Can AI be used to produce art?

[nikos79]

edited: "polling and stitching" are Alan Edward Klein's words, not mine; did you meant to reply to him?

Yes sorry for that

If true, you seem to be making the case that AI can create?

Yes I do. But whether it is art or not maybe we need a new definition of art.

It would for sure demonstrate excellent craftmanship.

But art in its deepest existential form? I am not sure. For me art in a very deep way is strongly related to death. And if they cannot experience this agony and metaphysical feeling of passing of time then why create and how should we approach it?

 

[Cholentpot]

Yes sorry for that


Yes I do. But whether it is art or not maybe we need a new definition of art.

It would for sure demonstrate excellent craftmanship.

But art in its deepest existential form? I am not sure. For me art in a very deep way is strongly related to death. And if they cannot experience this agony and metaphysical feeling of passing of time then why create and how should we approach it?

I don't even know if its craftmanship. Craftmanship denotes a certain quality to the work.

Art speaks to you via death and loss, art speaks to me as effort and something gained. Both can co-exist.

We've been conditioned in the past 100+ so years that art can exist as an idea. Like a banana taped to a wall. The idea trumps the work itself. AI upends that I think. In the coming future people will very much want to see the effort behind implementing the idea itself.

 

[runswithsizzers]

There's no rust or wear on the hammer. Hammer is the wrong side, not proportionate. Handle is too long, head doesn't fit. Also, what tradesman holds a hammer like this. The AI looked at thousands of tintypes from the era and stuck a hammer in where a six-gun would have been. The hand placement in correct, everything else it got wrong.

All true. But he is a murderer, not a tradesman. Pehaps his regular hammer was left at the crime scene, and the tintype studio provided this one as a prop?

My point being, some images are created by human photographers to evoke a feeling or create a mood, and strict adherence to factual details is not always a requirement. The movies we love to watch are good examples of how we sometimes like to be fooled by the fantastical -- to suspend our belief and just enjoy the spectacle. Not every movie has to be made to Ken Burns standards, and not every photograph has to be accurate documentation to be successful.

Personally, as a viewer, I can enjoy the tintype of hammer man for what it is, without paying too much attention to the details.

Now if it had been me asking Nanobanana to create this image, I probably would have added more prompts to make the hammer bigger, or whatever. If the hammer is inappropriate for this image, is it AI's fault for not getting it right? Or is it Sean's fault for not asking the right prompts?

 

[runswithsizzers]

But whether it is art or not maybe we need a new definition of art.

This is the heart of the problem will all discussions about art -- how do you define it?

Here are a few questions about the definition of art I see presented in this thread:

Is it art if it wasn't created by humans?

Is it art if was created by AI guided by humans?

Is it art if anyone can do it? ("My third grader could have made that")

If art is harder to make, does that make it better than art which is easy to make?

Does art have to be original?

Can art be original?

Who gets to decide if a work of art is good or bad?

I almost forgot this important litmus test:
Is it art if there is no wine and cheese?

And one more, inspired by the next post by @koraks:
Is it art if you can't touch it (intangible)?

 

[koraks]

I'd point these out as having AI red flags if I ran across them in the wild. There's just something subtly off about the photos.

I agree.
6-12 months down the road, or perhaps even today with a little bit attention to the prompting, that would probably disappear and neither you or I would recognize these images as AI-made.

We are then left with the inconvenient realization that much of the photography we're used to isn't very original, at least as viewed on a computer screen. Which might feel a little uncomfortable at first, but I think in the end, it'll bring a couple of possibly more comforting ideas.

Firstly, as we all know on this forum, a computer screen is a poor facsimile of a real print - and there's still a degree of magic to seeing, fondling and admiring a real, physical print. Of course, there's no real print underlying @Sean's examples; they're visually pleasing images as such, but no tangible artifact underlies them.

Secondly, once we have acknowledge that the vast majority of the photography that's made (including by us and certainly by myself) isn't particularly original. The creative aspect and perhaps also the intrinsic value lies in the hands-on nature of the process, and/or the fact that we have shaped the end result on every step along the way. I.e. it's not just the end result that represents the value, it's also and to a large extent the road that leads up to it.

AI can't touch that.
It's a different question, though, from the one this thread started with.

Also, as to your example of the comic group you're part of - your complaint the way I see it is with this one guy who doesn't want to play by your rules. The fact that he uses AI is of lesser importance. Had he made a poor ripoff version with his own hands, or had he outsourced it to Pakistan, you might have felt quite the same. It's not the car's fault that the dog was run over. The driver had a lot to do with it.

