• lime!@feddit.nu
    link
    fedilink
    English
    arrow-up
    1
    ·
    9 days ago

    idk if that’s the intended takeaway from those numbers.

    According to AllAboutAI analysis, global AI processing generates over 260,930 kilograms of CO₂ monthly from ChatGPT alone, equivalent to 260 transatlantic flights, with 1 billion daily queries consuming 300 MWh of electricity.

    according to the faa there are on average 5500 planes in the air every day, and while i couldn’t find an exact number there seem to be between 350 and 1 200 transatlantic flights every day, depending on season.

    260 tons is still massive, but let’s not kid ourselves. it’s about equivalent to producing 12 new american-size cars.

    • leftthegroup@lemmings.world
      link
      fedilink
      English
      arrow-up
      0
      ·
      9 days ago

      Yes, but there’s zero fucking actual benefit.

      Seeing memes posted here that use AI while sitting on it is the most confusing thing to me.

      Just… don’t use it, people. The hole burning in AI bros’ pockets will close up if you just stop making it profitable. Even the free ones are making money with ads. Don’t use it, even for a joke.

      • FreedomAdvocate@lemmy.net.au
        link
        fedilink
        English
        arrow-up
        0
        ·
        9 days ago

        People like you wouldn’t have seen the benefit in cars vs horse and carriages, computers vs typewriters and books, or watches vs sundials.

        I bet you think that the only thing AI is used for is ChatGPT style conversations too.

        • very_well_lost@lemmy.world
          link
          fedilink
          English
          arrow-up
          0
          ·
          8 days ago

          wouldn’t have seen the benefit in cars

          Yeah, because the widespread adoption of cars turned out to be such a great idea with no negative consequences… But even if you ignore the glaringly obvious negatives, AI still doesn’t come anywhere close to having the practical utility as the modern car. At least a car can carry out its advertised function without issues.

          I’ve been using AI almost daily for several years now, as a function of my job. It’s garbage tech. Most of the things it’s supposed to be good for it downright sucks at, and the stuff it is good at has already been possible using simpler, more reliable systems for years — sometimes even decades. The situation isn’t really improving, either. Models are using more energy, consuming more data, and doing more computation than ever before… but the results are still embarrassingly underwhelming. Anyone who’s bothered to educate themselves on the math and method behind the models knows by now that the current generation of AI is dead-end technology, and anyone who claims otherwise is either ignorant of the technical details, has a vested financial interest in AI, or both.

          It also really fucking irritates me to be constantly called a Luddite by people who don’t even know how this technology fucking works… No, I don’t hate AI because I’m scared of technology, or “progress” or whatever the fuck. I’ve made a career working in technology. I love tech… or I used to, before everyone lost their god damn minds praying to Sam Altman and his horrifyingly expensive golden idol. No, I hate AI because it’s demonstrably bad technology.

          • hisao@ani.social
            link
            fedilink
            English
            arrow-up
            0
            arrow-down
            1
            ·
            8 days ago

            I personally enjoy using it to the point I’m ready to pay for it. It helps me figuring out rather complex things where I wouldn’t even know what to type into a normal search engine to start tackling the problem. Imo, both forcing people to use it and forbidding people to use it, is making people unhappy. Just let people work the way they want to.

            Yesterday I got just a regular ChatGPT explaining me how to convert some geometry into screen space using inverted transform matrix of the camera and dividing x,y by z in camera space to get the projection and then normalizing using min/max x,y, then after I got some bad results trying to get those numbers into the place where I need them, writing me a script that transforms my node-based Blender geonodes setup into JSON (just for the sake of giving this JSON to ChatGPT for analysis), then after reading this JSON explained to me some advanced control and data flow intricacies of geonodes and recommended a setup I could use to reshape control and data flow the way I need. This is all rather useful and would take more time and effort to gather all this information by myself. And it’s not like I’m not learning anything, it just makes learning faster.

            spoiler