• yeahiknow3@lemmy.dbzer0.com
    link
    fedilink
    English
    arrow-up
    10
    arrow-down
    18
    ·
    2 days ago

    No, they detect AI use, which is correlated with paraphrases of the content they stole from.

    Also, this whole idea that AI detectors don’t work is tired. GPTZero has sub 1% false positive rate for creative writing. What more do you want.

      • yeahiknow3@lemmy.dbzer0.com
        link
        fedilink
        English
        arrow-up
        5
        arrow-down
        5
        ·
        edit-2
        2 days ago

        Yeah, me too. But until we get laws in place banning AI (and reducing their board of directors to mulch), we should try to support detecting and punishing its use. GPTZero specifically has very high accuracy. Some of the other detectors really suck.

        • Sarah Valentine@pawb.social
          link
          fedilink
          English
          arrow-up
          6
          arrow-down
          3
          ·
          2 days ago

          But until we get laws in place banning AI (and reducing their board of directors to mulch), we should try to support detecting and punishing its use.

          What exactly do you think drives the AI detectors?

            • Sarah Valentine@pawb.social
              link
              fedilink
              English
              arrow-up
              6
              arrow-down
              3
              ·
              edit-2
              2 days ago

              Answer the question.

              Edit: All right, since you won’t, I’ll spell it out for you. It’s AI. The AI detectors are AI models. You are suggesting we put the policing of AI use in the hands of AI. It’s not even close to a good idea. It’s nothing less than self-defeating.

    • SirLeToet@lemmy.world
      link
      fedilink
      English
      arrow-up
      3
      arrow-down
      1
      ·
      1 day ago

      Until you see it in practise. People around me that are studying are getting flagged constantly.

      Marketing.

      • yeahiknow3@lemmy.dbzer0.com
        link
        fedilink
        English
        arrow-up
        8
        arrow-down
        10
        ·
        edit-2
        17 hours ago

        Yeah, no. Accuse isn’t convict.

        Also, as a teacher I would literally already know they cheated. AI writing is easy to spot. Ironically, it’s always people who can’t write who’re super worried about getting “falsely” caught using AI. Never do I hear creative writing majors or English majors or philosophy majors express a shred of concern. I wonder why that is?

        • Mika@piefed.ca
          link
          fedilink
          English
          arrow-up
          6
          arrow-down
          1
          ·
          1 day ago

          You literally cheating doing your job when you use janky AI detectors that have no way to know if the work is made by AI and just hallucinate some score.

          The real AI detection like what is enforced by EU (secretly shifting biases towards certain words to appear more) would not have publically available validators cause otherwise anyone would be able to learn how to avoid getting caught. Plus, you’d have to validate for every LLM separately. Plus, open weight LLM would whoop your ass anyway.

          • jj4211@lemmy.world
            link
            fedilink
            English
            arrow-up
            5
            arrow-down
            2
            ·
            2 days ago

            When used for cheating by a student you see a lot, it’s almost certainly going to be super easy to spot.

            The appeal for cheating is that it skips the work. If they skip the work, they didn’t rework the GenAI output. Even if the GenAI output could pass for a human, it’s not going to pass for that human that you know.

            Also, advanced classes make ‘homework’ barely count, so even if you got away with cheating, the in-person assessments are going to ruin your grade.

          • yeahiknow3@lemmy.dbzer0.com
            link
            fedilink
            English
            arrow-up
            3
            arrow-down
            5
            ·
            2 days ago

            What do you think my accuracy would be for high school essays given that I’ve read maybe 10,000 of them? This whole conversation is making me cringe. I’m out.

            • Despair@lemmy.world
              link
              fedilink
              English
              arrow-up
              4
              arrow-down
              1
              ·
              edit-2
              2 days ago

              In my experience, highschool teachers could not be bothered to look over or grade assignments. They would do a lazy “what grade do you think you deserve” essay at the start of the year and carry that grade that you gave yourself through the remaining terms regardless of how much work you did or did not do, even missing that mandatory assignments were skipped.

          • yeahiknow3@lemmy.dbzer0.com
            link
            fedilink
            English
            arrow-up
            2
            arrow-down
            5
            ·
            edit-2
            2 days ago

            Look at those poor business and CS majors. Simultaneously illiterate and also very concerned about being accused of using AI to write.

            A score of 149 in the GRE Verbal is 37th percentile.

            You can achieve this score by leaving most of the test blank. The average American has a 6th grade reading level, btw. Google.

        • jj4211@lemmy.world
          link
          fedilink
          English
          arrow-up
          5
          ·
          2 days ago

          Suppose the question is why even use the tool if you can tell by looking.

          I’m guessing it’s because if you just say it, they’ll argue or maybe get their parents to argue and if you can have a ‘detector’ serve as ‘evidence’? Especially if they believe in the tool to use in the first place, they are likely to believe in the detector tool, no mater how flawed? Or to just serve as a general deterrent, your work is being checked by a machine and you’ll get caught if you try…

          I’ll say at least some kids are scared of being falsely accused of cheating because they don’t trust the tools to be accurate, perhaps more so in the very anti-AI kids that would avoid the AI like the plague. No perfect answer how you reassure that the tool isn’t immediately trusted without also defanging it to the intended audience in that case though…

          • yeahiknow3@lemmy.dbzer0.com
            link
            fedilink
            English
            arrow-up
            3
            arrow-down
            1
            ·
            2 days ago

            Luckily, previous drafts are automatically saved in most word processors, but you’re right. It’s a hideous new world.

          • DaleGribble88@programming.dev
            link
            fedilink
            English
            arrow-up
            1
            arrow-down
            1
            ·
            edit-2
            1 day ago

            I’ve used the detectors as well. At my university, accusing a student of cheating is a big deal and you want to go into it with as much documentation as you can muster. Even though I know it is AI, I’m going to be showing it to administrators without my area of expertise and experience. Administrators like reports and numbers, so I run whatever the assignment is through a checker, let it generate a score over 90%, and now I’ve got a number inside of my paperwork that will encourage my supervisor, or more often my supe’s supe, to take my accusation seriously. Before these AI checkers, I used “turn it in”, and before that I had LCS tools and diff command.

        • w3ird_sloth@lemmy.world
          link
          fedilink
          English
          arrow-up
          5
          arrow-down
          2
          ·
          2 days ago

          This just comes off like you’re biased against any students who are into STEM fields and favor the ones in the liberal arts.

      • yeahiknow3@lemmy.dbzer0.com
        link
        fedilink
        English
        arrow-up
        1
        arrow-down
        4
        ·
        2 days ago

        Why would I trust independently published research in addition to my own vast professional experience? You’re right. Internet forums are much more reliable.