• FiniteBanjo@lemmy.todayOP
    link
    fedilink
    English
    arrow-up
    0
    ·
    edit-2
    5 months ago

    Using machine learning to clone voices required building upon millions or maybe even billions of lines of dialogue to reach the current point, and only now that it is here can you easily approximate a person’s voice without them even knowing. So yes, it does magically generate sound with minimal input, that is how that works.

    The machines that are currently used to shape and paint ceramics are not Machine Learning models. They’re simple and precise automations. Like a program that packages audio to .mp3 format. The equivalent to that would be a machine that designs the vase based on thousands of older examples, designs the packaging, and designs the machinery that prepares it. It’s going to do a shit job that negatively impacts consumers but in the process it displaced thousands of workers in one go, so enjoy the profits.

    • richmondez@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      5 months ago

      You missed the point of what I said. Machines that manufacture goods put many people out of the job and yet you now very few people think that is an issue. At the time however the same kind of arguments I see made against LLMs putting people out of work were being made about these machines making soulless products that missed the human touch. LLMs are just a new tool we’ve invented to make life easier for ourselves. In time the same thing will happen with LLMs once the hype dies down and they just become part of the tool sets we all use without thinking about it.