OpenAI now tries to hide that ChatGPT was trained on copyrighted books, including J.K. Rowling’s Harry Potter series::A new research paper laid out ways in which AI developers should try and avoid showing LLMs have been trained on copyrighted material.

  • dx1@lemmy.world
    link
    fedilink
    English
    arrow-up
    4
    ·
    edit-2
    1 year ago

    Ultimately it’s a propertarian scheme of ownership imposed onto the realm of concepts and ideas. The first person to successfully lay claim to an idea is given a monopoly on that idea for some number of years. A book, an invention, a melody. To secure profit for that individual, the entire rest of humanity is prevented access to the idea except under his terms, and the naturally free exchange of information is curtailed by statute to accomplish this, via the imposition of punishments for anyone who goes against this scheme. I do not think that’s defensible. That is to say, I don’t think humanity sees a net benefit from this way of doing things. Even some hypothetical 20-30% reduction in the generation of different kinds of creative works would be well offset by the benefit humanity sees from being able to access them, and the funds that would be going to the artist still could if people saw fit.

    Is this being used to stop the development of generative AI? Yes, literally the imprint on an AI of having parsed the works and understood them in some symbolic capacity, they want to curtail that. And the existing models that have already done that would likely be rendered illegal, setting the entire technology back a year or two.

      • dx1@lemmy.world
        link
        fedilink
        English
        arrow-up
        4
        ·
        1 year ago

        Insulin is ridiculously expensive because of monopoly status over methods on how to produce it (on top of any other laws restricting supply). I personally contributed to a project to create an open source method to make insulin.