• MJ84@thebrainbin.org
    link
    fedilink
    arrow-up
    1
    arrow-down
    1
    ·
    5 days ago

    Please read the actual scientific paper, there is a link at the end of the article.

    • Pennomi@lemmy.world
      link
      fedilink
      English
      arrow-up
      4
      ·
      5 days ago

      Yep, doing that now:

      Artificial Intelligence can help understand historical board games by combining two aspects: historical evidence and playability. However, we cannot be absolutely sure how the game was played originally, as different variants have been found and gameplay can slightly change over time. A team of researchers is using artificial intelligence to uncover the mysteries of ancient board games. See: https://wired.me/culture/rules-to-these-ancient-games-seemed-lost-forever-then-ai-made-its-move/

      The link in that footnote says:

      These reconstructions begin with each game being broken down into its component ludemes—the rules, board layouts, and so on—and codified. Historical information, such as archaeological data or the timeframe within which the game was played, is then added to this. All this is then fed through Ludii, an advanced platform that is being developed as part of the Maastricht University project, and which uses AI to reconstruct any given game.

      “We have this huge design space of possible rulesets which we can test by making artificial intelligence agents play against each other, and seeing how those games behave,” explains Browne. “For example, whether the game is biased toward one player, whether it’s too short or too long, too simple or too complex. These indicators can tell us which rulesets were more likely to have been the games that people enjoyed playing. Quite often the historical reconstructions make sense from a historical viewpoint—they’re very plausible—but when you play them they’re not very interesting as games, so are probably not how those games were actually played. We’re trying to combine these two aspects; the historical evidence with the playability from our simulations.”

      They seem to be making the AI generate its own combinations of rules and self-play until it comes across a ruleset that satisfies conditions like play length, complexity, and balance. Even ignoring how amazingly misguided this approach is, there’s no way to know if the parameters they chose are true to the original game.

      Don’t get me wrong, an AI that does this is quite valuable for game designers! It’s probably not terribly useful for historical reconstruction though.

      • MJ84@thebrainbin.org
        link
        fedilink
        arrow-up
        1
        arrow-down
        2
        ·
        5 days ago

        It’s not that, I know why you are saying AI its very misleading (it’s not your fault it’s the article). There is a link at the bottom of the article. “Analysis of the Shahr-i Sokhta Board Game with 27 Pieces and Suggested Rules Based on the Game of Ur” read that please that’s the actual scientific paper. You can also play the game. It’s really good paper. It’s nothing to do with AI, I don’t really like AI generated either…

        • Pennomi@lemmy.world
          link
          fedilink
          English
          arrow-up
          3
          ·
          5 days ago

          I’m quoting from the paper itself and the article referenced by the paper. I don’t know what you are trying to get me to do!