‘Impossible’ to create AI tools like ChatGPT without copyrighted material, OpenAI says::Pressure grows on artificial intelligence firms over the content used to train their products

  • BURN@lemmy.world
    link
    fedilink
    English
    arrow-up
    10
    arrow-down
    4
    ·
    10 months ago

    Too bad

    Why do they have free reign to store and use copyrighted material as training data? AIs don’t learn as a human would, and comparisons can’t be made between the learning processes.

    • SCB@lemmy.world
      link
      fedilink
      English
      arrow-up
      4
      arrow-down
      5
      ·
      edit-2
      10 months ago

      Why do you have free reign to do the same?

      AIs don’t learn as a human would, and comparisons can’t be made between the learning processes.

      I think you’re going to have a hard time proving a financial distinction between them

      • BURN@lemmy.world
        link
        fedilink
        English
        arrow-up
        4
        arrow-down
        1
        ·
        10 months ago

        You don’t need to prove a financial difference. They are fundamentally different systems that function in different ways. They cannot be compared 1:1 and laws cannot be applied as a 1:1. New regulations need to be added around AI use of copyrighted material.

        • SCB@lemmy.world
          link
          fedilink
          English
          arrow-up
          5
          arrow-down
          5
          ·
          10 months ago

          I agree. For instance, it should be secured in law that you can train AI on anything, to avoid frivolous discussions like this.

          Output is what should be moderated by law.

          • BURN@lemmy.world
            link
            fedilink
            English
            arrow-up
            5
            arrow-down
            4
            ·
            10 months ago

            No

            Why are you entitled to use everyone else’s work? It should be secured in law that licensing applies to training data to avoid frivolous discussions like this. Then it’s an entirely opt-in solution, which works in the benefit of everyone except the people stealing data.

            Output doesn’t matter since it’s pretty well settled it’s not derivative work (as much as I disagree with that statement).

            • SCB@lemmy.world
              link
              fedilink
              English
              arrow-up
              6
              arrow-down
              4
              ·
              10 months ago

              the people stealing data

              No one is doing this

              Output doesn’t matter since it’s pretty well settled it’s not derivative work

              Cool, discussion over.

              • BURN@lemmy.world
                link
                fedilink
                English
                arrow-up
                4
                arrow-down
                4
                ·
                10 months ago

                It is stealing data. In order to train on it they have to store the data. That’s a copyright violation. There’s no way to interpret it as not stealing data.

                • 5too@lemmy.world
                  link
                  fedilink
                  English
                  arrow-up
                  3
                  arrow-down
                  3
                  ·
                  10 months ago

                  It is not stealing. The data is still there. It is, at worst, copyright violation.