How do you tell when something is AI art?

[UngainlyTitan]

This is trivial for them, but it’s a waste of processing power to have an AI produce arbitrary art no one is asking for.

That is not what I am talking about, that is still a human mandating something. I am talking an AI producing art of its own violation for its own reasons.

 

[Andvari]

That is not what I am talking about, that is still a human mandating something. I am talking an AI producing art of its own violation for its own reasons.

I don't know what that means. Like an AI that doesn't have the programming needed to make it capable of producing art on its own, but magically does so anyway? Or an AI created by another AI to produce art on its own, and therefore it's another AI mandating something, rather than a human?

 

[Cergorach]

How do you tell?

You don't.

AI art is pretty much an math question instead of an art question, we're using extremely powerful computers to first train a (math) model and then run slightly less powerful computers to actually run the models. You also don't do "x³+y³+z³=42" (Diophantine) from your head, you require a a supercomputer (and 65 years for people to make that supercomputer and figure out the math to do it on such a machine).

Why do keep people insisting that they can see what's AI art and what's not? Because it's "baby's first MS Paint job" at the moment, but they grow up so quickly! And then they only see what's flawed AI art and not what's not, or they resort to guessing. Going back to MS Paint, most of us produce horrible MS Paint jobs, but there are people who produce truly stunning artwork in MS Paint. If you didn't know, you wouldn't think it was made in MS Paint.

If people have no context in 'Art', would you rate "Birds and Sun (1954) - Karel Appel" as art by a famous artist or a finger painting by a slightly troubled 5 year old? https://www.tate.org.uk/art/artworks/appel-people-birds-and-sun-t04163

Throwing all 'AI Art' on one big pile is also not a good idea, just like throwing all 'Art' on a big pile is a bad idea. You have different artists with each their own styles, strengths, weaknesses, expertise, and popularity. 'AI Art' is the same, each solution has it's own styles, strengths, weaknesses, expertise, and popularity. Expecting to recognize them all without context is insane! Especially when development is so extremely rapid. You need tools to recognize patterns we can't see/recognize, you might even need many tools. And those tools might not always be as perfect either. See this as an arms race with 'AI Art generation' on one side and 'AI Art detection' on the other. A bit similar to IT security vs hacking, you implement a security feature and the question is not if someone will find a way around it, but when.

A better question would be, why would you care? All the clothing you buy is not handmade and hand weaved, the industrial revolution pretty much killed that industry (almost 300 years ago). Do you want to go back to getting your books hand made by monks that copy them one by one and illustrate them by hand? Might kill ENworld Publishing in production costs... Most artists, especially in the RPG industry already use digital/automated tools (often with some sort of AI process).

Let me give a personal example from here, ~20 years ago there was someone on here that made a map, I paid for what he made (after the fact). How did he make it? He made it from clay, photographed it and then did digital shenanigans to it. Vs. all the different mapping tools that are now commonly used and might even been used in your products. At what point do you say, I'm paying a fair hourly wage to someone to handcraft a map from clay, photograph it professionally and then make that into a presentable digital map? Or you pay someone who is making some very good and functional maps in an art-style you love in Wonderdraft? Would you pay both the same amount of money for a map? Would that be fair? And what of your customers, would they be willing to actually pay for that (vs. what they say)?

We, as humans have limitations on what we can do ourselves. Be that math, art, etc. But also in how flexible our thinking is (in accepting new thing or new realities) and that is very well illustrated by the current AI 'issue'.

 

[Morrus]

A better question would be, why would you care?

Why do people keep correcting my question?

 

[Prime_Evil]

The more you know about a subject or medium, the better the AI prompts you can produce. If I am trying to produce a photographic image via AI, I will specify details such as camera type, focal length, shutter speed, film stock, etc. I might go as far as including details like "ALEXA Wide Gamut color space, Log C gamma curve, 12-bit color depth, chroma sub-sampling, cinematic color grading, unified color scheme, black balancing, HDR 12, HDRI". For this reason, I suspect artists will produce far better works via AI than muggles. They will also have the skills necessary to manually correct any issues that arise. I suspect that in the long run, AI may augment artist workflows rather than replace them. The same is true with tools such as ChatGPT. Most Large Language Models produce awful prose riddled with passive voice, subordinate clauses, and unnecessary adjectives. But unless you know what you are looking for, you won't notice these deficiencies. But this is just my own opinion and YMMV....

 

[Cergorach]

Why do people keep correcting my question?

You don't ask if you should jump, you ask how high!

But I answered your question with "You don't" and the implied answer: Maybe with technological tools that are made for it (and not every tool is created equally, there is a lot of crap around) and they need to be constantly updated.

As for my question (about caring):

• Public opinion, that might influence your sales.
• Quality, either not being able to create the prompts you need to create the art you want or the solutions you use just don't produce the quality you want.
• Personal preference, you prefer to give artists money instead of a company that rents out it's AI art creation abilities.
• Legal reasons, as in they are not exactly clear at the moment.

Most of these are kind of out of our control. But if you prefer to pay artists for their art, why would AI art detection be an issue? You trust your artists, don't you?

As for other publishers you buy from, if they say they don't use AI art, you trust them... Right...

 

[Ryujin]

I finally managed to find an image that I recognized as being AI because it was "too good", so that I could offer an example.

 

[Andvari]

I've seen threads in places where people post a series of pictures and ask others to identify which one is AI-generated. For most of those I've seen, the AI-generated picture is usually a low quality one that at least to me was easy to identify, but even for those, at least half of those who reply turn out to guess incorrectly.

The one above I don't think I could determine was AI unless I studied ornothology.

 

[UngainlyTitan]

I don't know what that means. Like an AI that doesn't have the programming needed to make it capable of producing art on its own, but magically does so anyway? Or an AI created by another AI to produce art on its own, and therefore it's another AI mandating something, rather than a human?

@overgeeked pondered on the meaning of "art" in the context of an artificial entity producing art. My response was that this is not a factor unless the AI was doing it spontaneously. It is not a comment on the main topic, nor any kind of prediction regarding current AI nor, I should add, an observation on the potential development path of current AI.

 

[Ryujin]

I've seen threads in places where people post a series of pictures and ask others to identify which one is AI-generated. For most of those I've seen, the AI-generated picture is usually a low quality one that at least to me was easy to identify, but even for those, at least half of those who reply turn out to guess incorrectly.

The one above I don't think I could determine was AI unless I studied ornothology.

In this case I recognized it as being AI because, as a sometime professional photographer, the probability of getting a group shot this good is vanishingly small. I can count the number of times I've gotten the like on the fingers of one foot.