Against AI Art

[Hornbein]

I used to go to deviantart.com to look at recent offerings. But as soon as AI generated art became available it took over completely. I can't get interested. I think it's because on the surface it all has a certain style, it all has the same texture, as though the same person did it all. Boring.

 

[Lyro]

I get what you mean. Once AI art flooded DeviantArt, the "recent" feed started feeling oddly samey - polished textures, similar lighting, that familiar AI look, no matter the subject. It all blurs together after a while. What's missing is the sense of a person behind the work - the quirks and imperfections that make human art interesting. Without that, it's hard to stay engaged.

 

[Hornbein]

Yes, the thing that interests me most about good art is it communicates how the artist feels about life, sees life. I don't get that with AI art. The big anime/comic book influence doesn't help either.

I hired artists for a book. Today EVERY pro artist uses a computer to draw, so I had no choice. It seems that schools no longer teach the trad methods. I was though happy with the result. The artist's soul came through nevertheless. And it was very inexpensive.

 

[jack action]

Something is happenning in French literature right now. A book called "It Was Either That or Die" by French-Canadian-Haitian Thélyson Orélien has been praised all over the world:

His fiction novel, "C'était ça ou mourir," which translates to "It was either that or dying," is among 16 books on the initial shortlist for the Goncourt, one of the most important prizes in French-language literature.

The novel follows a Haitian man named Jonas Dorléon, who is forced to leave his country and undertake a perilous journey through the Americas.

For a Quebec writer — especially a debut novelist — the recognition is highly unusual, said Michel Lacroix, a professor of literary studies at the Université du Québec à Montréal.

The book has also recently been shortlisted for two other prestigious French literature prizes: the Femina and the Renaudot.

"It's extremely rare, almost unheard of in the history of Quebec literature," Lacroix said.

Twenty-six countries have acquired the rights to the book, which is currently being translated into 22 languages, including English.

It is based on the author's and others' immigration stories.

Now, accusations that AI was used to write the book have been thrown around. Even though the author denies it, the book has now been taken off the list for the Goncourt.

In a statement Tuesday evening, he [Orélien] attempted to soothe the scrutiny of his heavy use of metaphors and similes in the book, which many have pointed to as tells that AI was involved in the writing process.

"I come from a Haitian and Caribbean literary tradition in which poetry, imagery, repetition, and rhythm carry a living voice," Orélien said in the statement written in French and released by Éditions du Boréal.

"One might debate my style, but treating its techniques as evidence of artificial intelligence has nothing to do with literary criticism," he said.

As a French-Canadian, this is big news on the morning shows and the like. The general sense seems to be that because this book could be AI-written, it is not worth reading anymore. That perplexes me. What was so good about it that it isn't anymore?

For Florent Michelot, a Concordia professor who researches AI in education, Orélien’s ordeal raises a key question: Have we entered an era where the final product is no longer enough when judging someone’s work?

“I didn’t read the book, but maybe [Orélien] developed very complex personalities for his characters, maybe he developed a wonderful atmosphere … and the AI only helped to write the sentences,” he said.

Michelot says he's personally tested Pangram, the AI-detector tool, and like many other experts who have spoken out since the controversy broke, he says the tool is "amazing."

Basically, the tool identifies patterns based on probability, he explains. If every word in an excerpt is statistically the most likely word to follow the one before it, then AI was most likely involved in the writing, according to the tool.

So, as more people lean on AI for things like research, spellcheck and translations, Michelot wonders if the creative process will begin to gain more value over the finished product.

French essayist and commentator Samuel Fitoussi, the first person to throw these accusations, said:

He said any authors who produce "copy-paste" from AI while claiming it to be their own work "betray the moral pact binding them to their readership".

As a reader, do I really have a "moral pact" with the author? Do I really care about how the work was created if it reaches me emotionally?

And, assuming the creative process is now important, is it OK to put the burden of proof on the artist's shoulders? Should an artist check his/her work in AI-detection programs and scrap it if they return false positives? What level of false positives is acceptable, in percentage and quantity?

