AI Art Seething General

[imaginary friend]

AI could make the next mona lisa and knowing the thoughts behind its creation being some server rack in Finland pumping numbers is somehow worse than those "3 stripes on a canvas" paintings going for millions, cause there's a person and a story behind it.

I'm a complete amateur compared to people who sink hundreds of hours into making workflows, but personally I make AI art because I enjoy the process and also like the things I make because that's the reason I made them, for myself. AI art won't get you money or respect. It's worthless to everyone but the person making it. But people say that making art for yourself or for art's sake is the point of art. In that case, AI artists would actually get the point of art more than human artists who are all about comissions and clout that they get from other people who see their art and think "a human spent a lot of time making that".

Eh, it's a very autistic/ argument but the artists' "AI ARTISTS ARE MISSING THE POINT OF ART" is so annoying I feel it kind of deserves a rebuke.

AI porn will definitely be a giant industry if it already isn't one.

 

[Mad Hunter]

AI porn will definitely be a giant industry if it already isn't one.

I agree with this. While there will always be drawn hentai, AI porn for normies who don't give two shits about hentai and the rest of the subculture would be a great idea. Why?

1. Horny people are horny - most likely, they won't care about SUPPORT TEH STARVIN' ARTISTZZZ and just want to get off to something sexy.


2. Less people will be exploited, and of course, like hentai, at the end of the day, it's just pixels on a screen. No real people are being pumped with silicon bbl, getting pregnant through creampies, and going through the psychological pressure of having sex acts taped on a screen that will be spread around the Internet.

Just don't make porn of actual celebrities and only use actors and actresses for the pornographic vids. This strategy can also be used for child actors, considering how child stars endure so much pressure and abuse from the industry and even outside the industry.


Well, of course, the idiots who drink the Kool Aid and only want Internet points will ignore this.






Also, I'd like to add this silly thing I found to this post. Chill out son, it's only a joke.

https://x.com/voremaster4000/status/1883625409688010800?t=7YHAm3oGbEkBFG2p6yEDXA&s=19

I have a friend who uses the c.ai chatbot and yes, while it's for silly romantic roleplay (which I'm not into because I personally find a robot underwhelming), it's not only that. She gets story and art ideas from it. Both of us are very open-minded to this AI thing. Most of the people I have met and spoken to regarding AI were not morons who cry all day about stupid garbage on the Internet that's basically AI BAD AI BAD OH NOES.

Hell, I've even seen people on the webtoon subreddit recommend AI art for backgrounds when I saw someone complain he cannot draw backgrounds. Guess what? Never did I see a hateful comment on that suggestion.

 

[YautjaWeeb]

I'm a complete amateur compared to people who sink hundreds of hours into making workflows, but personally I make AI art because I enjoy the process and also like the things I make because that's the reason I made them, for myself. AI art won't get you money or respect. It's worthless to everyone but the person making it. But people say that making art for yourself or for art's sake is the point of art. In that case, AI artists would actually get the point of art more than human artists who are all about comissions and clout that they get from other people who see their art and think "a human spent a lot of time making that".

Eh, it's a very autistic/ argument but the artists' "AI ARTISTS ARE MISSING THE POINT OF ART" is so annoying I feel it kind of deserves a rebuke.

AI porn will definitely be a giant industry if it already isn't one.

It seems like AI's major draw is armchair dudes making stuff only relevant to themselves. Its basically instant "demoitis" from the music side of human creativity. AI is also good at making something that is a mockery, offensive, straight up coomer or coomeresque etc.

But I find its actually perfect for illustrations. In that space it doesn't have to be hard-hitting and profound like it would as "pure" artform seethe all it has to do is illustrate an idea you are trying to convey in fiction/non-fiction writing.

Or maybe you get lucky and hit some 1 in a 100 million stunning artpiece mashed out from Stable Diffusion's brain. It's possible, just a matter of how many times you need to hit generate over the course of your lifetime.

 

[SaltyFanta]

I watched this guy maybe once or twice a while back, and now I non-stop get him on my feed so I decided to watch an actual video of his, and this is the perfect example of the anti-ai soy rage. The funny thing is that he's a png slop tuber. People like this concern me more than ai.

He wants the government to be able to control the content companies like YouTube can host, and censor AI content. Synonymously, he defines a video of an ai generated white lion as stealing from Disney. I think that mindset more dangerous to artists than AI, imagine a day in which using Anti-Ai regulation Disney can declare your work copyright infringement on vague resemblence alone.

 

[AmpleApricots]

This is gonna happen at least in the US anyways. OpenAI will run the show, whetever you like it or not. They're currently demonstrating how far they're willing to go when a competitor dares to be the better deal. Deepseek can be glad they're in China.

 

[harmy.kook]

Deepseek can be glad they're in China.

Anyone dumb enough to use Deepseek might as well use Temu or TikTok as well.

 

[AmpleApricots]

Anyone dumb enough to use Deepseek might as well use Temu or TikTok as well.

