The ancient board games we finally know how to play – thanks to AI

[Iucounu]

https://www.newscientist.com/articl...mes-we-finally-know-how-to-play-thanks-to-ai/

Could this be true, or is it just another AI-hype article? It's partially behind paywall, so I can't read the details.

 

[Heroic Liberator]

Doubt it since the number of elements (a.i. board size, pieces, dice) say nothing about the rules. If you had a chess game with no instructions and gave it to an AI for analyzing, I don't see how it could discover how the various piece types move on the board.

 

[dbx]

I don't know. Some game board do give clues about potential rules (or rather a limited subset), like go or backgammon. Maybe even checkers.
With chess, you could derive some overall set of rules simply by seing how the board is made and that it has pieces with different shapes.
Obviously you can't derive the exact details about pieces movement, but you can come up with working ruleset that are pretty similar to real chess rules.
Removing pieces from the board is a common main objective, from that you derive working set of potential movements rules, start by eliminating movements that are too complex, movements that are too retarded, movements that unbalance the game, rules that leads to singular winning strategy, etc...

 

[Heroic Liberator]

Sure, but at that point can you still say that you discovered how that game is to be played or have you just come up with a new game using the same pieces as the original?

 

[dbx]

Well, clickbait title of the article apart, I'd say you've discovered a likely set of rules.
Details might vary but the overall experience do not, still don't know the actual rules, but we end up knowing at least the set of potential rules it might have had, the real rules are one of the possibilities in the set.
Which is more than enough for something we previously didn't know jack shit.

 

[Oreshnik Missile]

What does AI do, simulate playing matches according to randomised sets of rules and record which rulesets lead to the most diverse outcomes?

 

[Heroic Liberator]

What does AI do, simulate playing matches according to randomised sets of rules and record which rulesets lead to the most diverse outcomes?

Yeah, but that would just make for a good game. The original game OTOH wasn't some platonic ideal of what would be a good game with those pieces, just a game that people came up at some point with its own redundancies.

Well, clickbait title of the article apart, I'd say you've discovered a likely set of rules.
Details might vary but the overall experience do not, still don't know the actual rules, but we end up knowing at least the set of potential rules it might have had, the real rules are one of the possibilities in the set.
Which is more than enough for something we previously didn't know jack shit.

Eeeh, see above. And the possibilities are so many due to the number of pieces that knowing all of the possible rule permutations (many of which can't actually be inferred from the pieces themselves) is equal to knowing none of them. So in the end we didn't really learn anything. We just created something new which might be valuable in itself as a creative (and subsequently recreational) pursuit, but without giving us any new (historical) knowledge to speak of.

 

[wideman]

Sure, but at that point can you still say that you discovered how that game is to be played or have you just come up with a new game using the same pieces as the original?

yeah. Try to guess how cards are played.

AI could potentially come up with some 'games' from the resources (52 playing cards) given. But would anyone would be remotely interested to play.. Probably someone on the spectrum

 

[DemonKing]

Sure, but at that point can you still say that you discovered how that game is to be played or have you just come up with a new game using the same pieces as the original?

That's the most likely outcome.

Especially since in a mostly pre-literate world rules would probably vary greatly from community to community even if the basic pieces were roughly the same.