• Natanox@discuss.tchncs.de
      link
      fedilink
      English
      arrow-up
      2
      ·
      5 days ago

      It’s genuinely sad the information landscape for Linux is so bad due to time and lack of documentation (all hail the Arch Linux wiki), big LLMs got a higher hitrate.

      Unfortunately if they miss they might tell you to dd your disk as a speedtest.

      • Stinkywizzleteats@piefed.social
        link
        fedilink
        English
        arrow-up
        0
        ·
        5 days ago

        what LLM are you using that will tell you to do dd for a speed test? it has to have happened for you to talk about it right? you arent just making stuff up like a Karen calling 911 on a black person are you? that would be stupid to just make stuff up to dunk on AI for clout.

        • Natanox@discuss.tchncs.de
          link
          fedilink
          English
          arrow-up
          0
          ·
          5 days ago

          Your reply reads like a major car crash.

          It was Mistral AI (most likely their mistral-large model, whatever was live for logged-in users about 4 months ago), see picture. Given enough inconclusive context, a sufficiently niche problem and/or a persistent user request they’ll all generate dangerous nonsense or make shit up. ChatGPT isn’t any better, just more willing to answer something positive even if it ends up being wrong.

          • Stinkywizzleteats@piefed.social
            link
            fedilink
            English
            arrow-up
            0
            arrow-down
            2
            ·
            5 days ago

            no it doesnt

            we are in a time where Anti AI claims need to be verified as much as actual AI output. people are so full blown Anti AI clout chasers that they just make stuff up. that’s why “AI do X, make no mistakes” is a meme. lots of people are dogging on AI while putting the absolute minimum into trying to one shot some complex task then using it as “proof” AI is useless. they are the opposite side of the pro ai tech bro coin and they are making life easier for those companies by making low effort claims. I hate ai slop and I’m starting to hate anti ai slop even more because it delegitimizes my stance.

            you screenshot means nothing. you can get a LLM to say anything. since we are on the bullshit anti ai timeline you need more than a screenshot. AI companies need more than a claim their AI “broke out” of its sandbox. dig just a little deeper and you see it 100% didn’t in fact autonomously get out. I’m sure if we dug deeper into what lead to your screenshot we would find multiple reasons for it to give the output it gave. also it’s 2026 why would you think that is proof of anything?

            • Natanox@discuss.tchncs.de
              link
              fedilink
              English
              arrow-up
              0
              ·
              5 days ago

              Good thing that whole Chat (and therefore the screenshot) was literally vetted and used by a CCC member for a talk at the Hackover 2026 in Germany, a small but well-established hacker event. So I’m kind of sure your subtle insinuation that I might’ve just prompted it badly is bogus. 😘

              Either way, nobody owes you or anyone else proof unless they specifically aim to convince other people. If you feel the social media posts of other people “delegitimizes your stance” that’s on you.

              • Stinkywizzleteats@piefed.social
                link
                fedilink
                English
                arrow-up
                0
                arrow-down
                2
                ·
                5 days ago

                IT professionals. The ones who’s careers are hanging by a thread said AI bad and that’s your source because it was literally vetted.

                no one said you generated anything.

                there is nothing wrong with my stance that crying wolf hurts everyone.