Under US copyright law, only works created by humans can be copyrighted. Courts have (imho rightly) denied copyrights to AI-generated images.

My question is when do you think AI image tools cross from the realm of a “tool” (that, for example generates and fills in a background so an item can be removed from a photo) into the realm of “a human didn’t make this”?

What if an artist trains an AI so specialized it only makes their style of art? At what point do you think the images they create with it begin to count as their “work product”?

  • curiousPJ@lemmy.world
    link
    fedilink
    English
    arrow-up
    3
    ·
    7 months ago

    I’m pretty sure they would have to upload millions of pictures of their own creation first.

    From the YouTube guides in generating your own Lora models… Naa just a couple reference poses and it’s ready to go.

    • model_tar_gz@lemmy.world
      link
      fedilink
      arrow-up
      3
      ·
      7 months ago

      LoRA models still have the underlying fully trained base model underneath; it is not a complete replacement or complete modification of the model weights.

    • anarchost@lemm.ee
      link
      fedilink
      arrow-up
      1
      ·
      6 months ago

      What @[email protected] said, Sorry my original post was not so clear.

      I based my answer on the assumption that you would not be using somebody else’s art to train an AI. The Lora models are already brimming full of other people’s art.