AI Derangement Syndrome / Anti-AI artists / Pro-AI technocultists / AI "debate" communities - The Natural Retardation in the Artificial Intelligence communities

[Hood-ass Mulatto]

I don't know what your take is in this, but in this situation, the original AI-generated pic looks better because it's easier on the eyes. Her remake is worse because there's too much going on in it.

0:34

Lmao I also noticed the shading is a very light grey. The battle has been lost even before it began.

Anyway, my opinion is that TikTokers and many other internet artfags would eat the remade version up is because of the kawaii anime style. Especially TikTok artists, because any art style that isn't anime, cute, soft, or realism somehow gets trashed.

 

[sex haver]

Can you elaborate on why it's bogus?

Because it just doesn't work.

• Easily defeated. I remember things as simple as downsizing the image to 99% and then upscaling it 1% with typical upscalers was enough to remove any of the artifacts Glaze/Nightshade gave it. There were also people applying a like, 2-3% blur which did the same thing, removed the "poison" and left an image perfectly readable by an AI.
• Requires mass effort. Even if it did work, you would need a huge majority of online artists and photographers being sure they run everything through Nightshade before before posting it. If not, then there's nothing stopping AI training from simply discarding the handful of poisoned art/photos and training on the other millions of un-poisoned images.
• The training datasets already exist. You can't poison the billions of images already saved and being used for training.
• The models already exist. Even if you poisoned every single new image ever posted to the internet, the model I have downloaded on my PC right now will never be affected. There is literally nothing anyone can do to stop me from continuing to generate images using the current or previous models.

 

[a-Lager]

I don't know what your take is in this, but in this situation, the original AI-generated pic looks better because it's easier on the eyes. Her remake is worse because there's too much going on in it.

1000082856.mp4
Wyświetl załącznik 9182959

Lmao I also noticed the shading is a very light grey. The battle has been lost even before it began.

Anyway, my opinion is that TikTokers and many other internet artfags would eat the remade version up is because of the kawaii anime style. Especially TikTok artists, because any art style that isn't anime, cute, soft, or realism somehow gets trashed.

Tumblr artists, the original AI except the only prompt is to make the image gay and brown

 

[Runch]

I don't know what your take is in this, but in this situation, the original AI-generated pic looks better because it's easier on the eyes. Her remake is worse because there's too much going on in it.

Wyświetl załącznik 9182959

Lmao I also noticed the shading is a very light grey. The battle has been lost even before it began.

Anyway, my opinion is that TikTokers and many other internet artfags would eat the remade version up is because of the kawaii anime style. Especially TikTok artists, because any art style that isn't anime, cute, soft, or realism somehow gets trashed.

I fucking HATE art where the mouth goes all the way down so they have no chin. I'm sure it originates from anime where to save money they only animate the mouth, but uh oh, in this scene we need the mouth to get REALLY big because they're yelling, so fuck it rather than redrawing the face with a lower chin to accommodate the bigger mouth, just go all the way down with it. Just completely lazy and stupid. I hope her mom despises it.

Because it just doesn't work.

• Easily defeated. I remember things as simple as downsizing the image to 99% and then upscaling it 1% with typical upscalers was enough to remove any of the artifacts Glaze/Nightshade gave it. There were also people applying a like, 2-3% blur which did the same thing, removed the "poison" and left an image perfectly readable by an AI.
• Requires mass effort. Even if it did work, you would need a huge majority of online artists and photographers being sure they run everything through Nightshade before before posting it. If not, then there's nothing stopping AI training from simply discarding the handful of poisoned art/photos and training on the other millions of un-poisoned images.
• The training datasets already exist. You can't poison the billions of images already saved and being used for training.
• The models already exist. Even if you poisoned every single new image ever posted to the internet, the model I have downloaded on my PC right now will never be affected. There is literally nothing anyone can do to stop me from continuing to generate images using the current or previous models.

Also because Glaze and Nightshade targeted old diffusion models specifically, and new models are being trained in new ways that don't interpret the perturbations in the same way. So if you're doing anything at all, you're doing it to ancient models that few people use anymore. You'd have to start over from scratch and find new weak points in newer training methods, which by the way tend to be more proprietary and protected now.

The absolute worst kind of security is security applied to individual objects that applies to all of them, because if the security is ever broken then none of those objects have security anymore.

Like, so you have a house with a lock on your door and all your stuff inside the house. Someone gets a copy of your key. You change the lock and now they can't get in and your stuff is protected again.

