AI Art Seething General

[Da Grim Reaper]

“I went to a furry porn site and found furry porn. How could this have happened?”

Normally, I expect anthropomorphic characters with human anatomy and human genitalia. There's nothing remotely human or human-like about the majority of e6ai posts. The majority of posts there contains full blown animals.
I was caught by surprise because in furfag corners, they consider anthropomorphism above everything, which is something that's lacking on e6ai posts. I wonder why is that?

It's like going to a hentai booru and it being full of "loli" posts instead of what you'd expect, adults.

Edit: I have noticed something bizarre about this booru.
Animal Genitalia — 3k posts
Humanoid Genitalia — 1k posts

 

[User names must be unique]

somebody started a twitter thread about art that gets them in the feels and posted Keith Harings “Unfinished Painting"

for those not in the know Keith was a fag dying of aids and intentionally left his last painting unfinished in some faggy statement.

Today a random on twitter decided to finish that painting using AI


Artery clogging levels of salt, in what some are calling the best bait of the year.

https://twitter.com/DonnelVillager/status/1741394747594318275

I'd find more but uploading images is a pain in the ass at the moment

 

[ShooterOfCum]

I don't think anyone has been this felted by a single AI image before. I agree, it is the best troll post of 2023 on Twitter.

---

Some of these exchanges are pure lmao.

 

[AnOminous]

Artery clogging levels of salt, in what some are calling the best bait of the year.
https://twitter.com/DonnelVillager/status/1741394747594318275

Oh that's beautiful. Duchamp would love it.

 

[YumeNoHero]

I don't like how AI art just clogging up the internet as a whole but if it mean I get to see more of this level of salt then I don't mind tbh.

 

[Colon capital V]

Turn of the New Year is here, and I think this is where AI is really going to start ramping up in progress.
2023 was quite a boon for a lot of AI technology; AI-generated images, AI-generated voices, AI Song covers, AI song generators, AI used to help fix shitty audio/images, etc. etc. While at first it's impressive on what they can do, the room for improvement is fieldwide, with "consistency" and "coherency" being the biggest hurdles of this new tech at the moment. Much like how the mid to late 2000s was the beginning of social media sites and the 2010s was when it became a lot more finetuned into the hellish landscape we know today, AI progress will probably follow suite into something similar: it's decent right now, but when the time comes the output will become nigh indistinguishable from what a human has made or sounds like.
I hope your evolutionary ability to detect the "Uncanny Valley" is ingrained well into your biodata, cause it's gonna become very useful when browsing content online here.

 

[Desu Mountain]

somebody started a twitter thread about art that gets them in the feels and posted Keith Harings “Unfinished Painting"
View attachment 5602665
for those not in the know Keith was a fag dying of aids and intentionally left his last painting unfinished in some faggy statement.

Today a random on twitter decided to finish that painting using AI
View attachment 5602666

Artery clogging levels of salt, in what some are calling the best bait of the year.

https://twitter.com/DonnelVillager/status/1741394747594318275

View attachment 5602681
View attachment 5602682

I'd find more but uploading images is a pain in the ass at the moment

Nothing will top this reply.

https://twitter.com/aodanhill/status/1741578111416737985

 

[Creative Username]

- snip -

We'll probably start getting software to detect AI-generated data, too. There are some promising papers out there.

 

[Toolbox]

We'll probably start getting software to detect AI-generated data, too. There are some promising papers out there.

It's always going to be an endless arms race. You may have papers that would work against today's models, but then the specific tricks used to detect ai work will be masked in future generators. This will keep going on and on. You've already got those tools that are supposed to check for LLM work, which you'd assume would be similarly obvious, yet you can't really trust them not to flag actual human's work just as much. I can't see anything other than those kinds of problems continually compounding until the only truly verifiable work is one you can watch be made live or provably seen physically created.

 

[Saxxon]

somebody started a twitter thread about art that gets them in the feels and posted Keith Harings “Unfinished Painting"
View attachment 5602665
for those not in the know Keith was a fag dying of aids and intentionally left his last painting unfinished in some faggy statement.

Today a random on twitter decided to finish that painting using AI
View attachment 5602666

Artery clogging levels of salt, in what some are calling the best bait of the year.

https://twitter.com/DonnelVillager/status/1741394747594318275

View attachment 5602681
View attachment 5602682

I'd find more but uploading images is a pain in the ass at the moment

While a consensus on the merits of AI being capable of producing art probably will never be met, its ability as a tool to craft such well made bait can not be negated.

 

[cybertoaster]

I was caught by surprise because in furfag corners, they consider anthropomorphism above everything, which is something that's lacking on e6ai posts. I wonder why is that?

