• queermunist she/her@lemmy.ml
    link
    fedilink
    arrow-up
    1
    ·
    1 day ago

    Who do you think will be able to afford the training?

    Nobody. LLMs are already unprofitable now when it’s free from copyright restrictions, if they had to actually pay for the proprietary data they’re taking then basically all US-based companies will be unable to afford training. This pops the bubble.

    • Dr. Moose@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 day ago

      What about people who don’t care about IP law? Or trillion dollar companies? You think people just going to stop using LLMs? Lol

      • queermunist she/her@lemmy.ml
        link
        fedilink
        arrow-up
        1
        ·
        23 hours ago

        Trillion dollar companies want to make money, they aren’t just going to burn endless billions on super expensive and unprofitable tech that never turns a profit.

        And those massive data centers make it basically impossible for anyone to ignore IP law. Is Apple going to become an outlaw company? Or is some underground pirate server farm going to host an LLM? There’s no way to actually dodge the law on this for US-based companies.

        • Dr. Moose@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          13 hours ago

          Big tech absolutely will spend billions if Apple Intelligence or Gemini is the only legal commercial LLM, this type of moat is 100% worth it. It’s literally the dream.

          You got it the other way around. There’s no way to actually prove LLM used copyrighted material if it’s not a US based model and what then? Do you have a commercial allow list of models thst are legal to use? That would be completely unenforceable and collapse american tech advantage.