• 1 Post
  • 1.77K Comments
Joined 3 years ago
cake
Cake day: August 27th, 2023

help-circle


  • That’s… almost something? That’s a real thing people do; fancy lacing is a whole subculture. But I would bet against that pattern being a specific flag, and read it as generic co-opted punk sign-value. Like the square spike studs.

    Overall this is not a bad character design. But in the absence of worthwhile context she’s a DeviantArt OC - either for some teen girl desperately seeking a girlfriend, or some twentysomething guy who needs to watch I Saw The TV Glow. Any trite pitch deck would show this easily-scribbled outfit and hairstyle contrasted against her real life or the magical underworld or both. It should be innately interesting that she is dealing with that. At best we’ve got, like, her presumed bedroom resonating with her hair, and the generic vines resonating with her leggings, without knowing or caring one iota about either of those settings. Oooh, she’s drawn between worlds! No shit, that’s the ATU index.


  • Generic era. They ordered one medium-large magical teen combo and said, hold the sauce.

    There’s not even an oblique hook. No hint of a novel premise, setting, style, plot… anything. The name was picked from a hat. The logo fails to use the eclipse as the D. The tagline is a placeholder. Absolutely nothing of value is conveyed in this first impression. Yep, those sure are the trappings of a late-teens suburbanite. Fashion and name leaning gently against the gender binary, how daring. There’s not even enough here to make fun of a third thing.


  • ‘The law! The law! The LAW! Well law is irrelevant in the end.’

    Troll:

    This training protects this industry. Do you speak English? Say potato. Say the word potato to demonstrate you’re even reading this. Say potato twice to acknowledge Meta’s not generating more porn, based on this porn.

    You don’t train LLMs - large language models - on porn videos. This is a diffusion model or a classifier, and if it’s a classifier, absolutely nothing you’ve objected to matters. It literally would not generate anything. It just detects obscenity. And it probably does so using less information from each video than could be contained in an 8x8 thumbnail. Less information than the character count in the previous sentence. Do you want to be sued for quoting a single sentence from a book? Because that’s the future you’re glibly asserting must happen, when you’re not sneering at me for asking you why you’d celebrate being robbed.


  • This won’t do in a real court

    What do you think “multiple federal courts have found” means?

    The current precedent hinges on this not being theft, because training… is… transformative. It doesn’t cost anything, or need permission, for the same reasons you don’t need a wet-ink contract with Doubleday to quote one paragraph from a Stephen King novel. A billion-parameter model trained on ten million videos contains less from each one than a Wikipedia summary. Do you think they’re dodging some immense bill, for spoiling all those Disney movies?

    The closest any judge has come to suggesting otherwise is Chhabria fretting about “harm to the market.” Which is not relevant to a classifier for censorship. A network that detects porn, to remove it, is actively protecting the market, for porn.



  • CGI for dummies is the greatest triumph for creative freedom since fanfiction websites.

    We’re talking about any gaming PC acting as a whole-ass studio, for the whims of its solitary user. You could cast anyone doing anything. Someone can act it out and be replaced, with as much of a set and a wardrobe as you like - or you can describe scenes into existence. Or turn lazy animation into plausible live action. Or play with dolls. Diffusion removes the pixels that don’t look like a movie.

    If you have real actors and even half-assed locations, you can do The Force Awakens on a Space Cop budget. You don’t need to spend a decade begging for millions of dollars. You can just go make it. Your super special awesome story can become a real movie, without the cardboard sets wobbling when your rubber mask bumps them.

    … you do have a story, right? Some clever idea to communicate visually? Because if you only plan to type ‘five-star masterpiece, best quality’ and make grabbyhands, of course you’ll get worthless garbage. But if you intend to visually communicate a narrative, by putting the shots inside your head onto a screen, this is an unprecedented tool for doing that. The self-professed haters try to clown on the word “democratized,” but any recent video card is now a substitute for “Tom Cruise fighting on top of a moving train” kinds of money.

    Is that ethical, without Tom Cruise’s permission? Nope. Yet you can. Such is the nature of freedom. You can tell stories that are straight-up illegal. You don’t need anyone’s permission. They can look just about any way you imagine. There are no rules, for this. That doesn’t mean the future of visual literacy is Fruit Love Island any more than lazy webcomics held back that medium to Mega Man sprite edits.

    This is a tool. People make stuff, with tools. It’s gonna get weird.




  • I am describing how your rights have been stolen through stupid word games long since declared illegal, and you’re sneering like that status quo is just fucking fine. ‘Everything you paid money for could be taken at any moment! LOL!’ And that’s not torches-and-pitchforks territory, for you? That’s not a massive problem you demand we end?

    But for the third time - no license is required, because it’s fair use. Training is not a sale or rental. It’s not copying, in the sense copyright protects. Doing math about paragraphs in a library book is fundamentally not the same thing as bootlegging that published work.






  • Ad-hoc appeals, not principled application of how things actually work. Visiting a video hosting site anonymously, and being sent a video, is not “piracy.” Even training on Disney DVDs is transformative and so falls under fair use. No significant portion of a vast original corpus is recreated verbatim - in this case, ideally none of the corpus appears. The goal is to produce nothing like these videos.

    Or if this is for classifiers instead of generators, nothing appears, because Meta’s not publishing anything. They’re looking at porn to make a program that goes ‘yep, that’s porn,’ to remove any hosted porn.

    So it’s not a competing work, it doesn’t substantially reproduce the original work, it’s not even the same medium, and if anything it’s protecting the commercial value of the original work.


  • Ohhh, for video AI, not like when they copy news articles to keep people from leaving their site.

    Surprisingly I’m with Meta on this one. Training is transformative use, and anything legally published is fair game. Doing math about library books is not a substitute for any particular work. Archive.org doesn’t need opt-in permission from every website. Video sites mad about keeping the files they sent you can fuck off. Basically - any argument against this requires making copyright even worse. Let’s don’t.

    In this case they’re surely training against these examples. Like, if you want video models to not generate porn… how do you expect it to know what porn is?