 

[Greg Bernhardt]

There is a place for AI art (properly attributed), but in general it's off putting because it is generally derivative and quite frankly generated off the backs of stolen original work.

 

[Dale]

What was so good about it that it isn't anymore?

When you listen to a person’s own story then you feel empathy for the person. You feel connected to them by emotions that you share.

When you read the same words from a computer then you feel deceived. You feel manipulated into those emotions.

 

[jack action]

There is a place for AI art (properly attributed), but in general it's off putting because it is generally derivative and quite frankly generated off the backs of stolen original work.

In the case I'm talking about, the work has been praised by many and deemed worthy of prestigious awards. So it cannot be considered "derivative"; otherwise, all these people evaluating the works cannot be taken seriously.

As opposed to known AI art, the accusation in this case is that, after analyzing the text with a program, "it looks like something AI would have written". Even though no humans could tell the difference while reading it, before or after the accusation was made. Nobody can say "it looks like something [famous author of your choice] wrote".

When you listen to a person’s own story then you feel empathy for the person. You feel connected to them by emotions that you share.

As far as I know, the story, although based on actual events, has always been sold as a work of fiction.

And the author really went through the actual events. That was never questioned.

 

[Greg Bernhardt]

So it cannot be considered "derivative"

AI by definition is not derivative?

 

[Dale]

As far as I know, the story, although based on actual events, has always been sold as a work of fiction.

And the author really went through the actual events. That was never questioned.

I don't think any of that is relevant. The sense of connection is betrayed. It feels like discovering that you are being manipulated by a psychopath.

 

[DaveC426913]

AI by definition is not derivative?

No, he simply said a whole bunch of people deemed it worthy of awards.

Jack assumes (rightly or otherwise) that 'not derivative' was one of their criteria.

 

[jack action]

I don't think any of that is relevant. The sense of connection is betrayed. It feels like discovering that you are being manipulated by a psychopath.

There are plenty of biographies about someone else, authorized or not. There are also autobiographies with ghostwriters. Why does using AI to help you write your story - as opposed to another human being - make you a psychopath?

I can understand invoking a moral obligation to divulge who wrote it or helped write it, human or AI, but I failed to see your point from an artistic point of view, especially if it is recognized as being in his/her own voice by the "author".

If I want to show you a straight line and the ones I draw by hand are always crooked, so I decide to use a ruler to make one (or ask a better artist to make one for me). You can accuse me of cheating because you thought I'd done it myself, but I still feel that I passed my message as intended better with the help of a tool or someone else's help. Not only did I not cheat, but there will be no miscommunication between us as to what the line is supposed to look like. No psychopathy or deception involved.

 

[Dale]

You are missing/avoiding the key concept: connection

And I did not say that using AI makes the author a psychopath. I said that it feels to the reader like being manipulated by a psychopath, because they deliberately provoke emotions in others without feeling emotions themselves.

 

[A.T.]

Even before AI music, most music optimized for streaming was boring. And even before streaming most music optimized for charts was boring.

But still, the boringnees of it seems to be getting worse to me. Or maybe I'm just getting old and have heard it all.

 

[Matterwave]

If the author really used AI to help write the book and did not disclose it, I could see how readers would feel betrayed. The loss of that feeling of connection to the author is very understandable.

From a purely technological perspective, a significant portion of AI training is quite similar to (or some would even just say "the same as") textual data compression. So I could certainly see how some would view all AI work as derivative work.

But I also realize defining "the author used AI to write" becomes murkier and murkier. After all, if one simply talks to AI, then some time later, one can not be sure that some phrases from the conversation don't slip in, even subconsciously.

Drawing the line is difficult. From my personal feeling, without trying to be rigorous, I think if AI was used to actually put words to pages it should be disclosed. But if AI was not used to put words to pages, then it's unfair to legitimate authors that they are required to somehow pass their work through AI detectors which themselves can be quite flawed.

