• flatbield@beehaw.org
    link
    fedilink
    arrow-up
    6
    ·
    edit-2
    1 year ago

    I feel two ways about it. Absolutely it is recorded in a retrieval system and doing some sort of complicated lookup. So derivative.

    On the other hand, the whole idea of copyright or other so called IP except maybe trademarks and trade dress in the most limited way is perverse and should not exist. Not that I have a better idea.

    • FaceDeer@kbin.social
      link
      fedilink
      arrow-up
      3
      ·
      1 year ago

      Absolutely it is recorded in a retrieval system and doing some sort of complicated lookup.

      It is not.

      Stable Diffusion’s model was trained using the LAION-5B dataset, which describes five billion images. I have the resulting AI model on my hard drive right now, I use it with a local AI image generator. It’s about 5 GB in size. So unless StabilityAI has come up with a compression algorithm that’s able to fit an entire image into a single byte, there is no way that it’s possible for this process to be “doing some sort of complicated lookup” of the training data.

      What’s actually happening is that the model is being taught high-level concepts through repeatedly showing it examples of those concepts.

      • flatbield@beehaw.org
        link
        fedilink
        arrow-up
        4
        ·
        edit-2
        1 year ago

        I would disagree. It is just a big table lookup of sorts with some complicated interpolation/extrapolation algorithm. Training is recording the data into the net. Anything that comes out is derivative of the data that went in.