• SpanishSpaceAgency [he/him]@hexbear.net
      link
      fedilink
      English
      arrow-up
      49
      ·
      2 months ago

      But it isn’t even a contradiction. The more you understand how it works the less you like it. Once you’ve peeked behind the curtain you are aware the great and powerful Oz is just smoke and mirrors and bullshit and you don’t believe it’s magic anymore. That’s not contradictory that’s just rational.

      • space_comrade [he/him]@hexbear.net
        link
        fedilink
        English
        arrow-up
        29
        ·
        edit-2
        2 months ago

        I’d argue only at that point can the AI actually be useful to you because you’re aware of its inherent limitations and can just use it for the things you know it’s good at, which is not that many things actually so yeah you end up using it way less.

          • hello_hello [comrade/them]@hexbear.net
            link
            fedilink
            English
            arrow-up
            14
            ·
            edit-2
            2 months ago

            Its not even auto complete. I tried running qwen3:8b param model on ollama and it just felt so tiring having to cajole the program into producing the output I want.

            Everything was very underwhelming and could be done better by a 15 year old human personal assistant who actually likes programming.

            AI is the supposed bourgeois escape hatch for starving and mistreating workers when the reality of austerity hits them fast and no one is able to work.

            • space_comrade [he/him]@hexbear.net
              link
              fedilink
              English
              arrow-up
              4
              ·
              2 months ago

              Models that you can run locally on reasonably priced consumer grade hardware are always going to be crap, you really need the beefy models that destroy the environment for it to be somewhat useful.

                • space_comrade [he/him]@hexbear.net
                  link
                  fedilink
                  English
                  arrow-up
                  3
                  ·
                  2 months ago

                  Oh yeah this whole thing is very very much convenient to Nvidia and data center companies lmao.

                  I still think they’re incentivized to make the models more efficient as they could then squeeze out even more profit, it’s just that it’s a property of the technology itself that it doesn’t really work well until you have bajillions of parameters.

      • queermunist she/her@lemmy.ml
        link
        fedilink
        English
        arrow-up
        24
        ·
        2 months ago

        The more you understand how it works the less you like it.

        That’s the dialectic! There’s a mutual antagonism between understanding AI and wanting to use AI.