https://en.wikipedia.org/wiki/Solved_game (Ultra-weak)
https://en.wikipedia.org/wiki/Solving_chess
To bet on what the outcome will be, see:
@diracdeltafunk Doesn't Manifold provide a loan on this for long bets to not keep money Mana tied up? or is that only for certain markets?
Y'all have never tried actually coming to grips with the size of the chess game tree
Out of curiosity, have you ever come to grips with this yourself? How many positions do you expect White/Black to be able to reach if Black/White was playing perfectly? How much compute do you think is needed for a solution here?
@BoltonBailey I think there are probably relatively few games which are perfect play (probably few enough to fit on all hard drives ever produced combined, for example). However, it is very very hard to tell which games these are; that's the whole point. I think solving chess would take more compute than we will EVER have, unless someone discovers a really amazing invariant of chess game states. I don't think there is likely to exist an amazing invariant like this.
To answer why quantum computers would help: We could develop some heuristic algorithm that plays sufficiently well and then prove there is no combination of moves from the opponent that beats it by Grover searching for such a combination and failing to find one.
As to why we would do this, I suppose the answer is the same as for any kind of blue-skies research - we would hopefully acquire more practical knowledge along the way.
I wonder if @jonsimon, @anon and the other people who think this is near zero also think that /BoltonBailey/chess-with-queen-odds-solved-by-204 should be near zero. Is it a question of good heuristics being present, or is the game tree complexity the only thing that matters?
@BoltonBailey id best against us finding any structure that makes brute force unnecessary and also against brute force being available in the insanely short window of 16 years. I don't know about chess with queen odds, depends on state space size and if it seems like there could be some structure to it.
@anon What chance do you think there is of AGI before 2040? Most of the worlds in which this resolves yes IMO involve AGI
@dominic AGI means a system with human-or-better capabilities in a wide range of fields. It doesn't mean "solves all NP hard problems".
@jonsimon Definitely wouldn't solve all NP-hard problems. But solving chess to me seems in the ballpark of things that are hard for humans but a superintelligent AI would help a lot with.
@dominic general to human intelligence maybe dyson swarm or really effective fusion almost certainly not
@anon actually maybe fusion but I doubt we'll control enough energy transistors to do it by 2040 in any case even w fusion being somewhat viable before then
@dominic Thing is - we already have chess superintelligence, a chess engine running on a phone can can consistently outplay top players. Also current machine learning systems are very unhelpful in proving things rigorously.
Having a machine system that outplays anything else on the planet claim it has the optimal moves doesn't really solve the game, unless there's a rigorous proof that no better strategy exists.
/MartinRandall/when-will-7x7-go-be-solved
Pretty sure this should be easier than Chess.