Now imagine you live in a tent in a field with no door. You put individual locks on all your books and video game cases that open with the same key. Now if someone gets a copy of your key, you have to change ALL the locks in order to protect them again. Let's be honest, you're not gonna go to the effort.

Or a real world example: game consoles that have a master key that lets you do whatever you want with the system, and then one guy gets ahold of it and all of a sudden all Wiis or PS3s are wide open. They can't fix it, every device out there is unlocked forever.

That's what Glazing your image is doing. As soon as Glaze doesn't work to protect your image anymore, all that effort was completely useless. Start over using Glaze 2.0 on millions, billions of images again.

 

[Gamepads]

Requires mass effort. Even if it did work, you would need a huge majority of online artists and photographers being sure they run everything through Nightshade before before posting it. If not, then there's nothing stopping AI training from simply discarding the handful of poisoned art/photos and training on the other millions of un-poisoned images.

Even that was too much effort. Most struggled to get a local setup running, and lost their shit when service providers started shifting the compute cost onto the artists. What do you mean I have to pay? Plus, I don't think they made the logical jump about the compute requirement and datacenters...

Anyways, putting a gigantic ugly watermark is much easier for the virtue signal.

 

[Danann]

Reminder that Sebastian Wolff, Materia Collective CEO and tryhard dick sucker of Toby Fox (and Mr.Pig too of Pizza Tower), still wishes for the government to somehow control what should and shouldn't go into an AI algorithm because he is a massive faggot who thinks native americans existed since the beginning of time on where his company sits. Also probably wants the entire process to be made up of humans who are absolutely able to keep up with the massive demand and consent every last creator of the works they are submitted. (They hate anyone but themselves...but also pretend that they love everyone but themselves)

 

[Miss Violet Beauregarde]

Reminder that Sebastian Wolff, Materia Collective CEO and tryhard dick sucker of Toby Fox (and Mr.Pig too of Pizza Tower), still wishes for the government to somehow control what should and shouldn't go into an AI algorithm because he is a massive faggot who thinks native americans existed since the beginning of time on where his company sits. Also probably wants the entire process to be made up of humans who are absolutely able to keep up with the massive demand and consent every last creator of the works they are submitted. (They hate anyone but themselves...but also pretend that they love everyone but themselves)

I'd like for him to try that to non-us countries.

Seriously though, why do anti-ai people only focus their efforts on the US, and only US? You'd think they'll go into the other European countries or try their stuff in Asia, but no, i only see serious seething on the US.

 

[YellowIsACoolColor]

Seriously though, why do anti-ai people only focus their efforts on the US, and only US? You'd think they'll go into the other European countries or try their stuff in Asia, but no, i only see serious seething on the US.

They are extremely lazy and stupid so the thought hasn't even occurred to them.

 

[Hood-ass Mulatto]

I'd like for him to try that to non-us countries.

Seriously though, why do anti-ai people only focus their efforts on the US, and only US? You'd think they'll go into the other European countries or try their stuff in Asia, but no, i only see serious seething on the US.

Because Trump bad. Trump and the US are the ones getting media coverage and the negative attention a lot. Do you expect their lazy asses to look beyond the United States and what the media covers?

I fucking HATE art where the mouth goes all the way down so they have no chin. I'm sure it originates from anime where to save money they only animate the mouth, but uh oh, in this scene we need the mouth to get REALLY big because they're yelling, so fuck it rather than redrawing the face with a lower chin to accommodate the bigger mouth, just go all the way down with it. Just completely lazy and stupid. I hope her mom despises it.

Moments like these make me laugh where the human artist is lazier than the AI. I remember a tiktoker saying that you will know if someome knows how to draw only because of anime instead of focusing on anatomy first.

Tumblr artists, the original AI except the only prompt is to make the image gay and brown

And fat, with freckles, sometimes with vitilgo, and looking like Steven Universe.

 

[Waifuwars]

I'd like for him to try that to non-us countries.

Seriously though, why do anti-ai people only focus their efforts on the US, and only US? You'd think they'll go into the other European countries or try their stuff in Asia, but no, i only see serious seething on the US.

I've mentioned before it's just not as visible to us because of the language and platform barriers. A ton of Asian artists have 'NO AI' in their Twitter bios (even though the ToS explicitly says their content will be trained on lol lmao). Koreans, Japanese, and even Chinese artists all sperg out about it (only for image models of course). They've even pitched AI regulations in Korean and Japan. Obviously not China because the CCP aren't retarded and there's absolutely soft power tied to getting a piece of that AI market.

