AI Art Seething General

I only just realised that the people who develop the software (in general) are people who want their creation (their code) to be in as many things as possible and to contribute to a forward progression of computing, totally undisturbed by the idea that they may be credited simply as one of tens of thousands of names on a supply chain list of open source software that go into making a modern program.

And artists (in general) do not feel that way.
I used to write applications in Spring. Spring is one of the most popular Java frameworks. It isn't by Oracle, it is open source. I never knew or looked up who made it or contributed to it. Most people who use it never will. It's a work of art in itself, and it runs countless applications around the world.

Most modern artists of massive popularity draw pornography usually shared on image boards or special sites. Nobody cares about who they are jerking off to most of the time.

It's the same thing really. The art fills a need. The need eclipses the person who fulfills it.
 
So there are two things Anti-AI art believes.
1. On art-sharing websites, images that are AI-assisted need to be labeled appropriately.
2. AI training on AI-generated images will yield worse and worse results.
So with these two things being true, for the benefit of the anti-ai art side, shouldn't they encourage websites not put a label distinguishing what images are AI or not, that way if someone scrapes a bunch of images they get a poisoned training set?
 
The thing about AI training on AI generated images making it worse is only if the result it is being trained on is bad. For example, if someone feeds a bunch of AI art into a training data set and all of it has fucked up hands, yes, the hands will continue to get worse and worse as it trains more. If it is curated and done with good results only, however, no, there's no inherent reason why AI art will get worse by studying itself. Hell, I've personally trained LoRAs on entirely AI generated characters before. Takes a lot of effort and Photoshop to get a good, large data set, but it is 100% possible and made reproducing the character significantly easier and more consistent.

Basically, it would be no different from training it on bad "real" art. It gives out what you put in. You put in shit, you get shit back.
 
The thing about AI training on AI generated images making it worse is only if the result it is being trained on is bad. For example, if someone feeds a bunch of AI art into a training data set and all of it has fucked up hands, yes, the hands will continue to get worse and worse as it trains more. If it is curated and done with good results only, however, no, there's no inherent reason why AI art will get worse by studying itself. Hell, I've personally trained LoRAs on entirely AI generated characters before. Takes a lot of effort and Photoshop to get a good, large data set, but it is 100% possible and made reproducing the character significantly easier and more consistent.

Basically, it would be no different from training it on bad "real" art. It gives out what you put in. You put in shit, you get shit back.
I vaguely recall there has been research that has intentionally inserted AI generated or CGI images into a training set to fill in gaps to improve quality. This was years ago, probably before "LoRAs" were a thing, and may have mattered more when everything was more primitive.

At the scale of billions of images, there are always going to be problems with curation. The major models seem to continue improving, whatever they're doing behind the scenes.
 
Is this the place to post my masterpieces?
Colonel w Twain.jpg

A very rare photo of Colonel Sanders (1890-1980) dining with Mark Twain (1835-1910) on a Mississippi riverboat

Whales and Kraken and Zeppelins.jpg


Coon burglar.jpg
 
Tennessee Elephants.jpg

Abraham Lincoln accepted the King of Siam's offer of war elephants for the Union cause. They got loose and have become an invasive species across the Upland South. Now Hohenwald, TN (home of a real-life elephant sanctuary) is rebuilding itself in Reconstruction with elephant labor to assist construction and plowing the corn fields.

Red Bandit.jpg

The Red Bandit was a real-life apparation of Arizona. The Union introduced camelry, they got loose, went invasive. Had to hunt them down. One had a rider on him that died suddenly, probably a heart attack or something. He decayed in such a way - probably a mummy-like state from the dry heat - that his corpse remained sitting atop that camel for decades, occasionally observed by fearful farmers that considered an apparition.

Now he rides again, in a psychedelic haze fueled of peyote and mezcal.

OIG.jpg

The Dragon of Dixie.
 
View attachment 5430973
Abraham Lincoln accepted the King of Siam's offer of war elephants for the Union cause. They got loose and have become an invasive species across the Upland South. Now Hohenwald, TN (home of a real-life elephant sanctuary) is rebuilding itself in Reconstruction with elephant labor to assist construction and plowing the corn fields.

View attachment 5430977
The Red Bandit was a real-life apparation of Arizona. The Union introduced camelry, they got loose, went invasive. Had to hunt them down. One had a rider on him that died suddenly, probably a heart attack or something. He decayed in such a way - probably a mummy-like state from the dry heat - that his corpse remained sitting atop that camel for decades, occasionally observed by fearful farmers that considered an apparition.

Now he rides again, in a psychedelic haze fueled of peyote and mezcal.

View attachment 5430985
The Dragon of Dixie.
I was going to suggest this thread: https://kiwifarms.st/threads/stable-diffusion-novelai-machine-learning-art.130730

Also... THUMBNAILS (and spoilers so they're hidden)
 
itzmoepi returns with an unhinged, actually sexist rant

View attachment 5414606
Explanation 4: it requires enough technical know-how to go to a github page and follow the instructions there, searching online for help if you have problems, and then dealing with further technical tweaking and esoteric indie UI afterward, therefore most women are at a total loss even beginning to figure out all this techie shit
 
Explanation 4: it requires enough technical know-how to go to a github page and follow the instructions there, searching online for help if you have problems, and then dealing with further technical tweaking and esoteric indie UI afterward, therefore most women are at a total loss even beginning to figure out all this techie shit
Also, if they're doing it at home instead of "teh cloud" then you need a moderately powerful desktop or laptop. Which is unlikely except for techies or gamers. (Or people who bought the fastest and most expensive Mac because it had an Apple on the case.)
 
itzmoepi returns with an "apology" (for some definition of "apology" anyway)

View attachment 5435684

Then he recants it a mere few hours later in a fit of rage...

View attachment 5435686
Don't you love when people use places like Reddit as a personal blog, and then get assblasted when other people have the audacity to comment on it in disagreement? I think this guy would be a lot happier if he just bought a journal to write in, but I guess that doesn't come with e-validation. I mean, for fuck's sake, you're deliberately posting in a place dedicated to arguments, what kind of asspats did you expect?
"If StabilityAI dies, IT'S OVER FOR YOU AI BROS!", as if you wouldn't have another company/university research group eventually come along and throw a half-mil (or less, depending on how the tech advances) to train up a new model.
"If this replaces all human art, I'm KILLING MYSELF" This stuff isn't going to kill of human expression any more than Word killed off people picking up pens and writing their thoughts. It's absurd to think this is going to cause an integral part of human beings to just shrivel up and fall off like a gangrenous limb.
 
The fingerprints on a letter to Congress about AI
Sy Damle, a Washington lawyer representing OpenAI in copyright lawsuits, covertly organized a letter from tech groups and academics urging Congress to avoid new laws on AI and copyright.
Corynne McSherry, legal director at the Electronic Frontier Foundation, said Damle was “involved in circulating and revising the letter after input.” R Street spokesperson William Gray said that Wayne Brough, the organization’s technology lead on copyright policy issues, “was invited by Sy Damle to sign onto the letter, and after review, decided to add our name.”
 
For some reason, Twitter/X decided to recommend this Tweet to me. Furry artist demanding that people stop sharing AI art because showing it to other people is killing his career. Playing with it on your own is fine, but don't you ever post it online. So surely he must be really struggling, right?

Oh wait nope. Not only is he selling porn calendars for $60 a pop, his commission queue is absolutely packed, with at least 10 overdue commissions by his own schedule, and 32 pending total.

But yeah, AI is totally killing him. Won't you think of the porn artists who have so many commissions they can't keep up with them?
 
Last edited:
Back