax, alpha beta pruning, iterative deepening, whatever.

Now remove opening book.

Now remove endgame tablebases.

Now eliminate the game horizon by playing all training games to completion.

Now eliminate minimax and all associated evaluation functions and use calculus and back propagation instead.

Boom, you just eliminated human error and human assumptions which is why "AI" engines seem to play more natural chess.