Stable Diffusion, NovelAI, Machine Learning Art - AI art generation discussion and image dump

[King Dead]

It's the same idea as if someone gifts me a painting they painted themselves, you bet your ass I'm going to value that painting orders of magnitude more than if they had sent me a photograph of the same thing.

A lot of the screeching isn't coming from very successful artists, especially those who work in traditional media. Rather, most of it's coming from the lower-skilled DeviantArt randos that have been churning out basic commissions and charging more than they're worth. Now that a computer can produce something as good as they can or better in considerably less time, some people that would normally have had to pay for a commission will instead figure that it's good enough to let the AI do it for them. They're mad that their jerb is being taken over by robots when they were told that it was those damned obsolete farmers and factory workers that were supposed to be out of work.

It doesn't have to be this way, of course. For one, they can use this as impetus to improve their skills and produce better art that's actually worth the money they want to charge for it. You can't coast by on "good enough" anymore when that's the realm of AI art, so you need to show you've got what it takes to make something people will pay for. For another, they could transition to physical media and create art that no robot can (at least for a while). They can sell prints as well and make additional money off of every piece they complete. Or they can find a niche that it's harder to get an AI to do properly as yet, especially when it comes to porn artists. Fetish art always commands a higher price premium if you're willing to draw it, especially for those kinks that don't have a lot of artists tackling it.

They can also view AI as a tool instead of a threat. I have a friend that works primarily in charcoal and ink, and he's been using AI to generate picture ideas that he then improves upon and translates to physical media. The bot gets it most of the way, and all he has to do is clean it up and fix any problem areas as he draws it out. He's loving how well it works once he gets his prompts right, and he doesn't view it as a threat to his creative efforts at all.

I'll always put a higher value on art that human hands have created over AI art, doubly so for physical works. But I'm also loving how quickly I can take an idea from my head and get multiple well-generated version of it in only a few minutes, as opposed to waiting for days or weeks to get through an artist's commission queue. Both have their place, and both are valuable in their own ways.

 

[Pendulous Dong]

Happy 100 pages, guys.

 

[BirdUp]

for "disinformation researchers," openai suggests loisences for consumers to access remote AIs and to buy AI-capable GPUs (source)

 

[Sulla]

View attachment 4516985
for "disinformation researchers," openai suggests loisences for consumers to access remote AIs and to buy AI-capable GPUs (source)

https://arxiv.org/pdf/2301.04246.pdf

It does more than that, they go as far as to suggest limiting the amount of computational power people should be allowed to buy.

The silver lining in this is, Its proof that openai and other private machine learning companies are genuinely afraid of opensource competition.

 

[The Mass Shooter Ron Soye]

It does more than that, they go as far as to suggest limiting the amount of computational power people should be allowed to buy.

Governments Impose Access Controls on AI Hardware

I've been anticipating such a scummy move for a while. I'm not sure how successful they will be in the long term. In the short term, nobody is going to train their own trillion+ parameter GPT-4 at home. But what takes 100,000 datacenter GPUs today could shrink to a brain-like box in 10-20 years. Maybe you'll have to import it from China.

 

[WelfareNiggerQueen]

https://arxiv.org/pdf/2301.04246.pdf

It does more than that, they go as far as to suggest limiting the amount of computational power people should be allowed to buy.

These fag's really want to go to war with gamers don't they.

 

[vladimirovanova]

Obviously.
They're the ones behind all these AI art spergouts.

Notice how all the anti-AI art idiots are ganging up on Stable Diffusion (the open source one that the average joe can use without restriction) and Midjourney (not as open-source but still readily available and lightly censored) and NONE of them are targeting OpenAI or Google?

 

[AnOminous]

Notice how all the anti-AI art idiots are ganging up on Stable Diffusion (the open source one that the average joe can use without restriction) and Midjourney (not as open-source but still readily available and lightly censored) and NONE of them are targeting OpenAI or Google?

They know Google will stomp them flat like the tiny little bugs they are. So they're picking on the relatively little guys. Tough shit for them because Google isn't stupid enough to allow a precedent against a little guy that would harsh their own buzz.

 

[JoseRaulChupacabra]

They can also view AI as a tool instead of a threat. I have a friend that works primarily in charcoal and ink, and he's been using AI to generate picture ideas that he then improves upon and translates to physical media. The bot gets it most of the way, and all he has to do is clean it up and fix any problem areas as he draws it out. He's loving how well it works once he gets his prompts right, and he doesn't view it as a threat to his creative efforts at all.

This.

Another is the potential to photobash the shit out of the ai images. Generate terrain, followed by boobacharacters, and play with it in Gimp/PS. Should lead to some pretty interesting results.

That legit looks like a land card art in MTG.

Luminescent Lava Tube
Land
Luminescent Lava Tube comes into play tapped.
T: Add R or U or W to your mana pool.
Under the light of the moon called Yosh, the lava flows and dilates cavernous structure.

 

[Akashic Retard]

So which one can I use to make porn of people?

 

[ducktales4gameboy]

A thing I've learned messing with various loras and checkpoints is that if you like the aesthetic but not the subject matter you can usually just prompt hard against it and get interesting results.

