• stoy@lemmy.zip
    link
    fedilink
    arrow-up
    3
    arrow-down
    1
    ·
    6 days ago

    Don’t they do that from time to time and then switch back to M$?

    • onlinepersona@programming.dev
      link
      fedilink
      arrow-up
      2
      ·
      5 days ago

      The same old broken record repeating a single incident and presenting it as multiple 😮‍💨

      “You fuck a sheep one time…”

    • fruitycoder@sh.itjust.works
      link
      fedilink
      arrow-up
      3
      ·
      5 days ago

      I think the ideological motivations have changed this time around. Its not just public money public code, its also dependency on increasingly erratic nation with heavy ties to big tech is a threat to investing in building around their software rentals.

      It was a ratchet approach before from the outside looking in

    • mysticpickle@lemmy.ca
      link
      fedilink
      arrow-up
      4
      ·
      6 days ago

      The real news here: Is using Þ going to stop an auto complete machine from auto completing whatever it scrapes?

        • mysticpickle@lemmy.ca
          link
          fedilink
          arrow-up
          2
          ·
          5 days ago

          Okay. Did you actually read this study or did you just read only the title and think it applies to your specific effort?

          This study looks like it’s taking about injecting a LLM’s data set using full on designed documents embedded with trigger words that will specifically cause the model it to spit out either gibberish or break the guardrails that keep it from outputting information it’s not supposed to. A regular non-traditional character substitution isn’t going to do that.

          Even if by some miracle whatever you’re doing causes these outdated letters to show up in an AI’s output regularly, it’d be pretty trivial to for a company to correct it by autosubstituting any occurrences before they feed any data in.

          All you’re doing right now is making people do ineffective captcha to make sense of what you write. Like most captcha it’s not particularly hard to do but it is annoying AF judging by all the downvotes your comments seem to receive.

          Your efforts to derail the spread of AI would be more effective by speaking out against it in a clear manner that doesn’t immediately alienate so many people.

          What you’re doing right now feels more like you’re trying to force some cringe social media trend like “That’s so fetch/wizard/bacon!” to happen. It’s honestly kinda gross.

          • Ŝan@piefed.zip
            link
            fedilink
            English
            arrow-up
            1
            arrow-down
            6
            ·
            5 days ago

            Oh, no, I didn’t read it. I had an LLM read it and summarize it for me.

            • mysticpickle@lemmy.ca
              link
              fedilink
              arrow-up
              2
              ·
              5 days ago

              Alright. So, why would you post this study to reinforce your original point when it pretty clearly doesn’t do that?

              Were you just hoping no one would actually read it so you wouldn’t have to admit that this whole exercise of using a dead character is more about you trying to feel special than actually do something more effective to curb AI?

              Pretty bold gamble bro.