Like all automation, it would be cool if it didn't mean workers were staring down the barrel of capitol's gun. I love the cool art machine learning models can generate. Would be even cooler if we lived in a world where this didn't threaten anybody and couldn't be used as a cudjol.
We can make a world where this is the case. Suck my guts (girl nuts) for more objectively correct takes you slop hogs.
luddites didn't destroy machinery because machines bad, they did it because they threatened to make their craftswork which they spent their lives developing economically obsolete
Reposting my objectively correct and Marxism pilled comment from an earlier thread.
Digital art did not replace traditional art, and AI generation will not replace artists. AI generation is just a tool, like GIMP or a camera or a synthetic brush, and mastering the use of that tool is every bit as much of a skill as mastering the use of the tools that came before. I suspect that in the future mainstream art will be AI-assisted but artists will still be necessary to “finish” concepts, thus speeding up their output but not replacing them.
As with other industries, a massive increase in productivity will not produce an increase in wages - it will instead force artists to adopt AI assistance in order to keep up. Commissioning mid art will become dirt cheap, and artists who expect to make a living doing it will have to either master AI generation and increase their volume or specifically target the high end market like artisans of every other industry do. Also as with those other industries, the number of artisans will drop and the number of people doing that kind of work in an assembly line for a wage will increase.
If one doesn't already exist, soon there will be an app work grift paying artists pennies to generate cheap art to order.
I suspect that in the future mainstream art will be AI-assisted but artists will still be necessary to “finish” concepts, thus speeding up their output but not replacing them.
Yes. Specifically it'll help company illustrators to make their job way less annoying when you hit a creative block.
I'm already using it to bounce ideas off of while copywriting when I hit a wall writing the same SEO bullshit again and again.
The work it comes out with sucks shit, but it gives you the basic structure and speeds up the process and makes it all less teeth pullingish
I'm thinking a tier higher than that lmao, something like that ghost writing app.
It's not the same as other automations, though. That's where all these arguments towards AI art break down. AI art explicitly requires exploitation of mass amounts of artists to even WORK. These AI generators are getting better BECAUSE MORE ART IS BEING SHOVED INTO THEM WITHOUT PERMISSION.
Automation in art exists in digital tools already WITHOUT EXPLOITATION. Blending tools? Blurs? Smoothing? Color correction?
That's the automation. It does not require exploiting existing artists to work.
buddy I've got news for you about how every other process of industrialisation has happened under capital. Where do you think the cotton in the loom came from?
An automation being used to exploit people != an automation requiring exploitation to even work
plenty of public domain art out there! would make slightly weaker, less profitable, models but is still entirely viable as training data
Pretty sure a company could pay a license fee to access stock photos or any photos they wanted, to use alongside public domain photos.
All of these AI generators have a click wrap license forbidding commercial use. Any commercial version of this software will be using licensed works and the public domain..
I'm not sure where I fall on this. Couldn't you make the same argument about an artist who looks at thousands of other artists' works in the process of learning to draw, developing a style, etc.? And the same with musicians- there are such extreme similarities between different pieces of music, especially within a genre. Are the later musicians exploiting the earlier ones?
I'm somewhat being devil's advocate there. It's a huge, very techbro-y, leap to assume that a human artist's brain is the right analogy for what these algorithms are doing. But on the other hand it also doesn't seem that obvious to me that it's not. Generally we accept that people can look at two pieces of art and decide whether one is flat-out copying the other or just inspired by it. Maybe that same test is right no matter whether the second piece was created by a human or machine?
I agree that humans take inspiration from each other all the time in regards to art and I think that is a great and healthy thing.
The primary difference (and the difference between this being art/notart imo) is that humans bring their own experiences into the art. These machines cannot do that. The only thing AI art does is pull in art work, add it to the pile of referenced things, and create pictures from that.
We all agree that when one human directly copies another's artwork without adding anything to it, they are a hack. These machines explicitely cannot add anything to the art than what is input. Humans all have unique experiences and beliefs to draw from. AI art does not have it.
That's how I see the whole situation, at least.
Except people also produce vapid and derivative art that's strictly aesthetic or utilitarian in its purpose. The art-generating AI even works in a similar fundamental way to how humans do, being a neural network that recombines observed patterns to synthesize something new-but-derivative.
Even the limits of its training data are more a matter of scale than of its nature: when it gets to a point where the neural network is twice as big and being fed movies along with their screenplays and mountains of novels in a way that start to create a further layer of context and understanding, then what differentiates the AI from a human raised on that same media? Is there an actual innate spark to humans or is it just that we're currently unique in our scale and sophistication? When the output of an AI is being guided, curated, and edited by a human how is that distinct from the creation of media now, apart from removing the need for specific technical skills?
would be dope if things weren't commodified to the point where expressing yourself results in getting paid. then maybe we wouldn't need to shortcut the act of human expression and we could just appreciate it for the act that it is. art is both fetishized and monetized in ways that pervert it from being what i would even call 'art.'
I have real reservations about it because once people start using it for animation (video synthesis/generation? idk) in HD it's going to be more energy-intensive, and having that be widely available really kinda sounds bad. I mean it may not be the scale of crypto in terms of power consumption, but burning the earth to create a 4k "Judy Hopps Stuck in Dryer, What is Stepfox Doing?" clip doesn't sound great either.
It's not going to start using 0.2% of global energy usage like Ethereum did.
True, it's not gonna be that bad, but the ease of use means that people are going to make a lot of it. At the very least I think there's potential for the energy cost of AI-generated art (factoring in discarded/unused generated video/images) to be pretty high.
wonder how it does vs raytraced 3d renders. I know OptiX is a neural net demonising thing and it way reduces the time + energy my renders take
edit: denoising, not demonising lmao
This is an interesting point I don't think I've seen crop up in any of the other threads, similar to some big criticisms I have of IoT.
I feel like I would expect people here to get alarm bells from hearing the latest tech fad combined with a buzzword like "democratizing".
It produces treats and does it cheap, so the negative reaction overall is not as much as you might expect. Some even twist Marx's writings into a pretzel to make it sound like a leftist imperative to deny future income to artists because they're aristocrats or something.
what's the best AI text-to-art website to use ATM? optimize for cartoon outputs if that's ann option.
Probably depends on how technically minded you are and how big your GPU is. Waifu Diffusion might be what you are looking for if you can run it.
I think Diffuse the Rest (IMG2IMG + prompt) is the best and theres a bunch of google collabs out there
