• 27 Posts
  • 40 Comments
Joined 8 months ago
cake
Cake day: July 18th, 2024

help-circle
  • None of what you said is really contradictory to anything I said. Everything you just said makes perfect sense. I said it looks like a human did a bunch of them, and did a pretty excellent / creative job, and then that gen AI did some of the vast selection of others.

    I think I laid out pretty clearly why I think that. Jennifer Daniel didn’t make the fairy picture with a square sunset for a head and also a square sunset on the tip of its wand. For another thing, there are about 186,000 combinations of 2 of the 610 emojis on offer for this tool. It seems unlikely to me that any single human being would do every single one. It would start to multiply into years of full-time work time spent on them pretty quickly, even with some automation, and there are clearly AI tools that can fill in a bunch of the non-critical-to-get-perfect ones, so why not. Some seem clearly likely to be from a human, some sort of look like automated templates that aren’t gen AI (like the alternatives next to a “downward chart trend line”), and some are gen AI.

    Anyway, I’m not trying to argue with you. I agree with most of what you said including that the human-generated ones are awesome, which is why I posted this.


  • He didn’t make himself president. He had some natural talents that helped the process along, but being an angry idiot which means he could connect emotionally with a lot of American voters who are also angry idiots was one of them. It’s in no way a counter argument for him being an idiot.

    If you fire a sedated turkey buzzard into an airplane engine, that doesn’t make it a skilled mechanic, even if it then makes unprecedented changes to the characteristics of the airplane.




  • Except, it is not a template. Even leaving aside that “put the sunset image in a square and make it a fairy’s head” would never be the template that any human decided to use to combine a butterfly with anything, look at this:

    That’s moon + butterfly overlaid with 50% opacity over sunset + butterfly. It’s a different fairy image, not just a different scale, but a different shape with trivial differences. Which there would be no reason at all for other than the generator getting rerun with new parameters for the different input images. It’s gen AI.


  • It seems like you and maybe at least one other person are just not grasping what I am saying here.

    A human did not make that decision. Sounds like it’s possible that maybe you are reading a lie that Google is telling you, and believing it is truth, I don’t really know, but yes they’re using generative AI for at least some of them.


  • It sort of looks to me like some of them are human-created (or at least human-curated), and some of the crappier or more nonsensical ones are created by generative AI.

    I couldn’t find any ones that it would refuse to do – maybe it’s recently started using an AI image generator for those ones?

    Edit: Okay, now I am sure that it’s using generative AI. I’m not trying to talk smack about the majority of them that are pretty reasonable and even sometimes creative, but also, sunset + butterfly yields:







  • they aren’t authorized to speak to the media about the looming crisis

    What the fuck un-American nonsense is this?

    I do get it. We’ve normalized the “employer” to “employee” relationship, where you’re sort of a free person but also sort of a slave, and that’s carried over into government service. Fuck that though. You are a person. You’re allowed to talk if you want to, and any separate person who’s trying to tell you they are the one in charge of that decision is probably a big piece of dookie at heart.








  • I don’t know, man. Maybe it is lies. Cory Doctorow and Louis Rossman are not just random idiots on YouTube, but they are just as susceptible as anybody to reading some hearsay and repeating it.

    On the other hand, the claims are very specific. “When you install third party toner, on this particular model of Brother printer, it stops respecting the calibration settings meaning that your colors are misaligned and color prints will look weird.” That’s pretty verifiable. It doesn’t mean it’s true, but it’s also actually not directly contradicted by anything in the Ars article. The article just says that Brother said you can use third party toner. Which everyone (as far as I saw) was agreeing that you can. Your prints will just look bad, after a particular firmware update, for reasons that are completely separate from any theoretical issue with the toner being third party.

    IDK. I will not claim to know the truth here. But I’m not convinced by what I have seen so far that the story is definitely bullshit. And the whole thing of sidestepping the issue of “the printer starts refusing to apply color calibration”, and just talking in generalities, is also a little suspicious. That has nothing to do with third party toner being lower quality or something.




  • I think a lot of it comes down to finding a little space with specific like-minded people. There’s the whole “dark forest internet” theory that talks more about it and what some of the solutions can be.

    Mastodon seems to do a better job of this. Specifically because you have to sort of put some effort into building up a little network and finding people you like interacting with, the quality of the interaction is a lot higher. There aren’t these handy megaphones laying around where any given person can broadcast to everybody at any time, which then means that you have to have stressed and hassled moderators keeping up a steady stream of people which they are kicking out or muting, because they did something ostensibly bad.

    Content moderation on Mastodon is its own whole issue, but I think the core principle is, develop solid connections instead of just random interactions with random people and it will be much better.

    https://www.youtube.com/watch?v=QEJpZjg8GuA