The Fraternity of Shadows’ Position on AI Art

[TheArcaneWeirdo]

So few are willing to do so.

Generative AI as its proponents and creators define it replaces "art" by dispensing with the very process of creating art. Thus, to incorporate AI into the process of creating art defeats its very purpose. The wannabes and the AI Bro bros who vociferously tout AI as the inevitable, all-destroying wave of the future (of art) never do so because incorporating AI into art requires a level of effort and focus and competence they have no ability nor interest in meeting, and very few creatives bother to advertise that they use AI in their works for fear of rightly attracting negative focus.

We still have not seen the Mozart of AI music, Michaelangelo of AI painting. But we have seen the AI paragons of pure math, which has been like a nuclear bomb going of in that field of study. There is a narrow focus on how AI affects artists, while little thought is given to how it affects other people, and far more severely than the art field. We will likely have huge breakthroughs in technology and material science thanks to AI math, and no one will think to reject these advancements just because they were created by AI.

The difference is this: a scientific discovery is objectively wrong or right. You can't deny the truth just because of the source you discovered it by.
Art, on the other hand, is inherently subjective. Eliminating the subject within it by automatization defeats the point of creating it.

[Baron Von Stanton]

We still have not seen the Mozart of AI music, Michaelangelo of AI painting.

Given how AI makes art, in that it gobbles up information and jpegs that are both scraped off the internet and handfed to it in order to regurgitate collages, we will never see any equivalent paragons in AI beyond crassly obvious plagiarisms and superficially aesthetically adequate compositions that do not have any details correct.

[Baron Von Stanton]

The difference is this: a scientific discovery is objectively wrong or right. You can't deny the truth just because of the source you discovered it by.

Actually, you can call a truth into question or invalidate it if the source is understood to be unreliable, like how fossils from the black market are useless to paleontologists due to their location of origin being unknown, or how researchers who have been discovered to use AI illustrations in their papers are then later discovered to have produced their findings and text with AI, too.

Art, on the other hand, is inherently subjective. Eliminating the subject within it by automatization defeats the point of creating it.

Automating the process of creation defeats the purpose of art because it does not allow for the manual arrangement of the composition beyond deliberate plagiarizing -preexisting pieces.
And this is what makes AI Art less than useless, and this less than uselessness is best seen in AI generated scientific illustrations and paleoart, where you get pictures of lab mice with human testes and, at best, three-fingered T-rexes with seven legs.

[Le Noir Faineant]

Math is about counting. Art, music or literature are about the human experience and about human expression. There can be no Mozarts or Michelangelos in generative AI because the process of creation itself holds no meaning. You know, the difference between you telling your children a goodnight story, or an LLM telling them a goodnight story in your voice. One thing is you giving general directions to a randomizer. The other is you expressing yourself directly and in the moment. One thing is not the other. -- Tech-Bros usually pretend not to get the difference, and their counter-argument usually becomes some variation of "The Narcissist's Prayer". Let's not be those people.

[tomokaicho]

The thread is about generative art, as Rock stated, not math. But there is no difference in principle. The same LLMs produce math in the same way that they produce AI images. An AI can also draw campaign cartographer style maps, using mathematical principles.

[Le Noir Faineant]

The thread is about generative art, as Rock stated, not math. But there is no difference in principle. The same LLMs produce math in the same way that they produce AI images. An AI can also draw campaign cartographer style maps, using mathematical principles.

There is a difference in application, though. It's like calling a butter spreader a sword: Technically, both are made in the same way, and technically both are knives, yes. Yet so, knights are usually not recognized by their copious display and use of butter spreaders.

[doctornecrotic]

Day by day, the Butlerian Jihad becomes less and less distasteful to me...
Maybe I'll earn my own Darklord claim yet...

[Joël of the FoS]

Lisan al Gaib!

[Wolfglide of the Fraternity]

Math is about counting. Art, music or literature are about the human experience and about human expression.

Actually, there is a great deal of human expression that goes into mathematics. At a certain level, I would say that the process for writing a mathematical proof isn't far from the process of making art, music, or literature (i.e. a lot of critical thinking, revisions, and quirky inspiration). I think there are a lot of misconceptions about the sciences and how they work, and their humanity and creativity often aren't readily apparent. In both math and art, AI can use its training to reproduce material which humans have created a lot of data for, but it still relies on humans to push the boundaries. I have neither seen nor expect to see an AI Einstein or John Nash.

In the meanwhile, this thread is not meant to argue the uses of LLMs in the field of maths, or the way it's affecting other fields. It is specifically about the use of LLM-generated imagery and text.

Sorry, it is extremely bad moderator form of me to keep derailing the thread. The mathematician in me was twitching uncontrollably.

[alhoon]

The Fraternity of Shadows’ Position on AI Art

AI-polished but human-written articles? I found myself nowdays often writing something and then giving it to the AI to fix sentences, typos, and doing minor edits. ChatGPT is useful for this as it gives in bold changes it made so I can see and verify them.
Is this allowed?

