• Lovable Sidekick@lemmy.world
    link
    fedilink
    English
    arrow-up
    5
    ·
    edit-2
    2 days ago

    It’s actually called Symbolic AI, which codifies clearly defined rules and behaviors - think decision trees and flowcharts. I doubt machine learning will be used in this case, since the rules and reasoning used by air traffic controllers have evolved over almost a century and are well understood. Machine learning is more for situations where the software has to discover the best way to operate because the rules aren’t well established.

    • Logi@lemmy.world
      link
      fedilink
      arrow-up
      1
      ·
      2 days ago

      Machine learning is more for situations where the software has to discover the best way to operate because the rules aren’t well established.

      Or the machine needs to look at messy input data and recognise what the situation is so it can apply the more structured decision processes. You can’t feed a radar image to a bunch of if/elif/elif/else statements. That pattern recognition is fallible, but so are humans. Exciting times!

      • trebach@sh.itjust.works
        link
        fedilink
        arrow-up
        1
        ·
        1 day ago

        You can’t feed a radar image to a bunch of if/elif/elif/else statements.

        You don’t need to input the image. You just need to input call sign, latitude, longitude, altitude, heading, speed, and emergency status (none, pan pan, or mayday) for each aircraft on radar.

        The ATC will set the last one based on radio communications but the rest are either directly from or can be calculated from their transponders.

        • Logi@lemmy.world
          link
          fedilink
          arrow-up
          1
          ·
          17 hours ago

          Well, it’s neither. This is not about to be used in the tower per other comments.

      • Lovable Sidekick@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        2 days ago

        You’d be surprised what symbolic AI can do with messy data - in medical diagnosis, for example - but LLMs suck at it.