• Quick note - the problem with Youtube videos not embedding on the forum appears to have been fixed, thanks to ZiprHead. If you do still see problems let me know.

Merged Artificial Intelligence

It's equivalent in the sense of transformative work. In art it's normal to remix works of other people. From collage to simple inspiration. And that's the same with AI models. The might or might not think like humans, that's not important. What's important is they also transform works of other people, the same way many human artists do.

But the Disney case is not about imitating art. It's about selling service which can create copyrighted characters.
 
If we're looking for equivalences, I think the following is more representative:

If I were doing something for a company, for example programming, that company would have some recourse to protect that work from being exploited by its competition. Artists also do work, but they never imagined they would have to protect it in a similar way, because the technology didn't exist, so they have no recourse to prevent such exploitation.
 
It's equivalent in the sense of transformative work. In art it's normal to remix works of other people. From collage to simple inspiration. And that's the same with AI models.
No, it's not the same with AI models. AI is a machine, not a human. A human artist is allowed to reproduce an existing work provided they do it by their own hand, as this involves skill and creativity. Using a machine to copy works is a different story. That's what the 'copy' in copyright is about. Collage can run afoul of copyright if it uses copyrighted works.

The point of copyright is to prevent someone from using a machine to make copies of someone else's hard work, depriving them of the income they need to do that work. AI doesn't create art, it takes existing works and modifies them using various algorithms. Right now it needs artwork created by humans for input, otherwise the output quickly devolves into meaningless gibberish. In the future it may become 'smart' enough to largely overcome this limitation, but that's beside the point.

The point of copyright is to encourage human creativity. The artwork they create becomes part of our culture and our consciousness. If we replace that with art generated by machines we can't be surprised when human creativity goes down the toilet and we become more like machines. I for one do not want to see that happen.
 
tools and ideas are fuel for new expressions of creativity. AI doesn't need to be in opposition to human creativity, it can be useful to make up for lack of skill in certain areas as well as generate a lot of variations quickly on which to base the final work.

I find it also useful to get away from the idea that something has to be A Masterpiece. It perfectly enough if it manages to invoke strong and new emotions and thoughts.
 
tools and ideas are fuel for new expressions of creativity. AI doesn't need to be in opposition to human creativity, it can be useful to make up for lack of skill in certain areas as well as generate a lot of variations quickly on which to base the final work.
Problem is you are using the skill of others to make up for your own lack of skill. If everyone did this what would happen?
I find it also useful to get away from the idea that something has to be A Masterpiece. It perfectly enough if it manages to invoke strong and new emotions and thoughts.
Most real artwork isn't a masterpiece, but skill improves over time as the artist does more work. You see a masterpaice, but not the dozens or hundreds of previous works they created while developing their technique. Replace that with AI and there is no incentive to develop artistic skills. It's replaced with the 'skill' of promting the AI to produce works that look pleasing - which is something else.

Next step after that is to use another AI to chose the works, and another AI to appreciate them. Then humans aren't needed at all!
 
Problem is you are using the skill of others to make up for your own lack of skill. If everyone did this what would happen?
A lot of people would be employed in skilled professions.

I have no skill in plumbing. When I need some plumbing done, I call a plumber. I have no skill at art. When I need some artwork done, I call an artist.

Unless you are skilled in every skill ever, you're always going to use the skill of others to make up for it.
 
I find it also useful to get away from the idea that something has to be A Masterpiece.
I agree, which is why human artwork is preferable because someone's time and effort has gone into it.

It perfectly enough if it manages to invoke strong and new emotions and thoughts.
This doesn't follow the first sentence.

Thoughts and emotions inspired by AI generated images are wasted. They're a wank, and based on nothing.
 
so what?

ever-present Art, so cheap and common that we just assume that everything needs to have some artistic value, would be a good thing, even if most of it is done with minimal direct human contribution: we need to make our living spaces beautiful again, get people to expect more, and make Art just another type of mass media, to be consumed and talked about.

People believed that recording music would be the end of most musicians, and the opposite was the case.

At the end of the day, what matters is that the Artist, not the AI company, is getting paid and recognized.
 
A lot of people would be employed in skilled professions.

I have no skill in plumbing. When I need some plumbing done, I call a plumber. I have no skill at art. When I need some artwork done, I call an artist.

Unless you are skilled in every skill ever, you're always going to use the skill of others to make up for it.

Which is fine, as long as you pay them accordingly.
 
