Will computers ever solve chess?
Sort:
Does it matter? Would having a program that can do as the question posed, be able to teach a person to do the same? It would be interesting to watch and learn from those 100 matches. Regardless of weather or not computers improve to that point, enjoy what we have gotten thus far, chess ( although I am probably just an average skilled player) has got to be one of the best games ever invented and the computers today seem way beyond my skill on the harder settings.
tygxc wrote:
#6878
It is clear that number of possible chess games >> number of possible chess positions
10^120 >> 10^46
However
upper bound > estimate > lower bound
number of sensible games << number of possible games
number of sensible positions << number of possible positions
10^40 sensible games >> 10^20 sensible positions
You are most wellcome to present your own better video or paper of how many sensible positions there are in your estimated opinion.
"10^40 sensible games >> 10^20 sensible positions"
10^40, 10^45.88, and 10^46.7 all refer to positions, not games. So, *again*, this statement is meaningless. Just admit 10^20 is a number you pulled out of nowhere.
10^40 viable/evaluation-worthy positions is a very conservative estimate. It's not a hard estimate to figure out. There's 10^46.7 unique positions possible, and if only 1 out of every million positions is viable, that's still 10^40 right there.
For your 10^20 claim to stand up, only 1 in every 100,000,000,000,000,000,000,000,000 (100 septillion) chess positions would be considered "sensible". I think we can all have a good laugh pondering that notion, and then move on to some more reasonable discussion.
#6880
It matters. Many players quit checkers after it was solved to be a draw.
Jul 11, 2021
0
#6864
It's positions that count for a precise strategy. It doesn't generally matter how you got to a position if it is your list of positions that mate in 75 (or whatever). Likewise in the openings, the route to a position is of no interest once you get there.
Jul 11, 2021
0
#6865
I was only joking not that any one even saw my comment but quantum computers are really only good for simulating quantum states thats what i heard from leonard susskind in a video discussing what theyll be most useful for. So states where particles are superimposed and entangled. Not very useful for chess id imagine. Though i do kind of wonder, if theyre superimposed and entangled bits they should SEEM more efficient but as someone stated before in one of these, you can literally just use another super computer along side it. idk https://quantumcomputing.stackexchange.com/questions/5823/will-quantum-computers-be-able-to-solve-the-game-of-chess
Thanks, I pulled an interesting paper from that StackExchange thread:
https://www.cs.virginia.edu/~robins/The_Limits_of_Quantum_Computers.pdf
It's an older article but describes the limits of quantum computing better than most articles I have seen.
tygxc έγραψε:
#6880
It matters. Many players quit checkers after it was solved to be a draw.
The competitive checkers (100-square board) isn't solved yet, and quitting a game because the computer found a solution which noone will ever be able to memorize is ridiculous.
Jul 11, 2021
0
#6868
How are they going to have a 32-piece tablebase ready in only 5 years if they are having so much trouble completing the 8-piece version?
#6886
International draughts on a 10x10 board has not yet been solved, because nobody took the effort to do so. It is also conjectured to be a draw. After 8x8 checkers was solved also draughts on a 10x10 board took a dip in competitive play. Many strong draughts players flocked to chess or go.
#6888
To solve chess no 32 men table base is necessary. It is enough to calculate from the initial position towards a reasonably large table base. That is how checkers was solved: from the initial position towards a 10 men table base.
Jul 28, 2021
0
#6870
tygxc wrote:
#6886
International draughts on a 10x10 board has not yet been solved, because nobody took the effort to do so. It is also conjectured to be a draw. After 8x8 checkers was solved also draughts on a 10x10 board took a dip in competitive play. Many strong draughts players flocked to chess or go.
#6888
To solve chess no 32 men table base is necessary. It is enough to calculate from the initial position towards a reasonably large table base. That is how checkers was solved: from the initial position towards a 10 men table base.
Easy to say, but exchanging a substantial amount of material is non-trivial, and you are starting with 32 pieces. It would be a lot easier (if impossible in the forseeable future) to make a 16 piece tablebase than to generate an opening book all of whose leaf nodes have no more than 16 pieces, because the former is a much more complex problem.
Of course chess is much more complex than checkers.
However, in TCEC games the engines hit their 7 men table base around move 10.
There is already work in progress on an 8 men table base.
So a faster computer may hit an 8 men or 9 men or 10 men table base from the initial position.
It does not even need to be a full table base: illegal positions, non sensible positions with several promotions, and unbalanced positions with large army versus small army may be pruned.
The table base only needs to contain drawn positions.