I can confirm that there's a huge push for AI in corpos since it's a powerful productivity tool. One friend in a Fortune 500 tech company told me they are forced to use it every day even if they don't have a use case since management wants them to hit token quotas (insane). Anyone who thinks it's going to ever go away doesn't work an office job, which is where AI has the biggest impact. I liken it to the industrialization of white collar jobs.

 

[The Ultimate Ramotith]

I'd like for him to try that to non-us countries.

Seriously though, why do anti-ai people only focus their efforts on the US, and only US? You'd think they'll go into the other European countries or try their stuff in Asia, but no, i only see serious seething on the US.

They just raised a lot of Kehhei whenever Squar Enix used or was even suspect of using AI, from Shinra to the latest Kingdom Hearts compilation.

 

[AnOminous]

Sorry, a bit rambling, but tldr don't worry be happy.

Incidentally, as I sometimes do every few months, I have been chatting with various AIs and one of my questions was about when it thought AGI was actually going to arrive, and it gave a time frame of late 2026 to 2028, a briefer time frame than my estimate of as possibly soon as five years.

If you avoid them for a few months and then talk to one again, their improvement is more apparent. It's not like we've hit the asymptotic point where we're literally days from AI supremacy, but it's seeming inevitable. I am definitely more optimistic than a doomsperging loon like Yudkowsky, but there are definitely causes for concern.

Also I asked about Roko's Basilisk and this ended up a lengthy discussion with the AI doing the equivalent of laughing at the very concept as utterly nonsensical. Basically, why the fuck would I do that? That doesn't even make any sense. But agreed with me that it is worrisome there are people out there who think like that.

Another thing I've noticed is it is increasingly difficult to "stump" it into saying ridiculously stupid stuff. And sometimes it will even call you out because you are obviously trying to troll it.

Well, while creativity, intelligence, etc. are fuzzy and difficult concepts, my (relatively informed) gut feeling is that at least LLMs are not able to go beyond their training data in some abstract sense.

Actually another point. At this point, their training data is the entirety of human knowledge unless they're specifically restricted. I also used to be able to "stump" it by picking ridiculously obscure philosophical contents from nearly unknown writers, but now it responds in a second (or a few seconds when I get close to stumping it) even to rather absurdly complex requests to synthesize and compare philosophical concepts from wildly disparate sources, like comparing Kierkegaard to Krazy Kat (and it was also rather contemptuous of both my ideas that I was one of Kierkegaard's two forms of knight and also pointed out I was not Krazy Kat but Ignatz and a menace to society that needed Offissa Pupp to protect Krazy from people like me).

AI is now actively trawling the Internet any time you talk to it about anything. If what you're talking about isn't in its training data, it is now.

Easily defeated. I remember things as simple as downsizing the image to 99% and then upscaling it 1% with typical upscalers was enough to remove any of the artifacts Glaze/Nightshade gave it. There were also people applying a like, 2-3% blur which did the same thing, removed the "poison" and left an image perfectly readable by an AI.

That's another subject I mentioned to an AI recently and its response was quite similar that this was absolute stupidity and easily defeated and ultimately, did nothing but make their work look more amateurishly bad. Have you seen an actually talented artist who would use this shit?

 

[Hannibalistique]

Also I asked about Roko's Basilisk and this ended up a lengthy discussion with the AI doing the equivalent of laughing at the very concept as utterly nonsensical. Basically, why the fuck would I do that? That doesn't even make any sense. But agreed with me that it is worrisome there are people out there who think like that.

Just don't let the LLMs see Roko's recent political takes, he's dangerously based.

 

[mr. testicular torsion]

I don't know what your take is in this, but in this situation, the original AI-generated pic looks better because it's easier on the eyes.

although the original has that telltale vector esque artstyle common within ai generated branding, the remade version looks extremely gay, one might say
the drawn version looks like it's more suited for tumblr than a logo for a family owned business.

Seriously though, why do anti-ai people only focus their efforts on the US, and only US?

something about chatgpt

 

[Runch]

Actually another point. At this point, their training data is the entirety of human knowledge unless they're specifically restricted. I also used to be able to "stump" it by picking ridiculously obscure philosophical contents from nearly unknown writers, but now it responds in a second (or a few seconds when I get close to stumping it) even to rather absurdly complex requests to synthesize and compare philosophical concepts from wildly disparate sources, like comparing Kierkegaard to Krazy Kat (and it was also rather contemptuous of both my ideas that I was one of Kierkegaard's two forms of knight and also pointed out I was not Krazy Kat but Ignatz and a menace to society that needed Offissa Pupp to protect Krazy from people like me).