 

[256bits]

In a few years AI will become better at writing a good storyline, without the repetitive nature that most times will expose itself. A good story is a good story, but disclosure or category of AI written, or enhanced, should be the course.
Most of the news, sports outlets are using AI for their stories, and the outcry is minimal.

 

[Hornbein]

Something is happenning in French literature right now. A book called "It Was Either That or Die" by French-Canadian-Haitian Thélyson Orélien has been praised all over the world:


It is based on the author's and others' immigration stories.

Now, accusations that AI was used to write the book have been thrown around. Even though the author denies it, the book has now been taken off the list for the Goncourt.


As a French-Canadian, this is big news on the morning shows and the like. The general sense seems to be that because this book could be AI-written, it is not worth reading anymore. That perplexes me. What was so good about it that it isn't anymore?


French essayist and commentator Samuel Fitoussi, the first person to throw these accusations, said:

As a reader, do I really have a "moral pact" with the author? Do I really care about how the work was created if it reaches me emotionally?

And, assuming the creative process is now important, is it OK to put the burden of proof on the artist's shoulders? Should an artist check his/her work in AI-detection programs and scrap it if they return false positives? What level of false positives is acceptable, in percentage and quantity?

A similar thing happened long ago with Jerzy Kosinski. He wrote autobiographical fiction that was highly praised. Then it was discovered he had exaggerated his real-world victimhood in what he had told others and his work was declared no good. I come from music where it is OK to lie about one's biography, how much success one is having, and so forth so I didn't see what the big deal was. Either his fiction is good or it isn't. I don't care about the artist's stated biography.

In my experience AI-generated musical performances are better than what the great majority of humans can do. I dunno about song writing but performances, definitely.

 

[symbolipoint]

In my experience AI-generated musical performances are better than what the great majority of humans can do. I dunno about song writing but performances, definitely.

A person should see and hear a live performance of the artist of his interest. Many many of such performances are excellent; and the conditions assure that none of it were artificial intelligence. But maybe all of those which I had seen and heard were too long ago. Today attendees might not know if any A.I. was used in the compositions or arrangements.

 

[BillTre]

The issue of using AI in writing is a fraught and often misunderstood thing.
There are no flawless detectors of AI writing and the ways to prove you did not use AI in your writing are not that convincing to me. The use of AI in researching issues, spell checking and grammar to be OK is considered OK by some.
To me its a confused mess. (I have been using spell checkers for years since I am a lousy speller but for now am avoiding other AI uses.)

Here is an article by Jane Friedman on many aspects of this issue in publishing. She is my publishing info go to person.

Here is an excerpt for that article:

Why do some agents ask whether AI was used in a submission or to prepare submissions materials?

Agents who ask about authors’ AI use are either primarily concerned about copyright protection for the work (typically a necessary condition for selling any work to a publisher), and/or they do not want to work with writers who use AI in any way, shape, or form. However, given the complicated nature of AI and copyright as discussed earlier, it’s not true that if you use AI for certain tasks (brainstorming, outlining, editing) that you are giving up copyright protection. Some agents and publishers prefer to be cautious or take it off the table as a concern, assuming writers disclose and are honest about their AI use. Today that is a big assumption.


Can an agent or publisher reliably detect AI use in a manuscript?

AI detection software exists and is being used by some publishers (especially academic/scholarly and professional publishers) on contracted work, but it operates in probabilities, not certainties. Also, not all AI detection software is created alike; some tools are much better than others. The current industry leader is Pangram. Some AI detection sites try to scare people into purchasing “humanizing” services, which is a good way to identify the less credible services.

AI detection is often dismissed as completely unreliable, yet well-regarded studies have shown otherwise, at least in certain contexts. It is most reliable when evaluating long works, such as book-length manuscripts. To understand how AI detection works, this is an understandable and mercifully brief explanation. Furthermore, due to EU transparency laws around AI-generated content, Claude has added watermarks and digitally signed provenance metadata to outputs, offering another means of detecting AI-generated and AI-assisted content.

