Adobe Firefly: AI Art Generator

Yeah I was just using DALLE2 for the first time in awhile because I hoped it would integrate more simply alongside my code for the GPT4 api. Results are just objectively underwhelming after having used stable diffusion for so long.

It’s also still under beta and given the recent codex termination, I wouldn’t be surprised if they get rid of it.

It's fine we disagree. Let's avoid the trap of right or wrong. Ai is new, and our opinions will evolve over time. With that said...

> "indisputably Hayao; any fan of anime will recognise his style immediately"

No they wouldn't. If you're talking about a still frame, nobody can be certain who the artist is. Why? Because LOTS of artists make similar work. You think there's an indisputable signature in every drawing from every artist? There really isn't.

There may be an obvious signature, such as placing an "upside down umbrella" in every piece they make. But the Ai won't copy the umbrella, because that would be replicating the work rather than style.

There are books about Miyazaki's style, authored by him that go into detail and share everything from concept sketches to the use of computer graphics in his films. It's all there in the books he wrote explaining and SHARING his techniques with the world. But what he can't share, and what can't be replicated with prompts, is the brilliance of the storytelling that threads the vast body of visual work to create the whole.

> unique style

Again, you're claiming an artist's style is unique, when in fact the vast majority of "styles" are blends of other styles and influences. No artist was born in a bubble.

Even if there is a unique visual signature, so what? It's not the heart of the work. It's a preferential technical characteristic.

Not only that, but this world is saturated with images. Even in music, it's difficult to come up with an original riff that doesn't sound like an existing song. The musician, in order to be awarded "originality", needs to use the full 3 minutes of the song to involve lyrics and other elements before something truly "unique" emerges.

You think there's an infinite bucket of "styles" visual artists can tap into? There isn't.

Content choices and narrative choices are more important anyway. For example, the artist may have something to say about humans being cogs in the machine. Sounds interesting. Such concepts form part of the style and reputation of the artist. People follow artists largely because of their artistic substance.

You're too focused on visual technique. The guts of any good artwork is the whole package, including the symbolism and balance of meaning in the work.

What is the artist SAYING with the work. Ai won't help prompters with quality of symbolic meaning, concepts and interlinking symbolism, subtext, balance of imagery, subtle visual tension... all the things that make ART ART -- the ESSENCE of art is found in the body of work.

So... Ai can generate a sci-fi robot running through the city, and your concern is this impedes the profit margin of real artists pumping out the same generic stuff? Sorry, but I won't shed a tear for the loss of artistic mediocrity in a world saturated with stock images.

Artists are adaptable, and should be doing a lot more than selling robot images to stock art collections anyway. Time to level up. Nothing wrong with Ai keeping us on our toes.

> If this kind of abstract copyright regime of 'I had the idea first, and anyone who uses a derivative of my idea must pay me money!' is a very sillpery slope

This is a cartoon conception of both existing copyright and the proposal I described.

Copyright applies to works, not "ideas" (William Gibson doesn't have a copyright claim on the "idea" of a virtual reality but he does on _Neuromancer_ as a work). It gives creators incentives and a stake in their work by allowing for some degree of control in how works are used and especially how they're distributed.

What I'm proposing isn't that different, and it's simple:

People should be able to decide if their work is put into a training set, and what they get in return for it.

That's it. No "flavor copyrights." No ownership of ideas. Just a claim on the creator's work and how it's used/ distributed. Specifically updated to cover a case that didn't exist 10 years ago, a case whose explicit intention is reproduction of not just one narrow aspect of the work but a combinatorially large number of aspects of the work (there's no other reason for adding it to the training set, that's the nature of these models).

> it will become a spotify where each creator gets a sum total of $0.00000000001 per AI their media item was trained on

"the artists will get such a small payout so we should make sure they get nothing" is quite the take.

I think "artists should be able to negotiate what consideration they want in return for their work for a training set" is a better one. It's certainly better than "people collecting training data should be able to take anything they can get their digital hands on without any consideration other than possession." Maybe some artists would take a nanocent per use. Maybe they wouldn't. That should be up to them.

Part of the reason why services like Spotify are so terrible economically is the way digital rights were assumed by labels and streamers often as extensions of older pre-digital conceptions without much in the way of negotiation by artists. The current moment is a chance to do better across an equally large (if not larger) technological shift.

It'll be useful, I'm sure, but if it can't match styles then you can only use it with itself. Clashing styles inside the same image is a terrible look.

What lot of folks are missing here is just how easy it is to add "AI" to products for large companies. The APIs are prevalent and easy to integrate with. Right now we're in the hype cycle, but soon people will be finding and reporting on all sorts of limitations and security issues.