• grue@lemmy.world
    link
    fedilink
    English
    arrow-up
    18
    arrow-down
    3
    ·
    10 months ago

    Nevermind paying for content; AI firms should be required to abide by the terms of copyleft-licensed training data. In other words, all output of an AI trained on even a dataset containing even a single copyleft work should be required to be copyleft itself.

    • Rogers
      link
      fedilink
      arrow-up
      4
      arrow-down
      2
      ·
      10 months ago

      Yeah that won’t work. Any country that lets their companies use whatever data will flat out have better models. Potentially meaning grater economic output of the whole country if ai is as big as i think. Unfortunate but i don’t see an alternative yet unless they make it so you can use amy data but models have to be free

      • grue@lemmy.world
        link
        fedilink
        English
        arrow-up
        3
        arrow-down
        1
        ·
        10 months ago

        The alternative is to take the good training data and then accept the output being copyleft without whining about it.

      • a2800276@lemmy.world
        link
        fedilink
        arrow-up
        1
        ·
        10 months ago

        Doesn’t that argument apply to any instance of ignoring intellectual property. Books, records and movies will also be cheaper in countries that let companies do what they want. Medicine would be more accessible, ignoring patients will greatly accelerate innovation in countries where permitted…