Discussion about A.I.

[fallenmystic]

You repeat that claim but fail to reply why the large companies should be "above law".

Don't put words into my mouth. I never said large companies should be above the law. Companies and researchers have used publicly available images for ages since there was no law explicitly prohibiting such usage by non-human agents.

The voice urging regulation only recently began to gain traction, and the last time I checked, it was still a hotly debated matter. In short, it's still in a grey area until they make a law that prohibits such usage. You can express your opinion in favour, of course, as I did in objection in my post above.

However, if you paint it as already illegal to train AI over such images or say that AI users like me are either breaking the law or promoting others to do so, you are just misleading people.

And how should this change the situation that material cannot be used for machine learning unless there is consent by the copyright owner?
It would just be a bit more difficult to enforce, as the companies then would have to provide their own databases, along with proof of consent.

Because it'd be both impractical and pointless at that stage. When an AI agent can spontaneously collect visual data, identify objects from it, and improve its understanding of the world through the process, it's basically doing what humans do to learn what things like "car" or "orange" mean, i.e. by looking at them and hearing what other people talk of them.

At this stage, it doesn't need to be given photos of oranges and cars all cropped to the same resolution and tagged with the corresponding words to learn such concepts. Then how do you define "stealing images" if all it does is look at some photo just like people do?

Humans can look at images from the internet and obtain information without restriction. You may even inspired by them to create your own art as long as it doesn't clearly clone the original style. And if AI agents can do most of what people do in the way they do, where do you draw the line between "art theft" and permitted operations? Is it stealing art when it looks at a photo? Or is it when it learns something from it? Or maybe when it creates something from the information learned by looking at it? Do you think it'd be feasible to give AI instructions like "When you need to browse the internet, turn blind eyes to any images you see"? Would it make a good AI agent if you can do that?

It's much easier to enforce copyright on what people produce and publish rather than on publicly available source material based on future usage or by whom it'll be used.

If you believe that an open source AI can be even remotely relevant in comparison to one of the large players, that really is wild assumption.

If you believe it differently, you obviously don't know much about the history of the open-source movement or its implications for the AI field.
As I mentioned several times in this thread, it's simply an irrefutable fact the free and open-source movement has won big time over the big corporations in the past. And whether you are aware of it or not, we've all benefited immensely by this victory. In short, the whole IT industry has been literally dominated by the open-source community for the past few decades.

Granted, the advent of AI poses a new challenge to this trend since training AI models requires centralised hardware infrastructure that only a few large corporations can afford. Despite this difficulty, however, the open-source camp has been rapidly closing the gap with proprietary competitors like OpenAI for the last few years.

In fact, ALL the models I use for generating images and text are open-source variants. And when making kinky art is concerned, you simply have to use open-source models because all proprietary models are heavily censored.

You should have known this basic fact if you were to dismiss what I said about this subject with such confidence.

Tell me how you protect that style from being taught into an AI. It happens already, as DavidTX has outlined, as a revenge on artists who dare to open their mouth against abuse of their art..

Again, you seem to have missed my point. The quoted part was about developing your unique style to make it stand out among mediocre AI images most people generate, not about protecting it against copyright infringement.

As to the issue of copyright protection, I repeatedly pointed out that copyright infringement is copyright infringement, whether you copied the original by hand or using an AI tool.

The race is on because its a must, because the obvious theft is being tolerated by politicians.
Besides, don't mix up things. AI capabilities can also be filters, thats a completely different thing. Generative AI is only a subset.
AI can be a great help for example when generating seamless textures, as it is much better at detecting and eliminating repeating patterns than the human is.
I have yet to know somebody who would complain about this use.

If it's a subset, then what? Are you claiming that users of those tools will refrain from using the generative functionality? Have you seen how many Photoshop users suddenly changed their opinions on AI after it introduced a generative AI feature?

When practically all popular art tools are either implementing or planning to implement the generative AI feature, I don't understand how saying that there are also other AI features would strengthen your argument in any way.

 

[windar]

However, as I argued at length above, this assumption is flawed. Do you think photo manipulation work can be legitimate art? And what about Daz3D renders? In both of those cases, the author does not create any of the assets they use (at least in 99% of cases) but only arranges them, which you also do in AI art. AI tools allow far more controls than either of those mediums do, so I don't think you can justly deny AI as a legitimate tool of art while approving of the others.

I'm not expert enough to say what is art and what isn't. I'm rarely moved by photomanipulation and likely will always prefer hand-created art to anything done on a computer, whether by DAZ or AI or anything else. But overall, I am really much more into stories than anything visual. And so far at least, no story I have seen done by ChatGPT or other AI comes off as other than bland and generic. The words are there, the grammar is fine, but something is missing, not that that isn't the case with many stories written by humans, also.

Part of it is the experience of seeing the art in a public space, like a gallery or museum, rather than sitting alone in front of a screen. That's how musicians have found ways to make a living-live performance, something which a great many people crave. Bands used to tour to promote their albums, now they release albums to promote their tours.

So when @DavidTX says there is no way to protect your art from theft, if you produce physical art there is, at least to a reasonable degree. Exhibit in galleries, public spaces, etc. Most ban photography, so it would be hard for AI to access. Keep it off the internet. Of course it may be difficult to make a living doing that, but hasn't making a living as an artist always been difficult? The ones I know generally have teaching jobs that pay the bills. The same is true generally of writers, and this is long before AI came along.

 

[fallenmystic]

I'm not expert enough to say what is art and what isn't. I'm rarely moved by photomanipulation and likely will always prefer hand-created art to anything done on a computer, whether by DAZ or AI or anything else. But overall, I am really much more into stories than anything visual. And so far at least, no story I have seen done by ChatGPT or other AI comes off as other than bland and generic. The words are there, the grammar is fine, but something is missing, not that that isn't the case with many stories written by humans, also.

I'm not a native English speaker, so I may not be a good judge. However, I can easily imagine what you said above to be true because it's not incompatible with what I mentioned about how the level of control determines how much "human" what AI generates can be.

Unlike the case with image generation, in which you can control almost every aspect of it to its finest details, generating stories relies much more on randomness. You usually write things like the general outline of the plot, a summary of each character's personality, etc., in a prompt, and then AI does the rest. As such, I wouldn't argue that AI-generated stories can be as good as those written by skilled human writers, at least with the tools we have and how people use them now.