Quantum computers do in 200 s what a supercomputer does in 10,000 years.
tygxc wrote:
Of course chess is much more complex than checkers.
However, in TCEC games the engines hit their 7 men table base around move 10.
There is already work in progress on an 8 men table base.
So a faster computer may hit an 8 men or 9 men or 10 men table base from the initial position.
It does not even need to be a full table base: illegal positions, non sensible positions with several promotions, and unbalanced positions with large army versus small army may be pruned.
The table base only needs to contain drawn positions.
Quantum computers do in 200 s what a supercomputer does in 10,000 years.
Your last sentence has been debunked many times already.
You are also being misleading when you say engines in TCEC "hit" their tablebase around move 10. They might touch the tablebase in some very top lines they are calculating, but that is miniscule. Especially when you are positing that this is an indicator chess will be solved 
.
Nothing debunked at all
https://www.docdroid.net/h9oBikj/quantum-supremacy-using-a-programmable-superconducting-processor-pdf#page=5
Look at TCEC
https://tcec-chess.com/#div=sf&game=42&season=21
move 10: 1500 table base hits
move 20: 2.5 million table base hits
move 45: 100 million table base hits
Think about it, even in the first move, there are a total of 400 diferent board combinations, after 2 moves its 197,742, and by the third it's over 121 million! And the number just increases as you go on, so its impractical to think you can 'solve' chess. Because each turn has so many combinations, and to compute that many combinations would take a lot of time, space, and energy, all spent on what? Solving a game.
In all, its possible, just not practical.
Quantum computers do in 200 s what a supercomputer does in 10,000 years.
As much ignorance as possible stored in a short sentence.
And the TCEC games, where the engines are forced to play fixed opening poositions with limited time are completely useless to reach any sort of conclusion.
Jul 28, 2021
0
#6876
tygxc wrote:
#6880
It matters. Many players quit checkers after it was solved to be a draw.
Thats just stupid. Quitting a game because a computer did something no human will be able to. That makes as much sense as beginners here complaining about how engines, and opening prep have ruined chess. Nothing in chess has changed for the 99.99% of us...no matter how much you want to think it has impacted your chess.
Yes.
tygxc wrote:
Nothing debunked at all
https://www.docdroid.net/h9oBikj/quantum-supremacy-using-a-programmable-superconducting-processor-pdf#page=5
Look at TCEC
https://tcec-chess.com/#div=sf&game=42&season=21
move 10: 1500 table base hits
move 20: 2.5 million table base hits
move 45: 100 million table base hits
Thanks for proving my point.
1500 hits...geez.
#6896
"the quantum processor takes 200 seconds, while an equalfidelity classical sampling would take 10,000 years on 1M core" That is what the quoted scientific paper says. The quantum computer is faster and not just a factor 2 or 10, but 10^11.
The fact that the TCEC engines with limited time already hit the 7 men table base 1500 times by move 10 indicates that a faster quantum computer with more time can hit a table base with more than 7 men consistently from the initial position.
#6897
Checkers and even draughts players have flocked to chess and go after checkers was solved. Also competitive play of Nine Men's Morris and Connect Four has stopped after these games were solved. Capablanca already proposed his Capablanca chess and Fischer proposed Fischer Random aka Chess960.
tygxc wrote:
#6896
"the quantum processor takes 200 seconds, while an equalfidelity classical sampling would take 10,000 years on 1M core" That is what the quoted scientific paper says. The quantum computer is faster and not just a factor 2 or 10, but 10^11.
The fact that the TCEC engines with limited time already hit the 7 men table base 1500 times by move 10 indicates that a faster quantum computer with more time can hit a table base with more than 7 men consistently from the initial position.
#6897
Checkers and even draughts players have flocked to chess and go after checkers was solved. Also competitive play of Nine Men's Morris and Connect Four has stopped after these games were solved. Capablanca already proposed his Capablanca chess and Fischer proposed Fischer Random aka Chess960.
Quantum computers are only much faster for a very narrow set of applications that does not include solving chess . Plus, IBM has disputed Google's assertions. Their supercomputer solves the quantum-friendly problem in 2.5 days, not 10,000 years. Slight difference of opinion.
As before, there's no leg for you to stand on until quantum computers actually achieve something that requires looping iterations and intermediate storage of results. So many posters have already pointed this out to you...me, BlueEmu, Elroch (I think, anyway), Pfren, Llama, and so on. Pretty much anyone with notable science chops or engine expertise has come down opposite your position, as far as I have seen.