Purely AI-generated art can't get copyright protection, says Copyright Office | Hacker News Reader

Purely AI-generated art can't get copyright protection, says Copyright Office

theverge.com

41 pointsSerCe17 comments

> No matter how many times a prompt is revised and resubmitted, the final output reflects the user’s acceptance of the AI system’s interpretation, rather than authorship of the expression it contains.

If the photography was invented today would it be judged to be creative and containing authorship?

The art of photography is mostly the art of selection of outputs.

How's Henri Cartier-Bresson different from prompt engineer when he took hundreds of photos of the same street to ultimately pick his great photos:

https://external-content.duckduckgo.com/iu/?u=https%3A%2F%2F...

he is still choosing what to take a picture of, rather than being presented 10 pictures and going "that one". In more practical terms, he had to expend the effort to actually be there in person and do the work. I know AI people pretend that coming up with good prompts is somehow a ton of work, but does anyone actually believe that?

Prompt engineer still chooses to type in prompt describing something specific and presses enter.

And after photographer generates 10 photos he obviously goes "that one".

Photographer makes no effort contributing to determining the color of each of millions of pixels. He just turns few knobs, points the thing and presses a button. Many, many times. Then he picks. The bulk of the art is in picking afterwards. That's when you see the product, evaluate its artistic value and claim it as your art. Another part of art is having interesting idea up front.

I don't really know about AI art but I know how photography is done. The kind that is displayed later in galleries.

And all the objections hurled against AI art would hit photography in the face if it stood in the way.

I don't really agree. I can in the span of 10 minutes get you 100 ai pictures of a mountain without lifting a foot. But if I want to give you 100 photographs of a mountain, i have to actually go to one. And obviously "the photographer doesn't decided the color of the pixels" is only somewhat true at best. You can actually use your legs to manipulate perspective to have the rooftop of a church be centered on the sunset (or whatever). If you want to do it with AI, you can put it in the prompt and hope it actually does it.

But to be honest, I don't really care about this distinction. I am a utilitarian, to me it's not about whether the processes are comparable, but whether it takes any effort. Copyrights, like patents and other IP mechanisms are ultimately there to make sure the effort you put into intangible goods can be rewarded. Everything else is just intellectualizing it. No effort = no reward, that's my ultimate stance.

On the other hand you can take 100 photographs of whatever without knowing any words in any language at all and without sitting in front of any computer. I don't think what's art and what's creativity shpuld be decided on the basis of doing some specific "leg work".

No effort, no reward is a good rule but I apply it to copies as well as originals.

I think "what's art" is another discussion, this is just about ownership. And of course I can also take a camera and take 100 pictures of my ceiling and I'll have copyright without effort, but that's just the price we pay to have photographers who do put effort in be able to copyright their work. And the photos would be shit anyway, so good luck making any money off them.

So why don't prompt engineers get copyright on 100 images they effortlessly generate?

because generating those definitionally takes no effort, rather than there being circumstances (which with photography i would say are not the norm) where this is the case.

If you believe AI generation is effortless try to generate an image that will win any AI art competition and if you succeed not how much time and effort it took.

I'd say it's same as for photography. There's huge gap of effort between taking a picture of your ceiling with your phone and taking a photograph that people appreciate.

> How's Henri Cartier-Bresson different from prompt engineer

Does the prompt engineer gets the copyright ? Or does the company (i.e. the photo camera) ?

When is something "purely" AI-generated?

I suggest to read the article because it's quite detailed about it

But incomplete, since it doesn't address provenance.

When the model was not trained on human creative output.

Art produced by an model that was trained using Google Street View might be an example.

Now AI “artists” are worried about protecting their “content”?

Lol