Internal emails, Slack conversations and documents obtained by 404 Media show how Nvidia created a yet-to-be-released video foundational model.

Nvidia scraped videos from Youtube and several other sources to compile training data for its AI products, internal Slack chats, emails, and documents obtained by 404 Media show.

When asked about legal and ethical aspects of using copyrighted content to train an AI model, Nvidia defended its practice as being “in full compliance with the letter and the spirit of copyright law.” Internal conversations at Nvidia viewed by 404 Media show when employees working on the project raised questions about potential legal issues surrounding the use of datasets compiled by academics for research purposes and YouTube videos, managers told them they had clearance to use that content from the highest levels of the company.

  • ArkyonVeil@lemmy.dbzer0.comEnglish
    132·
    3 months ago

    I love it when marketing manages to spin Armageddon levels worth of copyright infringement into “spirit of the law” just because a program is magically called “AI”. Machine Learning is just pattern recognition software.

    Software that runs on data assembled from petabytes of copyrighted information… And then promptly resold to us.

    We may decide later on if it’s okay to do this. But I’m pretty sure that if it wasn’t for the labels we’d have legal WW3 happening right about now.

  • hoshikarakitaridia@lemmy.worldEnglish
    9·
    3 months ago

    in full compliance with the letter and the spirit of copyright law

    That is some real semantic acrobatics. The law is supposed to follow societal norms and reflect boundaries accordingly. Yeah, AI laws take time, and obv there hasn’t been enough legislation done. That said, the EU for example already has a law for AI but the member states need to adapt that into national laws now.

    There is law here. And even though I’m sure what they are doing rn will be illegal or at least very heavily regulated in the future, they might be doing something illegal today. Depending on how eager governments are to litigate, this might already get dicey in the coming months.

  • AdamEatsAss@lemmy.worldEnglish
    4·
    4 months ago

    Technically a human could watch all the videos on YouTube. But if a human were to cut together bits of each video to make a new one that would probably be an infringement. But if a human saw a video and wanted to recreate it that might be allowed depending on how close to the original they got. It points out some glaring issues with modern copyright law. I don’t know what the solution is, but if a multinational tech giant is doing it, it should probably be illegal.

  • tal@lemmy.todayEnglish
    4·
    3 months ago

    I wonder if that’s actually fast enough to keep up with the rate at which video is being uploaded?

    • Kevin@programming.devEnglish
      3·
      3 months ago

      I remember this old website that the YouTube team had made which visualised the amount of video time getting uploaded per day on YouTube over the years of its existence, and it was on the order of several years per day or something. Gotta find that site again

  • werefreeatlast@lemmy.worldEnglish
    51·
    3 months ago

    Hey Nvid-AI, solve all world hunger please!

    Okay removing all homeless people…

    Noh I mean feed them!

    Okay feeding them homeless…