AI companies have all kinds of arguments against paying for copyrighted content::The companies building generative AI tools like ChatGPT say updated copyright laws could interfere with their ability to train capable AI models. Here are comments from OpenAI, StabilityAI, Meta, Google, Microsoft and more.

  • Uriel238 [all pronouns]@lemmy.blahaj.zone
    link
    fedilink
    English
    arrow-up
    19
    arrow-down
    3
    ·
    1 year ago

    Here is what deliberative experts way smarter and more knowledgeable than I am are saying ( on TechDirt )

    TLDR: Letting AI be freely trained on human-made artistic content may be dangerous. We may decide to stop it so long as capitalists control who eats and lives. But copyright is not the means to legally stop it. This is a separate issue to how IP law is way, way broken. And precedents stopping software from training on copyrighted work will be used to stop humans from training on copyrighted work. And that’s bad.

    • ryannathans@aussie.zone
      link
      fedilink
      English
      arrow-up
      11
      arrow-down
      6
      ·
      1 year ago

      Agree, it’s not much different from a human learning from all these sources and then applying said knowledge

      • realharo@lemm.ee
        link
        fedilink
        English
        arrow-up
        9
        arrow-down
        1
        ·
        edit-2
        1 year ago

        Scale matters. For example

        • A bunch of random shops having security cameras, where their employees can review footage

        • Every business in a country having a camera connected to a central surveillance network with facial recognition and search capabilities

        Those two things are not the same, even though you could say they’re “not much different” - it’s just a bunch of cameras after all.

        Also, the similarity between human learning and AI training is highly debatable.

        • Both of your examples are governed by the same set of privacy laws, which talk about consent, purpose and necessity, but not about scale. Legislating around scale open up the inevitable legal quagmires of “what scale is acceptable” and “should activity x be counted the same as activity y to meet the scale-level defined in the law”.

          Scale makes a difference, but it shouldn’t make a legal difference w.r.t. the legality of the activity.

          • lollow88@lemmy.ml
            link
            fedilink
            English
            arrow-up
            2
            arrow-down
            1
            ·
            1 year ago

            Scale makes a difference, but it shouldn’t make a legal difference w.r.t. the legality of the activity.

            What do you think the difference between normal internet traffic and a ddos attack is?

      • topinambour_rex@lemmy.world
        link
        fedilink
        English
        arrow-up
        7
        arrow-down
        2
        ·
        edit-2
        1 year ago

        When Google trained their playing neural network, they trained it to starcraft2 . It was better at it than professional gamer. It trained by watching 100years of play. Or 36500 days of play. Or 876000 hours of play.

        Does a human can do that ? We both know it’s impossible. As the other person said, the issue is scale.