AI is trained by analyzing artists’ work and then instructed to replicate art in a particular style, therefore, from the beginning of the process it wouldn’t be original.
If an AI could create art without being fed galleries of images first and develop its own style that might be considered original.
What do you think human artists do, exactly? You think they just learn to create art in a vacuum? It just magically appears?
Humans can, in fact, create art without having seen others do it first. (e.g.: cave paintings from several millennia ago)
I don’t understand why anyone would assume humans only have the same creative capabilities as a computer when we have free will and all that good stuff that comes with being a conscious, intelligent living being.
Computers can create the equivalent of cave drawings without models as well.
So any artist that went to art school isn’t an original artist?
Yeah, but one went to school to learn how to hone their skills and learn from the masters, the other stole it off of artists who will never see a dime off of it.
Don’t get me wrong, I have fun with AI art, but the moral question that hasn’t been solved yet
Why is it actually different?
If I study a painting (train a model) and then replicate the style am I stealing the painting off of an artist?
If I illegally obtained a copy of the painting that i studied, would the piece that I generated belong to the artist of the painting i studied?
If I go to a wine and design thing and paint a picture after being instructed how and following specifically with a template, does that make my painting no longer mine?
Is a person sitting in a free museum sketching in their notebook, a version of the painting that they see on the wall stealing?
Ai is not copying, the work that it generates is novel. The training data may have been obtained illegally (debatable and not settled in law) but that doesn’t make the generated work any less new or novel.
In your own example, the people who 'went to school and learned from the masters" also don’t pay the original artists. Art students aren’t paying the Gogh estate for permission to study his paintings and they aren’t paying royalties for making something that looks and feels like his paintings either.
All of those the artist knew what they were doing and how their art could be used to inspire new people.
Artists has no way of consenting to thos before it was done. Their art wasn’t taken and used as inspiration for one person, it was taken and is now being mass produced for the masses in some cases.
You’re not sitting in a lecture absorbing what a professor is telling you and filling out an essay question. Your copying someone else’s homework and changing it a little to come off as okay. In private and for private use I’m okay with that, but these big studios and content creators have no right to do that to artists. There’s no way they could have consented to that.
How you don’t see that in principle AI data training and human learning is the same process, is fascinating to me.
People like that surely do see it, they just deny it publicly because they feel threatened by the technology.
No person with even a basic education can legitimately come to another conclusion and be honest. The only way I can see this happening legitimately is to not understand even the basics of how AI art works. Like, not even the first thing about it.
Depends on how it’s synthesized. Some programs, like Midjourney, allow you to use to your own art as material to synthesize new art.
Aside from that, no. It’s not OC.
It’s an interesting thing to ponder and my opinion is that like many other things in life something being ‘OC’ is a spectrum rather than a binary thing.
If I apply a B&W filter on an image is that OC? Obviously not
But what if I make an artwork that’s formed by hundreds of smaller artworks, like this example? This definitely deserves the OC tag
AI art is also somewhere in that spectrum and even then it changes depending on how AI was used to make the art. Each person has a different line on the spectrum where things transition from non OC to OC, so the answer to this would be different for everyone.
If you take the art and just trace and polish it and nobody is any the wiser, in that situation yes. At least until that is found out, in which I will refer to it as derivative work over original content. It’s why I am calling some of the digital art I am working on AI derivative rather than full-on original content.
If all you do is generate an image, do no edits whatsoever, and then act like you did it, then I couldn’t in good faith considering “OC” since you did nothing but type a few words and maybe click a few buttons or moved a slider 3 pixels to the left.
AI art is not OC. It cannot be.
How so? What is it that makes art OC that cannot be applied to AI created art? I think it would take an extremely narrow definition which would also exclude a significant amount of human created art.
Why would human art be then?
Because humans can have new ideas.
Those ‘new’ ideas can be inputted as a prompt into an AI image generator. Would the output of that satisfy your criteria for OC?
No. Every parameter in the LLM, not just the prompt, is or was a new human idea at some point.
And would you say that an idea formed from the combination of multiple old human ideas is not original? If the influence of an existing idea disqualifies it from being original then very little could be considered original. If something additional to existing ideas is needed for originality then that what is that thing which is beyond the capability of an AI?
Personally, I would argue that any new combination of existing ideas is inherently original (i.e. a fresh perspective.)
Talking specifically about image generators (rather than LLMs) which are trained on billions of images - some of which would be widely considered as artwork (old ideas?) and others documentary photographs.
How new is a new idea, if it must be formulated in an existing language?
Language is itself invented by humans.
Not by me though, or you. Can’t have new ideas 😞
“Humans can have new ideas” isn’t “humans only have new ideas”.
Those are not new ideas. Those are based on persons experiences up to that point. There is nothing magical in human brain that we cannot eventually implement in AI.
No. Large Language Models only regurgitate what they’ve been fed.
As do most humans most of the time.
As someone who has been trying to get my vision for a piece to fruition using AI for months…I absolutely think AI is OC. The argument that it references existing work cracks me up because all of art history is derivatives of what has come before. I do think there is “low effort” pieces, but you get that in other mediums as well such as photography. Also…need I mention Duchamp and the urinal?
I do not.
I’m sure there’s plenty of people who just want to play around with art generators to see what wacky stuff they can get and that’s fine. But anyone who bends over backwards trying to convince others that AI generated images are genuine art are ultimately just resentful of the fact that there are people who can create things that they can’t.
Absolutely not
It’s new, but not original. With the recent influx of AI content that doesn’t seem to be slowing down, I’d say we should make a new designation of GC - generated content.
What people make is not original as well, you’re always inspired by something.
Inspiration isn’t the same. It’s more like if I found a bunch of pictures I liked, then traced my favorite parts from each one onto a single piece of paper to make one image made up of lots of small copied pieces of other people’s work.
That’s not how AI works. For example, just a while ago I was generating one image for fun (which I don’t claim is art, by the way), it’s this:
The prompt was quite simple, “anthropomorphic bean standing in a field of beans”.
This is not created from a bunch of pictures, this is created from the AI understanding what a bean is, what anthropomorphic means, what a field is and so on. Try to find me any one image this is created from if you claim it’s just slapping together parts of images. This is an original image (which presumably was never done before, at least I don’t think anyone would create something like that very often), I can’t find any that looks enough like the one I created to claim it was copied from that. I looked for visually similar images using Google, Bing and Yandex.
That leads me to believe, that it’s indeed the same process as a human would do - take an inspiration (from real world or different paintings) and create something new.
Steam bans games that contain such AI content because they are not near OC. Except you train the AI on only your own Copyrighted Images, which mid journey and various other AI aren’t. They are all trained on copyrighted images without asking.
They accept it when you trained on data you had the right to [train and republish on]. That isn’t limited to only your own content.
Yes.
But most of it is, sadly, low effort. Or at least that’s what I’d call it if a human did it.
In general, yes.
No
I think so, there’s still a lot of creativity that goes into designing the prompts from what I’ve seen. AI is just another tool for artists to use and I think it could honestly be considered it’s own medium, like oil painting or wood burning. But I do also understand the hesitation people feel around AI art and calling it OC.