There's a vintage bondage comics model up on civitai that's stark black and white high contrast with that does an amazing job with abstract landscapes it you ask for them and put something along the lines of 'woman, person, people' in the negative.

So which one can I use to make porn of people?

Just go to civitai.com and fish some boobs out of the endless stream of T&A.

 

[Allanon]

View attachment 4516985
for "disinformation researchers," openai suggests loisences for consumers to access remote AIs and to buy AI-capable GPUs (source)

I think, personally, that we should do nothing to stem the tide of deepfake and ai. Let the flood of false reality wash over us so that people stop believing literally everything they see. If every image, video, and piece of text that they see may be entirely fake, then people will be forced to do their own research every time to see if indeed it's actually real or if it's fake. If it's common enough, people might actually learn to do it themselves instead of rely on waiting for someone else to tell them, so it will also kill the cancer of "fact checkers" that don't really check facts at all.

Unfortunately I assume this is the real reason why people like OpenAi are interested in censorship and control of all that. If only a select group of people have access to being able to falsify 'reality' with AI, then their control over media, public perception, and the public itself will be unprecedented. I'm sure at least one government in the world has a group of tech nerds seething that this tech ended up publicly released and acknowledged at all, considering the sheer power this would give someone who had it if no one else knew it was possible.

 

[Sulla]

I think, personally, that we should do nothing to stem the tide of deepfake and ai. Let the flood of false reality wash over us so that people stop believing literally everything they see. If every image, video, and piece of text that they see may be entirely fake, then people will be forced to do their own research every time to see if indeed it's actually real or if it's fake. If it's common enough, people might actually learn to do it themselves instead of rely on waiting for someone else to tell them, so it will also kill the cancer of "fact checkers" that don't really check facts at all.

Unfortunately I assume this is the real reason why people like OpenAi are interested in censorship and control of all that. If only a select group of people have access to being able to falsify 'reality' with AI, then their control over media, public perception, and the public itself will be unprecedented. I'm sure at least one government in the world has a group of tech nerds seething that this tech ended up publicly released and acknowledged at all, considering the sheer power this would give someone who had it if no one else knew it was possible.

I agree with you.

But, it isn't just about control. OpenAi is just extremely butthurt over the fact that their technology can or will be largely replicated by anyone in the future.
Let me explain, ML requires significant computational power. So the barrier to entry is relatively high for the moment, however what most people don't realize is there is a ceiling. EG: running an algorithm for a trillion effective years isn't going to turn up a model much different than an algorithm that ran for a billion effective years.

Eventually the technology will progress to the point where it doesn't matter if private companies and governments have computers exponentially more powerful than basic bitch consumers.
That said, I'm fairly confident OpenAi won't be effective in their attempts to legislate computers, and even if they were successful. People will just smuggle hardware in from whatever unregulated country, and there is no fucking way nvidia or amd suddenly stops R&Ding new consumer hardware.

 

[JoseRaulChupacabra]

If every image, video, and piece of text that they see may be entirely fake,

As if that isn't already the case.

then people will be forced to do their own research every time to see if indeed it's actually real or if it's fake.

NOOOOOOOOOOOO. You can't do your own research. You have to trust The Experts™. Only they can interpret The Science™.

I do think that people will eventually learn to spot fakes as part of the adaptation process. They will have to. The real pressing question is what sort of damage can be done in the mean time.

 

[AnOminous]

As if that isn't already the case.

NOOOOOOOOOOOO. You can't do your own research. You have to trust The Experts™. Only they can interpret The Science™.

I do think that people will eventually learn to spot fakes as part of the adaptation process. They will have to. The real pressing question is what sort of damage can be done in the mean time.

We can all be killed as Consensus™ determines that our opinions are too vile for us to be allowed to exist. Scan your opinion code here, chud.

 

[IsaacShraeder]

Wow, Donald Trump wants to gift ME a gold bar? Sign me up, patriots. [A, L.]

 

[Stephanie Bustcakes]

Goth Zelda

 

[Epoch]

Goth Zelda
View attachment 4553854View attachment 4553857

If you don't mind me asking, what model? I've been out of the AI generating 'biz' for a little while so I'm behind on all the new improvements

 

[Stephanie Bustcakes]

If you don't mind me asking, what model? I've been out of the AI generating 'biz' for a little while so I'm behind on all the new improvements

It's a weighted sum checkpoint merger between two models. 67% Dreamlike Photoreal (a realism model) and 33% Kotos Abyss Proto (a cartoon/anime/hand drawn model). Both models are specifically for making characters and as such are good at anatomy.

 

[Epoch]

It's a weighted sum checkpoint merger between two models. 67% Dreamlike Photoreal (a realism model) and 33% Kotos Abyss Proto (a cartoon/anime/hand drawn model). Both models are specifically for making characters and as such are good at anatomy.

Thanks, but I'm having trouble getting faces to not be a horrible nightmare. Apologies if this sort of stuff has been explained earlier in the thread, but any tips for non-nightmare faces?