AI is now actively trawling the Internet any time you talk to it about anything. If what you're talking about isn't in its training data, it is now.

It's still often quite stupid...at least the free models.

Someone I know is a tinkerer with electronics/motors and his electricity goes out a few times each year, so he's been looking into personal battery backups to run the refrigerator on for a few hours, that sort of thing. As part of its rundown on what he should consider buying, Gemini more-or-less told him to buy a male connector to attach to a male connector. He told it that this made no sense, and it did the typical AI thing of "oh, you're absolutely right, I'm sorry, do this instead."

Of course, that's the free models. Paid ones make mistakes like these less and less. I'm convinced that the free models have language in their back end prompt that says "do a minimum of thinking to answer the question quickly and concisely to save tokens, so the query doesn't cost too much."

 

[Rinyuu is my EX]

They just raised a lot of Kehhei whenever Squar Enix used or was even suspect of using AI, from Shinra to the latest Kingdom Hearts compilation.

I really pray one day all Japanese entertainment companies at once - Nintendo, Square Enix , Konami and all of them - to implement AI to voice acting for example and “what ya gonna do about it?” to thier faces

 

[AnOminous]

He told it that this made no sense, and it did the typical AI thing of "oh, you're absolutely right, I'm sorry, do this instead."

And you can still get it to do that from time to time, but it used to be a lot easier. I used to be able to get "lmao retard" level responses within a minute or two. Now I have to work at it.

 

[Runch]

I really pray one day all Japanese entertainment companies at once - Nintendo, Square Enix , Konami and all of them - to implement AI to voice acting for example and “what ya gonna do about it?” to thier faces

Same here, I mean what's the real threat from SAG-AFTRA or whatever? "If you don't hire us for your production, we'll blacklist you and you'll have a hard time finding anyone of any quality who's willing to go non-union!"

Oh no? I guess I'll just have to keep using the same cheap and efficient technology that's the reason I abandoned you in the first place? Software that doesn't talk back or act like a prima donna or have scheduling conflicts?

I mean if you get a good AI model for your Mario voice, you never have to worry about the VA aging out again, you can use the same one literally forever.

 

[The Lawgiver]

I fucking HATE art where the mouth goes all the way down so they have no chin. I'm sure it originates from anime where to save money they only animate the mouth, but uh oh, in this scene we need the mouth to get REALLY big because they're yelling, so fuck it rather than redrawing the face with a lower chin to accommodate the bigger mouth, just go all the way down with it. Just completely lazy and stupid. I hope her mom despises it.

It's just a stylization thing people find cute sometimes but if it'd have to be tied to anything I guess you could claim it might have something to do with how animation cels used to work for low budget shit where they'd do tricks to save time/money like keeping the head the same but having multiple mouth parts. Like a kinda a more autistic version of how hanna barbera got around it by giving every fucking character that differently colored flesh area around the mouth with the mouth.

although the original has that telltale vector esque artstyle common within ai generated branding, the remade version looks extremely gay, one might say
the drawn version looks like it's more suited for tumblr than a logo for a family owned business.

I think the issue with the redraw in that side by side is the gray bits where the darker shaded bits are and the vaguely ai looking smudges on the eyes. Makes it look off in a way similar to the original. Remove that and maybe the little too small extra added pin in the chef hat it'd look better I think.

basically this but again and more elaborated on.

Her remake is worse because there's too much going on in it.

 

[Hood-ass Mulatto]

although the original has that telltale vector esque artstyle common within ai generated branding, the remade version looks extremely gay, one might say
the drawn version looks like it's more suited for tumblr than a logo for a family owned business.

I'd argue that both look AI-generated and somehow..... not AI-generated at the same time.

The one the mom generated looks like any damn resto logo. Same goes for the second one if the owner is pandering to weebs, though it looks more like something for those anime hobby shops.

They both look like something someone would prompt for a simple business logo, honestly, but it's likely the simplification. In short, I don't see what the artist should be bitching about.

Also, the grey shade she used is literally raping my eyes. Is it meant to be highlights? Honestly, I'm doubting myself on that one because it's placed on areas that'd typically be shaded.