• 7 Posts
  • 255 Comments
Joined 3 years ago
cake
Cake day: June 23rd, 2023

help-circle

  • Download all existing literature to build a library for preservation and you’re called a pirate.

    Said library contains petabytes of the exact text of each and every piece of literature.

    Download all existing literature from aforementioned library to train an LLM and you’re a tech innovator.

    Said model contains gigabytes of a bunch of weights that can never go back to the exact words of the book.

    What a strange world we live in.

    It’s not strange at all. It’s degrees of compression. You compress a JPEG to the point that it’s unrecognizable, and it’s no longer breaking copyright. It’s essentially like trying to write a book you just read based on memory.


  • This is a deliberate oversimplification to try to excuse derivative and copied works of artists who have had their art stolen.

    It’s not. You misunderstand both copyright law and how LLMs work.

    Models are GBs of weights, typically in the 4GB to 24GB range. LLMs do not look at a picture and then copy that picture into the model. There’s not enough disk space to do something like that. It’s used for training, adjusting weights here and there, based on how the image links to the description. You can’t just say “recreate the Mona Lisa” and have it give you a pixel perfect copy of the original.

    When you do it, it’s copyright infringement.

    It’s not copyright infringement to copy a style. People do it all the time. You wouldn’t believe the amount of times I’ve seen something that I thought was some unique style, and thought that one artist did it, but it turns out it’s just another copycat artist “inspired by” the more popular artist.

    Because that’s what people do to something unique, or even remotely rare: Copy it a thousands times and drive it into the ground until you’re fucking sick of it.


  • If looking at a picture is stealing, then I’m doing so every day when I browse a web page or Google Image Search.

    Energy needs are a concern, but it was never a new problem. Our energy needs have been ramping up ever since we learned how to make electricity.

    The ones who “contribute to human culture” are the 1% who are lucky enough to make a career out of making art or making music or whatever other creative talent they had. The problem is oversaturation, not AI. AI makes the problem worse, but so does the Internet and every other technological leap we’ve seen.

    “Once it runs out of new content to plagiarize it will be unable to produce anything new.” Sounds like humans in a nutshell. Good artists borrow, great artists steal. Creativity itself is not sustainable at the rate we consume it. Every new thing is drilled into the ground, and beaten into a bloody pulp, until we find the next new thing. This is not a new problem.

    Capitalism is the enemy of humanity. Capitalists wield AI as a weapon, and we treat it as a scapegoat. We think that we can just get rid of AI, and then the enemy is gone. But, AI isn’t going away, and the same enemy we’ve always had still exists.

    Use it to your advantage. Use local models. Support open source LLMs. The biggest failure of rich capitalist assholes is sheer, absolute overconfidence and an inability to relate to the people they are trying to fleece.


  • I think we, as a society, need to do a better job separating out the real issue. The real issue isn’t AI. The issue is laziness. It’s the “slop” part, not the “AI” part.

    This is just CGI arguments all over again. People fucking hated CGI back in the 90s and 2000s. They hated how it was a crutch for VFX, hated how people wouldn’t bother hiring an animal to put into a simple scene, but they’d spend $10K to make a CGI sheep for a few seconds. Practical effects were suddenly novel. People praised Mad Max: Fury Road for its practical effects, but completely ignored the fact that Fury Road very much had CGI effects throughout.

    And that’s the secret: people stopped talking about CGI when it became invisible. If you can’t tell it’s CGI, then CGI has done its job. If you can’t tell it’s AI, then AI has done its job.

    But, quite often, you can tell it’s AI, because lazy hacks pretend it’s supposed to replace things that it’s not made for. They spend five minutes trying to generate something, and call it “good enough”. The creative art/video models are getting there, but they aren’t there yet. It still requires a ton of work to get certain styles out of the uncanny valley, and inpainting isn’t perfect. Voice models are okay, and better than the old TTS ones, but they don’t know how to act out a scene well enough. 3D modeling might get somewhere, but it shouldn’t be used for primary characters.

    This hype train needs to crash into a brick wall, so that we can use it in a more reserved manner. Some companies are quietly doing so, but that’s not what pushes the headlines.



  • A while back, I told myself that I wasn’t going to watch any YouTube video over an hour. But sometimes, really good YouTubers that I respected ended up putting out great two-hour material, like Folding Ideas or RLM or Grimbeard, and I just ended up watching these things, anyway. I might watch it for an hour, switch over to something else, and then watch the rest later. I just did that with the RLM Christmas video they just put out a few hours ago. It’s a hell of a lot better than the fleeting TikTok garbage that is geared towards maximum overstimulation and minimum education.

    But, I also watch a lot of YouTube. And I still have my limits. I don’t understand these videos that go up to 5-6 hours. That’s just a lack of restraint in the editing department in my opinion.