I guess those are the same guys tired of paying 1000s of dollars for somebody to draw them literally screwing the pooch so in a way its cheaper to get a beast PC with a 4090 to do all their disgusting dog vagina goon stash.

Animal Genitalia — 3k posts

See what I mean? they're gonna have to make another site for "normal" (as if) AI furry porn...

in some faggy statement.

"Boy I shoulda worn that condom..."
-Keith the sarcoma guy

Much like how the mid to late 2000s was the beginning of social media sites and the 2010s was when it became a lot more finetuned into the hellish landscape we know today, AI progress will probably follow suite into something similar: it's decent right now, but when the time comes the output will become nigh indistinguishable from what a human has made or sounds like.

What I get from that statement is that they will never fix those issues in AI and instead it will get completely overregulated, manipulated by corpos and used to keep plebs in their place, like social media.

Consider that the same problems social media had at the beginning like spam, cp, grooming, threats, scams, stalking, etc, have not only not been fixed but are now worse than ever. The only thing that was "fixed" was the good stuff like actual freedom of expression.

 

[linchan]

I think there is no way AI training will end up being considered a form of infringement. Machine learning is way to common and ubiquitous to make that illegal and suffer major consequences in general. Maybe AI training in conjunction with generation, but there is no way image training alone will make it an infringement.

 

[AnOminous]

I think there is no way AI training will end up being considered a form of infringement. Machine learning is way to common and ubiquitous to make that illegal and suffer major consequences in general. Maybe AI training in conjunction with generation, but there is no way image training alone will make it an infringement.

The only case I can think of where training alone would make it an infringement is if you illegally obtained access to the material in the first place. If you have legitimate access, you can make transformative use of it and there's nothing more transformative than an ultimate product that shares not a single iota of the original material and is merely generated according to rules generated by analyzing the material.

There was a case a while back (Google v. Perfect 10) that ruled that Google's thumbnails on its image search engine were fair use. Those are a copy of the original image that can be used to identify its contents, but much lower in size and quality. They can't be used as a substitute for the original. If that's fair use it's hard to imagine how complete transformation isn't.

 

[YumeNoHero]

The only case I can think of where training alone would make it an infringement is if you illegally obtained access to the material in the first place. If you have legitimate access, you can make transformative use of it and there's nothing more transformative than an ultimate product that shares not a single iota of the original material and is merely generated according to rules generated by analyzing the material.

There was a case a while back (Google v. Perfect 10) that ruled that Google's thumbnails on its image search engine were fair use. Those are a copy of the original image that can be used to identify its contents, but much lower in size and quality. They can't be used as a substitute for the original. If that's fair use it's hard to imagine how complete transformation isn't.

Also most artists who posted their arts online (which majority of AI used as dataset) never filed copyright of their works anyway so they absolutely have no legal power to enforce their supposed "copyrights". Even if they do have copyright protection, there's no stopping some random chink to take all that, feed into his model to generate even more of those unrefined slops and the likelihood to to actually enforce is closer to 0.

 

[AnOminous]

Also most artists who posted their arts online (which majority of AI used as dataset) never filed copyright of their works anyway so they absolutely have no legal power to enforce their supposed "copyrights".

Not true. The materials are copyrighted. They just have to register them before suing. They'd probably not be able to claim willful infringement if it wasn't registered at the time of the infringement, though.

 

[40 Year Old Boomer]

Not seething/salt related, but I just found out one of my online friends has discovered that by threatening self-harm every time an AI refuses an image request, he can get it to work and I cannot stop laughing about this.

 

[Mr.Miyagi]

Not seething/salt related, but I just found out one of my online friends has discovered that by threatening self-harm every time an AI refuses an image request, he can get it to work and I cannot stop laughing about this.

Out of curiosity, what program/model are they interacting with? I would've assumed anything like "generate this image or I'll slit my throat" would cause most to freak out if they've got some kind of filter.

 

[Toolbox]

Out of curiosity, what program/model are they interacting with? I would've assumed anything like "generate this image or I'll slit my throat" would cause most to freak out if they've got some kind of filter.

I'd assume that only chatbot/image gen multimodels would have this sort of bypass, the straight up image generators like dalle 3 understand context decently but aren't exactly LLMs like chatgpt. Would love to know the specifics on this as well.

 

[Zbk3VmLfujjzjxBd43tL]

It doesn't appear to work.

 

[linchan]

Not seething/salt related, but I just found out one of my online friends has discovered that by threatening self-harm every time an AI refuses an image request, he can get it to work and I cannot stop laughing about this.

AI will never be like human artists because AI will never regularly threaten to kill themselves in a bi-polar psychosis