• NaibofTabr
    link
    fedilink
    English
    615 hours ago

    Every single generative tool is built on theft. No one has a training database that only contains code they were given legal permission to use for that purpose. Moreover, if these tools were trained on any GNU GPL code, then the tool and all of its output should also be GNU GPL:

    if you distribute a derivative work or modification, you must provide the source code to those recipients under the same or equivalent license terms

    https://en.wikipedia.org/wiki/GNU_General_Public_License

    None of the companies developing these tools are obeying the terms of the licenses. The trained models are the product of theft.

    You cannot be a moral person and approve of the use of these tools, they are diametrically opposed.

    • AwesomeLowlander
      link
      fedilink
      English
      014 hours ago

      Given that you’ve presumably looked at open source code in the past, and got more proficient (however marginally) as a result, does all your future code now belong to the GPL?

      • NaibofTabr
        link
        fedilink
        English
        26 hours ago

        This comparison is invalid. Training a neural network algorithm is not equivalent to human learning. We are talking about data stored in machine learning models owned and controlled by multi-billion dollar corporations.

        It has already been demonstrated multiple times that original training data can be reproduced completely from models, so yes, they are data storage systems. When they reproduce code which they have previously stored, even only in part, that is a derivative work. Adding extra steps to the transcribing process doesn’t make it any less a copy of the original.

        • BJW
          link
          fedilink
          English
          1
          edit-2
          3 hours ago

          Only when the data in question is so obscure to the point that the original source is the only place in which the data exists; which is neither common, nor useful to regurgitate, in practice.

          It’s like if someone asked an artist to draw Trump, and then they did. It’s not that they studied what he looks like, but they’re familiar with that walking pusbag and there’s only the one, so it’s going to be a recognizable drawing.

    • BJW
      link
      fedilink
      English
      -214 hours ago

      Hard disagree.

      This is like saying anyone who has watched a Disney movie now owes Disney a check any time they illustrate anything, even entirely unrelated, because they’ve learned from their commercial products in the past. If they don’t pay Disney, then their creations are based on theft.

      • NaibofTabr
        link
        fedilink
        English
        16 hours ago

        Digital reproduction is digital reproduction no matter how many extra steps are added to the reproduction process. It’s just an algorithm that sorts through a collection of stored data to find specific pieces of data which best fit the keywords supplied by the user, then regurgitates the results that are the best match based on correlation.

        In spite of common meaning overlap and popular metaphors, the human brain is not a computer. Ask any neurologist.

      • @Einskjaldi@lemmy.world
        link
        fedilink
        English
        413 hours ago

        Fundamentally that’s a purely ethical decision about whether a machine doing something should be treated with the same understanding the same as a human would. But we don’t consider remembering something with a neurochemical storage the samd as taking a picture of something even if they’re basically the same.

        • BJW
          link
          fedilink
          English
          -16 hours ago

          That’s because taking a picture is an idetic copy. Machine learning is NOT copying, it’s learning - hence the name.

          • NaibofTabr
            link
            fedilink
            English
            06 hours ago

            This is a misunderstanding based on confusion between technical and colloquial terminology.

            A machine learning model “learns” information in the same way that a curve fitting algorithm “learns” the shape of a data set.

            This is not the same as the colloquial meaning of human learning. It is a mathematical process.

            • BJW
              link
              fedilink
              English
              14 hours ago

              That’s splitting hairs on definitions, with no change in meaning. It’s still not copying the data, and is far closer to a person learning than to a picture taken by a camera.