• brygphilomena@lemmy.dbzer0.com
    link
    fedilink
    English
    arrow-up
    17
    arrow-down
    2
    ·
    4 days ago

    I just don’t get what AI is doing in this context. Build normal fucking algorithms.

    I saw that the average savings was 10 minutes and $375 a flight. For a system that cost $1billion. It won’t pay itself off until 2.6 MILLION flights have completed.

    What a farce.

      • datendefekt@feddit.org
        link
        fedilink
        English
        arrow-up
        1
        ·
        2 days ago

        Optimization is a classic computer science problem since the start of the field, and an intensive field of ongoing research.

        Pattern matching (not LLMs!) is good at recognizing recurring situations and could be used to recommend procedures.

        LLMs are non-deterministic, don’t use variables and are known to short-cut to their goal. You can’t test them and expect the same answer twice, and they might forget things. Good for emails, bad for air traffic.

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

          It’s the main computer science problem. Cybernetics. Counting cars, trips, metal bars and all other things on a factory for optimal throughput. Operating a government responsive to events. Predicting weather. Predicting war. Calculating a nuclear blast without making a nuclear blast. Or a space ( or subspace ) flight without building a rocket.

          A-and also building a cybernetic democracy before cybernetics in the fields mentioned above make voting a formality because the electorate has been calculated. That thing seems to have failed to develop fast enough.