AI/LLMs AI art bans are going to ruin small 3rd party creators

[pawsplay]

There are different kinds of taste leaders. There are people out there who have strong opinions about the McDonald's menu.
If you think there's no such thing as a good generic elf ranger illustration versus a bad one, perhaps that argument holds. But, while people might have different preferences, for any given set of preferences, there are pieces of art that are going to be more or less successful at generating an aesthetic response.

 

[Justice and Rule]

This sort of reminds me of the argument a friend of mine used to make...a professor of history and political science...that it is important for political parties to have strong influence over elections because they serve the role of vetting candidates to make sure that qualified people get into office.

This was from about 2013.

Oops.

I mean, as a guy who has a degree in political science, he wasn't wrong. We're just sort of living in the failure of that sort of gatekeeping. And that's all I'll say on that to avoid becoming more political.

In fact, the more I think about this, the more non-sensical that argument is, and the more I'm surprised that anybody "liked" it. (Except that often we click 'like' on things that agree with our general stance, without necessarily interrogating the logic itself.)

The phrase "but that's why taste leaders exist" suggests intentionality, as if people have gotten together to agree that this is an important rheostat, and therefore appoint certain people to be taste leaders. As far as I'm aware, authoritarian societies that have attempted to do exactly that have not produced great art.

Rather, taste leaders emerge because there is a market for them, and they emerge through market forces, I suspect because people are afraid of making poor choices and want a trusted source, or they don't even realize they are being influenced by taste leaders. But it's all market driven; there's no guarantee that the people get in that position actually have good, you know, taste. Surely we don't have to argue about whether the historical landscape is littered with "bad taste" that nevertheless became exceedingly popular.

Taste leaders are absolutely no guarantee that good taste prevails.

I'm not sure that's really what @pawsplay is saying as people come to the fore as taste leaders. The idea that it's somehow done with "intentionality" feels like you putting something in that isn't there, and only for the purpose of making it mutually exclusive to having market forces. It feels like their statement perfectly blends with the idea of market forces creating taste leaders because there is a desire for people to help come to decisions, to prejudge whether something is worth their time, or help them understand why they might connect with a piece of art.

At the same time, while you might say this could create taste leaders who have bad taste, but that also exists: there are plenty of blocks that will disagree as to what makes a good movie or a bad one, a good show or a bad one, a good book or a bad one. There will always be niches that want representation and go to people who like, I dunno, Zack Snyder films (just to use an example beyond AI art boosters), where a lot of people seem to disagree with them. Ultimately tastes shift and we can reappraise certain things, but living in the here and now we probably can accept why certain things are popular versus unpopular. We can also see where critics can differ with audiences, and how that might play out over time.

But at this point, it feels like we are really stretching things to try and fit AI into the idea of "art", saying stuff like "most art is purely commercial" which is basically trying to dismiss anything that might get funded by a studio. Like, Netflix might not be the most artistic studio, but I think telling everyone that their art is purely commercial misses what the creatives put into their work. But I also feel like that idea links up with the serious disdain for creatives that seems to be an undercurrent of this whole debate, as though they were somehow gatekeeping artistic skills by pointing out the problems with AI art and why it shouldn't be used.

 

[Bill Zebub]

But I also feel like that idea links up with the serious disdain for creatives that seems to be an undercurrent of this whole debate, as though they were somehow gatekeeping artistic skills by pointing out the problems with AI art and why it shouldn't be used.

To a large extent I agree with, or at least sympathize with, what the creatives are saying, But my experience so far is that to not agree 100.0% is interpreted as loathing for creatives combined with stupidity.

 

[Morrus]

I'm bored of this now, I will disengage.

That’s how it works.

 

[Umbran]

My point was that once upon a time Wikipedia had lots of errors, and people made a lot of noise how you couldn't possible trust it. But people used it anyway, the world didn't end, and Wikipedia got better over time.