Major international scientific journals accept AI-polished scientific articles, that's why I ask.
And with disclosure, AI-translated articles (which are bad, BTW. The AI translation is marginally better than google from what I have seen in articles I reviewed; AI-polishing works especially with human behind it, AI-translation not so much).

We haven't thought about it, but indeed, perhaps the prompt will kept in a separate "source" document, so we can review it as editors, but it won't appear in the netbook.

Ahh... can we the people also have a look at the prompt? Suppose someone makes an image of an awesome mummy, in juuuust the right style, with pencil and charcoal or whatever. And we find we need a similar but not quite mummy. Say, wearing a crown. Or a skull-faced mace. Would be useful to have access to an addendum with the prompt.
Perhaps not the ENTIRE history of the prompt, that as others said may be 1 hour of prompt-working, but the final product or something.

[alhoon]

I understand the need to weed out AI slop. And the need to not publish cloned AI art which is signature to a particular artist. The best way is to just weed out anything that looks low effort. If it looks like someone could have belted it during their lunch break, just exclude it.

But also, at the same time, try it out yourself. See if you can go to ChatGPT with a prompt and produce a similar work right off the bat to test your suspicions of low effort AI slop. If ChatGPT just belts out an equivalent just like that, you've got your slop indicator.

Why, if I may ask? If I generate an image of my NPC like a portrait, and it is of acceptable quality why exclude it because the prompt was:
"Oil portrait of middle-aged man, with graying temples, strong jaw and piercing gaze dressed in quality but somber medieval atire." (Yes, the attire written wrong!) on chatGPT, without agentic help to look for images, nor parameters, nor using image-specific AI tools etc. Why reject it if it works?
Sure, if someone else uses it the AI of their choice would look on their history or what they usually do and make something different. But it would work for that author. So, why exclude it?

===========

Something else that needs to be stated:
- If you use ChatGPT, your own history and what ChatGPT knows about you, will be used. Don't expect AI prompts to give the same for two people with radically different interests.
- If you use some other programs or sites, their own tool may "enhance" your prompt! Thankfully the one I use gives me the prompt IT USED to generate the image (not the one I used). But you GOT to tell it "Do not copy particular artists!" else you would see in the prompt "In the style of X, Y, Z artist!"


And something else that would be helpful to those remiscing about the Butlerian Jihad here:
You can use AI to LOOK FOR OPEN ART. Yes, you can find portraits of 16th century gentlemen that are now open art and you can legally use them.

[Le Noir Faineant]

Actually, there is a great deal of human expression that goes into mathematics. At a certain level, I would say that the process for writing a mathematical proof isn't far from the process of making art, music, or literature (i.e. a lot of critical thinking, revisions, and quirky inspiration). I think there are a lot of misconceptions about the sciences and how they work, and their humanity and creativity often aren't readily apparent. In both math and art, AI can use its training to reproduce material which humans have created a lot of data for, but it still relies on humans to push the boundaries. I have neither seen nor expect to see an AI Einstein or John Nash.

My not-so-serious answer to this would be here: Imgur

My more serious, and hopefully more balanced answer, however, is that this really pretty interesting conversation lamentably takes us back to the Tech Bro baseline:

"LLMs can create the same results as people who have spent their lives studying liberal arts and humanities."

The misleading tangent being here - as I observe the tedious, years-long public debate on the topic, at least - that we're debating whether an accounting clerk wouldn't be equally able to create, whatever, cool paintings despite not going to, dunno, art school.

What we are actually talking about, however, is - or should be - is whether we want to continue using services by private companies that are provided through openly unethical means like content farming, and whether the content provided through their services should, by default and for the future, be accepted as equal to content created through more ethical means.

If we divide the problem like that, I'm pretty sure all of us are more or less on the same page.

(Except when you are, ugh, Azalin, or Daclaud Heinfroth, or something. )

[Rock of the Fraternity]

... something else that would be helpful to those remiscing about the Butlerian Jihad here:
You can use AI to LOOK FOR OPEN ART. Yes, you can find portraits of 16th century gentlemen that are now open art and you can legally use them.

But you can't trust it without double-checking its findings, can you.

[Rock of the Fraternity]

Ahh... can we the people also have a look at the prompt? Suppose someone makes an image of an awesome mummy, in juuuust the right style, with pencil and charcoal or whatever. And we find we need a similar but not quite mummy. Say, wearing a crown. Or a skull-faced mace. Would be useful to have access to an addendum with the prompt.
Perhaps not the ENTIRE history of the prompt, that as others said may be 1 hour of prompt-working, but the final product or something.

Or you could talk to the author of the article, talk to them about the things you liked, and ask them? I'm not saying a hard no, but ... usually the creators of articles love to talk about their process. I know I do.

[Rock of the Fraternity]

(Except when you are, ugh, Azalin, or Daclaud Heinfroth, or something. )

Or a techbro. Or a shill for those unethical companies.