 

[DaveC426913]

As an aspiring writer, this is an issue that affects me. It is discussed at great length on the writing forum I frequent.

One of the more recent long running threads was started by me, not about whether to use AI but about whether to acknowledge it in one's work.

I choose to acknowledge my use of AI. I use it for SPaG (spelling, punctuation and grammar), but it also use it for research - essentially a fast, verbose search engine.

I make clear that I never let it generate any content and that every word on every page is my own.

I was quite surprised at how much pushback I got from other writers saying "why let the reader know - it's none of their business", which immediately makes me think they are using it more than readers would tolerate.

 

[symbolipoint]

DaveC,..., at least you care about your own ability to use your own language and thinking abilities. AI should not be a part of any of it, if you intend to be a normally functioning and communicative human.
I have a bit more to say but am short on time.

 

[BillTre]

As an aspiring writer, this is an issue that affects me. It is discussed at great length on the writing forum I frequent.

One of the more recent long running threads was started by me, not about whether to use AI but about whether to acknowledge it in one's work.

I choose to acknowledge my use of AI. I use it for SPaG (spelling, punctuation and grammar), but it also use it for research - essentially a fast, verbose search engine.

I make clear that I never let it generate any content and that every word on every page is my own.

I was quite surprised at how much pushback I got from other writers saying "why let the reader know - it's none of their business", which immediately makes me think they are using it more than readers would tolerate.

Your stance is a lot like Friedman's. Its also nice that it's a honest approach.

From what I've read (from other sources) there is a lot of pushback on AI. Seemingly for different reasons and in different domains. Among the writer's posts I have read are:

• Concerns about cover art (mostly based on stealing or taking jobs away from artists).
• Concerns about stealing of other's work by the AI. The companies are working around this by only training on materials they have permission to use for that purpose.
• Concerns about taking work away from writers.
• Concerns about poor quality writing (common).
• Concerns about deception.
• A general concern about it being an affront to human civilization.

 

[BillTre]

DaveC,..., at least you care about your own ability to use your own language and thinking abilities. AI should not be a part of any of it, if you intend to be a normally functioning and communicative human.
I have a bit more to say but am short on time.

In many ways it is merely a tool, like a spell checker.

 

[DaveC426913]

In many ways it is merely a tool, like a spell checker.

Exactly.

Virtually all technologies are neither good nor evil. The good and evil are in how they are used.


*Except AI. AI doesn't require humans to use it. It acts itself. And it can do evil.

 

[symbolipoint]

In many ways it is merely a tool, like a spell checker.

I might assume that is both informative and beneficial. Seems very limited, justifiably.

 

[BillTre]

Seems very limited, justifiably.

Letting your spell checker take over SAC sounds like a bad idea.

 

[symbolipoint]

Letting your spell checker take over SAC sounds like a bad idea.

What is "SAC"? Spell And Check?

 

[DaveC426913]

Letting your spell checker take over SAC sounds like a bad idea.

You never let any tool take over.

I specifically tell it to make suggestions only (and forbid it from making any edits). I review and change only what I find is warranted. All it's ever doing is drawing my attention things I missed on my tenth or twelfth review.

That is the only sane way to let a tool be involved in the writing process.

Lordy, that's not AI-specific. I would never let any spellchecker tell me to correct anything unless I knew it was wrong. (Here in Canada, we get used to that, since most spellcheckers default to Americanese).

 

[BillTre]

What is "SAC"? Spell And Check?

SAC = Strategic Air Command.
The part of the air force that runs big strategic bombers and intercontinental missiles. Could start a war.

 

[AlexB23]

Me personally, I am against AI art. If people absolutely want to use it, the art should be clearly labeled as AI generated, be able to be blocked from search results, and must not be monetized. There is a slight niche for AI art. People could take a photo of their room, then a color swatch for paint, and furniture and be able to visualize what a home remodeling might look like. Other than that, I am strongly against AI art.

 

[Hornbein]

With AI art everything looks like plastic, especially faces.