Except already back in September, OpenAI admitted that generative AI hallucinations are mathematically inevitable. They are a result of the fundamental math of how the thing operates, and cannot be removed by improving engineering.

OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws

In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limits.

www.computerworld.com

 

[pawsplay]

Also, a true fact and a convincing lie are equally "successful" from an algorithmic standpoint.

 

[dead]

Sort of like an axe, sword and saw, but all three are sharp tools. It doesn't matter if they have different names, they still all function as a form of automation.

Not really. For example, I have a really good intuitive grasp of what looks good. I can't draw it, but I could get there via directing someone else. Maybe that's intuitive skill.

That's what they would be volunteering to be. It's not something I would ever do or expect someone else to agree to do, but it was your hypothetical.

Edit: Two things.

First, the whole comparison of a living artist to a machine tool like AI is a False Equivalence(An AI isn't a person or artist) and a Red Herring. It's nothing more than a distraction from the real discussion.

Second, nobody is going to do your hypothetical. A person with the creative idea isn't going go to an artist, direct him every step of the way, and let the artist control the creation, so it simply won't be done. That or the person with the creative idea will talk a friend who is willing to let the creative person also have control of the final product, or pay the artist for the final product. Nobody is going to be treated like a slave in this distraction hypothetical of yours.

I think you’re taking for granted a lot of conscious choice that an artist makes - colour, proportion, composition, how light falls on a subject. If you’re not trained in all of that, your exact vision will be lacking (at least for professional use, anyway). A real artist might be able to humour you and bite their tongue and apply none of their skill and knowledge to help you out as they reproduce your vision. You may be disappointed with the results though. An AI would be a better choice as it won’t cripple itself. It will do its level best to help you out where you lack an understanding of light and form and colour, etc. It’s not as simple as “make eye bigger”, “make mouth smaller,” and, now, oh, I’m 100% the creator of this image down to the last AI-generated brush stroke. I would love to see one of your AI-generated “exact visions” where the AI hasn’t helped at all with any artistic knowledge along the journey.

Either way, it looks like you feel like you can claim ownership of the work. If you used AI, do you think you need to disclose its use? Why would you? It’s just a tool like a paint brush, right?

 

[Maxperson]

I think you’re taking for granted a lot of conscious choice that an artist makes - colour, proportion, composition, how light falls on a subject. If you’re not trained in all of that, your exact vision will be lacking (at least for professional use, anyway). A real artist might be able to humour you and bite their tongue and apply none of their skill and knowledge to help you out as they reproduce your vision. You may be disappointed with the results though. An AI would be a better choice as it won’t cripple itself. It will do its level best to help you out where you lack an understanding of light and form and colour, etc. It’s not as simple as “make eye bigger”, “make mouth smaller,” and, now, oh, I’m 100% the creator of this image down to the last AI-generated brush stroke. I would love to see one of your AI-generated “exact visions” where the AI hasn’t helped at all with any artistic knowledge along the journey.

I don't need to be trained to know what looks good. Personally, I have a good eye for that sort of thing and would not be disappointed in the end. The artist that had to constantly redo the entire piece of art to change it to my vision as I tweaked it would likely strangle me in the process, though.

Either way, it looks like you feel like you can claim ownership of the work. If you used AI, do you think you need to disclose its use? Why would you? It’s just a tool like a paint brush, right?

Of course you do. Why? Because a lot of people would use it to just do all the work and generate pictures from a prompt. You would need to show it to differentiate your true AI artwork from the hordes of pictures that aren't AI art.

 

[Bill Zebub]

Except already back in September, OpenAI admitted that generative AI hallucinations are mathematically inevitable. They are a result of the fundamental math of how the thing operates, and cannot be removed by improving engineering.

Whereas Wikipedia's improved processes has made it immune to false edits, right?

Touché. I concede defeat.

 

[Umbran]

Whereas Wikipedia's improved processes has made it immune to false edits, right?

Of course not. However, Wikipedia can always improve its material, but there's a point beyond which LLM output will not get better.