Deepseeks models are open weight. You could literally run them at home on an airgapped system without ever sending a single word to deepseek (or anyone else, for that matter) if you wanted. You cannot and will never be able to do that with OpenAIs models. You'll always be beholden to send your data to their cloud computers with all implications this has. You can also do nothing but hope they will not change their side of their deal if e.g. your companies' infrastructure depends on them. They can unperson you like every other of these shitty cloud platforms. This can by definition not happen with something you can just run in-house, on your own hardware. But yeah, I'm sure OpenAI will not abuse that kind of power they can wield over their customers, because they're 'murican. Be careful not to choke on that corporate boot you're deepthroating.

 

[harmy.kook]

Be careful not to choke on that corporate boot you're deepthroating.

Who said I use OpenAI?

Plus this ignoring the fact China is openly hostile to the West along with the fact the CCP has openly been stealing people's shit through state-backed hacker groups.

All cemented by the fact they have access through the CCP's laws to the data Chinese tech companies have.

But maybe you're just deepthroating the Chinese government because it somewhat inconveniences the Tech Industry.


Like Redpillers™ embracing Islam because they're misogynistic

 

[AmpleApricots]

Who said I use OpenAI?

Plus this ignoring the fact China is openly hostile to the West along with the fact the CCP has openly been stealing people's shit through state-backed hacker groups.

All cemented by the fact they have access through the CCP's laws to the data Chinese tech companies have.

But maybe you're just deepthroating the Chinese government because it somewhat inconveniences the Tech Industry.


Like Redpillers™ embracing Islam because they're misogynistic

What are you even on about, what data? You download the model and run it on your computer. You never need to be in direct contact with deepseek, ever. They don't even have a way of knowing you use it. Chinas or even Deepseek's stance towards the west does not matter, at all. They don't get any money from you for this.

The american tech industry is not a fan of deepseek because of that fact. They prefer proprietary, pay-to-play, cloud-based solutions which puts them in direct control over your data and the way how you use their AI. Every single request. They want you to be dependent on them.

OpenAI/Anthropic/Microsoft (via Azure) has done all kinds of thought-policing on their users pretty much as long as this tech exists. OpenAI banned accounts for asking o1 what it's system prompt is or for just talking about the wrong topics even though nobody was there to even witness it, completely cutting you off from their AI because you have no route to using it anymore (vs. just running the model locally and not giving a shit about what the company wants you to do with it). There was a proven instance back in GPT3 days where OpenAI gave private logs of API interaction to a 3rd party for manual classification and moderation. Random fiverr workers basically read through personal logs to judge them on wrongthink. Absolutely no oversight or respect for privacy.

Get used to OpenAI if you're american, since they're pretty much cementing their position as the one true AI provider with your current administration. But yeah, I'm pretty sure they won't abuse this, at all.

You have absolutely no idea what you're talking about.

 

[sex haver]

Who said I use OpenAI?

Plus this ignoring the fact China is openly hostile to the West along with the fact the CCP has openly been stealing people's shit through state-backed hacker groups.

All cemented by the fact they have access through the CCP's laws to the data Chinese tech companies have.

But maybe you're just deepthroating the Chinese government because it somewhat inconveniences the Tech Industry.


Like Redpillers™ embracing Islam because they're misogynistic

My man, it is incredibly clear you have zero fucking idea what you are talking about and should really stop while you're ahead. Faggot.

 

[Willie B. Hardigan]

People too fucking lazy to make their own AI models are now unironically sucking the teats of whatever large governmental/corporate entity will hand-feed them the quickest. It's like arguing whether Walmart or Target has less chemically poisoned food instead of doing the smarter thing of growing your own. At least one can make the excuse that doing the latter requires land and big tools, the only thing stopping a person from doing the former is a couple of tutorials and a mid-range PC. Self-sufficiency is dead I guess.

 

[linchan]

Looking at this thread Im convinced Deepseek is a psyop to make the Western world demonize open source.

 

[Runch]

Looking at this thread Im convinced Deepseek is a psyop to make the Western world demonize open source.

It's working.

https://www.hawley.senate.gov/hawley-introduces-legislation-to-decouple-american-ai-development-from-communist-china/

Hawley Introduces Legislation to Decouple American AI Development from Communist China​

Wednesday, January 29, 2025
Today, U.S. Senator Josh Hawley (R-Mo.) introduced legislation to protect America’s artificial intelligence (AI) development from China.
“Every dollar and gig of data that flows into Chinese AI are dollars and data that will ultimately be used against the United States,” said Senator Hawley. “America cannot afford to empower our greatest adversary at the expense of our own strength. Ensuring American economic superiority means cutting China off from American ingenuity and halting the subsidization of CCP innovation.”
Senator Hawley’s new legislation comes after China unveiled “DeepSeek,” a data-harvesting, low-cost AI model that sparked international concern and sent American technology stocks plummeting.
Senator Hawley’s Decoupling America’s Artificial Intelligence Capabilities from China Act would:

• Prohibit the import from or export to China of artificial intelligence technology;
  
  
  
• Prohibit American companies from conducting AI research in China or in cooperation with Chinese companies; and
  
  
  
• Prohibit U.S. companies from investing money in Chinese AI development.

