A scientist admitted he'd "lost" two years' worth of "carefully structured academic work" after turning off ChatGPT's "data consent" option.
    • Flauschige_Lemmata@lemmy.world
      10·
      6 hours ago

      If he only used it as a tool, I’d still say it’s his work. It should of course be included in the methodology section.

      But considering he lost 2 years worth of work, just because his chats got wiped? That’s more than just a tool

    • filcuk@lemmy.zip
      27·
      3 hours ago

      Is it the LLM’s research? Do guns shoot people?
      Researching other people’s work is still research. If they start claiming other people’s work is their work without having added any value, then it’s obviously stealing, but otherwisethis argument makes no sense.

      • Nalivai@lemmy.world
        81·
        2 hours ago

        When you type “dear robot girlfriend, do my work, make no mistakes” into a chatbot window, and then use the slop that the average word predicting machine shat out, you’re not doing research. In fact, you’re not doing anything besides frying your brain while frying the planet.
        It’s not “research” at all, but whatever that is, llm generated it, you did nothing.

        • filcuk@lemmy.zip
          33·
          2 hours ago

          So if I just quickly search Wikipedia or list through a few books, is that a research? How many prompts or hours until my unnamed activity becomes research? Is it a hard limit, or just based on how hateful you’re feeling for the day?

          • Nalivai@lemmy.world
            31·
            41 minutes ago

            If I go to the restaurant and order something, is that counts as me cooking? How many times I need to point to the waiter at the menu and ask them to bring me something, until I am officially count as a professional cook? If I ask them to make it less salty and add cheese, is it counts as the restaurant employing me as a chef or only as a liner cook?
            Just in case your chat “research” fried your brain completely, and it needs to be spelled out: no, to be called cook you need to cook the food. To do research you need to do research, not ask a word prediction machine to do it.

  • Grimy@lemmy.world
    180·
    16 hours ago

    He lost his chat logs. I was going to mock him him for not backing up his work but I’m not sure I want to punch down on someone who considers his LLM chat logs research.

    • flango@lemmy.eco.br
      1·
      14 minutes ago

      I think that scientists embracing AI so naively as this guy contribute to a larger discredit in science. Probably we will have in the future antivax people saying that " such and such vaccines were made of AI slop".

    • MajinBlayze@lemmy.world
      36·
      9 hours ago

      Wait wait wait. ALL of his research was in his chat logs in chat gpt‽ So NONE of his research was outside of chat gpt.

      I’m not sure it’s meaningful to call that research

    • RustyNova@lemmy.world
      4·
      15 hours ago

      Same for me but the reason is different

      … I don’t want to punch down someone for not using backups while I probably have a lot stored on sites and ready to be wiped accidentally

  • mech@feddit.org
    49·
    14 hours ago

    ChatGPT’s chat history isn’t “carefully structured academic work”.
    He basically used the 0/0/0 backup strategy, with not even one instance of his data saved anywhere.

    • TheTechnician27@lemmy.worldEnglish
      332·
      15 hours ago

      As someone who knows a decent bit about machine learning, I can say Dr. Collier seems to know very little of how it works on a technical level, which presents such a beautiful contrast when I’m watching her be fucking dead-on about its societal consequences. Seeing someone who isn’t familiar with the field but still knows what they’re talking about is so cathartic in a subject strewn with misconceptions from laypeople. Angela’s integrity is really admirable; you can tell that critical thinking is a virtue to her and that she takes a lot of care not to overstep into areas she doesn’t understand.

        • TheTechnician27@lemmy.worldEnglish
          112·
          14 hours ago

          I always do. I write as a hobby, and I’m not stuck here, so I don’t see why I’d be here if I didn’t give enough of a shit to write what I want.

        • Kogasa@programming.dev
          23·
          8 hours ago

          At some point, we need to stop caring or stop interacting entirely. It’s so exhausting to have to question the authenticity of every interaction.

  • gustofwind@lemmy.worldEnglish
    51·
    16 hours ago

    He disabled the feature because he “wanted to see whether I would still have access to all of the model’s functions if I did not provide OpenAI with my data.”

    Well I guess you found out

  • Flying_Lynx@lemmy.ml
    12·
    14 hours ago

    But now you can also question the validity of his papers, the validity of his educational value in general, and, if you do research, the safety issues of having no contingency plans in case an hypothesis fails…

    • Nalivai@lemmy.world
      1·
      2 hours ago

      Yeah, the guy’s credentials should be scrutinized if not revoked. This is a disgrace.
      And it’s the one who is public about it, imagine how many of his colleagues do the same, but have enough brain cells left to not actually brag about it