Mark Zuckerberg gave Meta's Llama team the OK to train on copyrighted works, filing claims | TechCrunch
techcrunch.com
Meta CEO Mark Zuckerberg gave Meta's Llama team approval to train on copyrighted documents, according to a new court filing.
  • Grimy@lemmy.world
    4·
    16 hours ago

    Won’t the goverment step in and rid us of them. Force them to pay, so meta can give millions to reddit, getty and all the other data brokers and just keep doing it.

    You guys are responding to openai propaganda. The open source scene dies if you can’t train freely. Meta was the first player to actual dump their stuff, even if it was just to fuck over Microsoft and Google.

    I can train and run their model on my shit laptop, as well as the hundreds of fine tunes. I can talk dirty to an uncensored fine-tune which are explicitly banned on actual platforms. I can fine tune it on my own data or wtv else as I please.

    Getting mad and giving the goverment green light to legislate for the benefit of copyright laws will fuck all of us. At least with an open source scene, you can turn around and compete against your boss if he ever fires you because you’re made redundant because of AI.

    If only the companies able to afford the data can build models (and there’s like 3), you end with a soft monopoly with subscription models that can eventually actually replace people and are priced to be just low enough to make economic sense but way too high to afford on an unemployment cheque.

  • JackGreenEarth@lemm.eeEnglish
    213·
    19 hours ago

    Llama is open source, so they should be allowed to train on public data if they’re going to release it to the public.

    The problem is with them using non licensed data to train proprietary models, not models in general.

      • JackGreenEarth@lemm.eeEnglish
        412·
        19 hours ago

        Well, the licence tries to prohibit people doing various things with it, but the model is open weights. Anyone can physically run it on their hardware, not something they can do with ChatGPT or Claude for example. You’re right, I shouldn’t have implied it was fully open source, but at least it only tries to legally, rather than physically, prevent people running and modifying the model themselves.