Full text of the bill is available here.

https://www.hawley.senate.gov/wp-content/uploads/2025/01/Hawley-Decoupling-Americas-Artificial-Intelligence-Capabilities-from-China-Act.pdf

 

[harmy.kook]

People too fucking lazy to make their own AI models are now unironically sucking the teats of whatever large governmental/corporate entity will hand-feed them the quickest.

I've actually looking into talking to a friend of mine's who's deeper into the tech than I am who's built his own model.

But back on topic, Deepseek is some of the most artificially( ) pushed shit I've ever seen.

Overnight there are several 10 to 20 minute videos popping up in my feed glazing and worshiping the tech that's openly pushed by China.

 

[Slurred]

But back on topic, Deepseek is some of the most artificially( ) pushed shit I've ever seen.

Overnight there are several 10 to 20 minute videos popping up in my feed glazing and worshiping the tech that's openly pushed by China.

People are legitimately excited that there is now an open-source, MIT-licensed reasoning model with performance on-par with the closed-source GPT o1. There are questions around whether it actually was as cheap to develop as is claimed, and whether it used GPT o1 outputs for training, but regardless of either of those things its release represents a massive step forward for open-source AI. That is objectively a big deal in the space and worth making videos about. Not everything is an astroturfed psyop.

As has been mentioned already, you can download this model and run it locally, so it being made in China is irrelevant. The hardware requirements are quite steep, but there are already multiple smaller distillations out there that can run on fairly standard gaming PCs and perform quite impressively. You can even run it on an airgapped setup without an internet connection if you're paranoid about Chinese state hackers having somehow backdoored the model files.

Deepseek wouldn't matter as much if OpenAI had lived up to its name and original purpose by actually being open and releasing its model weights, but they decided not to.

 

[Whatevermancer]

People too fucking lazy to make their own AI models are now unironically sucking the teats of whatever large governmental/corporate entity will hand-feed them the quickest. It's like arguing whether Walmart or Target has less chemically poisoned food instead of doing the smarter thing of growing your own. At least one can make the excuse that doing the latter requires land and big tools, the only thing stopping a person from doing the former is a couple of tutorials and a mid-range PC. Self-sufficiency is dead I guess.

What kind of hardware and training data do you need to teach a model comparable to ChatGPT or DeepSeek?

 

[Runch]

You can't really make your own model independently without millions of dollars in hardware. You can fine tune existing models without exorbitant costs, though.

A lot of people make and publish "models" on Civitai which are nothing but glorified LoRAs of a certain concept baked in very strongly, or else combinations of multiple other models at different strengths.

 

[AmpleApricots]

But back on topic, Deepseek is some of the most artificially( ) pushed shit I've ever seen.

Once again, you have no idea what you're talking about.

Deepseek R1 is not even only a very capable model, synthetic data created with it improved Deepseek V3 and even several models (llama, qwen) of competitors. Even though these models aren't as impressive as R1, the sheer degree of improvement through the synthetic datasets deepseek (the company) created are very intersting.

Then on top, deepseeks in-house models are very efficent and have many novel approaches in optimization, for example training in 8 bit, to reduce hardware requirements, or direct us of PTX, ~10x improvement, just like that, from Deepseek's V3 technical report:


While you can argue that OpenAIs top models are somewhat better (which I think is most likely subjective) they absolutely do not own models that are this efficient. People know this and talk about this because contrary to these american companies, deepseek open-sourced a ton of their stuff, enough to independently reproduce it, which is the gold standard in science. Everyone talks about it because not only is the model one of the best performing LLMs in existence currently, it also has tons of novel approaches in architecture that make it a lot more efficent to run on hardware. It is unique.

The best way currently to learn about LLMs in-depth is to read chinese papers because american corporations release nothing. This is a fact. This fosters an atmosphere of collaboration in the scientific community that helps progress overall, something OpenAI et al are not interested in if they can't directly profit from it. This is also fact. They most likely also don't want to give others a leg up because they are aware how tenuous their position at the top is.

Maybe you should just stop talking before you embarass yourself more.

 

[crowabunga]

Deepseek wouldn't matter as much if OpenAI had lived up to its name and original purpose by actually being open and releasing its model weights, but they decided not to.

Thanks to the Rationalist cult making AI scares appear "reasonable" to pseuds through a lot of undergrad-level philosophy masquerading as legitimate prognostication, Altman had it easy convincing folks "AI" was too "dangerous" and seguing into opaqueness, then regulatory capture, then for-profit rent-seeking.

Chinese or not, I'm glad someone knocked the legs out from under that little rat faggot, and I also hope Altman gets Allahu Ackbar'd by Schlomo Yudkowsky: the world would be better off without either of them.

 

[Runch]

People know this and talk about this because contrary to these american companies, deepseek open-sourced a ton of their stuff, enough to independently reproduce it, which is the gold standard in science.

Though when you're reproducing it, be sure to skip step 18,302, which is the one where you implement the block on all data regarding Tiananmen Square.