Then you shouldn't keep bringing up things like human artists analyzing images as some sort of equivalence.
For relevant parts of an argument it is entirely appropriate to bring up how copyright is currently being dealt with, especially when discussing the Disney & Universal legal action that is about outputs not inputs based on current copyright laws and legal precedents and would apply no matter the mechanics of how the copyright breaching images are produced. Which is why I don't think Midjourney has a leg to stand on and why Midjourney has been singled out by D&U, and in my view rightly so. As far as I am aware all of the other USA based image AI companies do try and ensure that their generators can't reproduce works with recognisable copyrighted elements.
 
It's not copyright infringement if it's done by hand? I'd like to see a citation for that, please.
What is Copyright?
Works are original when they are independently created by a human author and have a minimal degree of creativity. Independent creation simply means that you create it yourself, without copying...
And always keep in mind that copyright protects expression, and never ideas, procedures, methods, systems, processes, concepts, principles, or discoveries.

The word 'copy' may be confusing you. In New Zealand copyright law it is defined as:-
copying—
(a) means, in relation to any description of work, reproducing, recording, or storing the work in any material form (including any digital format), in any medium and by any means; and
...
(d}includes, in relation to a film or communication work, the making of a photograph of the whole or any substantial part of any image forming part of the film or communication work—
In this context, 'copy' means a duplication of the original by some mechanical means. In the case of digital artwork that generally means copying the bitmap electronically, not by 'eyeballing' it and attempting to draw something similar by hand.
 
It's equivalent in the sense of transformative work. In art it's normal to remix works of other people. From collage to simple inspiration. And that's the same with AI models. The might or might not think like humans, that's not important. What's important is they also transform works of other people, the same way many human artists do.

But the Disney case is not about imitating art. It's about selling service which can create copyrighted characters.
But for AI it's more like a lad going into the Louvre, robbing a hundred or so prints of the Mona Lisa, squiggling a curly moustache on each and selling the resulting "work" off as his original creations.

Edit: Or Greg Land.
 
Last edited:
It's not tracing machine. Diffusion synthesis can't reproduce original picture. But Disney doesn't own specific pictures. They own the whole idea of Darth Vader. I can't draw a Darth Vader and sell it. The issue is not that the AI can draw Darth Vader, nor that it used images from the web to learn. The issue is they provide generator of Darth Vader for money, and Disney doesn't get a cut.
It's possible to have list of copyrighted characters and styles, the same way AIs already try really hard not to be able to do celebrities, by filtering them from training sets or even poisoning the data set by mistagging the images.
But IMHO the system where copyright owners would get a cut would be much more feasible. Multimodal AIs understand what they see. You can give them 1 image of a person or a character .. and they will spit out any variation you want.
Unbelievably the biggest vampire in the creative digital world, Adobe has taken what I think is the most ethical stance on the dataset they used to train their generative AI Firefly, artists have been paid for their images being used.
 
As original as most people. Certainly original enough for pop.

I got recently trolled hard. I decided to make a collection of light japanese city pop, to get rough overview about the artists, pick 10-20 best songs and make a playlist. There's ton of playlists on youtube, so it's just listening, and marking down what you liked, and checking more from those artists.
And I found a goldmine. Channel with like 50 hour long mixes, and what's crazy, I like most of the songs. But they don't list authors .. only song names. So I google the names. Nothing. Some of them I could find, but it's different songs. I try to Shazam them, google audio search them, nothing. It's because the whole thing is AI !
I play with image generation a lot, so of course I immediately spotted the cover art is AI, but that's pretty common. They are also quite nice (minus the texts). But the music !? I listened to it for few hours before I started to be suspicious, and the suspicious thing was how uniformly I liked it, which is usually not the case with the real music.

Here's one of the videos:
That's amazing. The music sounds real. The Japanese is legit. At least it sounds legit (maybe easier to understand than real music. I often have trouble understanding lyrics, even in my native language.) The thumbnail image also looks pretty good to me at first glance. Until you look closer at the letters on signs and license plates. Then you realize that's not real writing, it just has a superficial appearance of it.
But if you told me this is City Pop, I would agree.

I got curious to see if there's something similar for Country Music, because that's in English. This sounds pretty generic but it's almost like the real thing:

ETA: I'm also amazed by how positive all of the YouTube comments are. I didn't read them all, but I don't see anyone calling it AI Slop or even mentioning that it is AI generated. Do they realize?
 
Last edited:

Back
Top Bottom