 

[Cholentpot]

I agree.
6-12 months down the road, or perhaps even today with a little bit attention to the prompting, that would probably disappear and neither you or I would recognize these images as AI-made.

We are then left with the inconvenient realization that much of the photography we're used to isn't very original, at least as viewed on a computer screen. Which might feel a little uncomfortable at first, but I think in the end, it'll bring a couple of possibly more comforting ideas.

Firstly, as we all know on this forum, a computer screen is a poor facsimile of a real print - and there's still a degree of magic to seeing, fondling and admiring a real, physical print. Of course, there's no real print underlying @Sean's examples; they're visually pleasing images as such, but no tangible artifact underlies them.

Secondly, once we have acknowledge that the vast majority of the photography that's made (including by us and certainly by myself) isn't particularly original. The creative aspect and perhaps also the intrinsic value lies in the hands-on nature of the process, and/or the fact that we have shaped the end result on every step along the way. I.e. it's not just the end result that represents the value, it's also and to a large extent the road that leads up to it.

AI can't touch that.
It's a different question, though, from the one this thread started with.

Also, as to your example of the comic group you're part of - your complaint the way I see it is with this one guy who doesn't want to play by your rules. The fact that he uses AI is of lesser importance. Had he made a poor ripoff version with his own hands, or had he outsourced it to Pakistan, you might have felt quite the same. It's not the car's fault that the dog was run over. The driver had a lot to do with it.

On your first points I agree, most photography is not unique in itself other than it was made by a person. My photography is unique because I made it. It's me. The same way I enjoy my own veggies from my garden. I did it. The fruits of your labor and all that. I didn't write the song, I covered it badly but it's my effort that makes it valuable.

As for the second part. No it's not the AI's fault but its so much easier to rip off now and it's also possible to see outright how little effort the ripoff was. It's obvious that this rando is using a platform to make the comic look like it was drawn by Gary Larson, person doesn't avoid the redflags, they embrace them.

I think this is helping me get further to my ideals and the issue. It's not AI, it's lazy AI. It's slop. Well curated AI work that shows thought and effort from the author can be no different than a pen or brush or Tmax 100 exposed just right. The issue is the huge and growing mountain of slop that makes people just stop caring. Which may be the answer. In the world of slop a hand made PB&J might be king.

 

[Sean]

Can you share them with us?

These are two recent examples. I would not be surprised if there are many more. We could be in a situation where LLMs take a back seat or become one of many layers to something more (something we may not even understand "black box" issue).

https://www.integral.ai/agi

• Grounded Reasoning: Because it builds a "Universal Simulator," it doesn't just predict the next word; it simulates the outcome of an action before taking it.
• Self-Correction: It can identify when an action fails and update its own "world model" internally, which is a hallmark of general intelligence.

(VL_JEPA) https://arxiv.org/abs/2512.10942
"The release of VL-JEPA (Vision-Language Joint Embedding Predictive Architecture) by Meta AI represents a significant shift in the philosophy of AI development. While it is not "AGI in a box," many researchers—most notably Meta’s Chief AI Scientist Yann LeCun—believe it addresses the fundamental flaws of current Large Language Models (LLMs) that prevent them from reaching human-level intelligence."

AGI may require a convergence of emerging technologies

"It is unlikely that any single model—be it VL-JEPA or the Integral model—will be AGI on its own. Instead, the consensus in 2026 is that AGI will likely emerge from an integrative system:

1. VL-JEPA provides the visual/linguistic world model.
2. The Integral Model provides the ability to plan and act.
3. A Reasoning Engine (like OpenAI’s "o" series or DeepMind’s AlphaProof) provides the formal logic."

Then there are the things going on behind closed doors, with some of the big players hinting at "new science" capable ais in early 2026.

I'd also add I don't think we even need AGI for incredible advances. Have a look at what Google's Isomorphic Labs and AlphaProteo are doing. https://www.isomorphiclabs.com. Their AI designed proteins could solve countless diseases.

The next 2yrs are going to be super interesting.

 

[nikos79]

These are two recent examples. I would not be surprised if there are many more. We could be in a situation where LLMs take a back seat or become one of many layers to something more (something we may not even understand "black box" issue).

https://www.integral.ai/agi

• Grounded Reasoning: Because it builds a "Universal Simulator," it doesn't just predict the next word; it simulates the outcome of an action before taking it.
• Self-Correction: It can identify when an action fails and update its own "world model" internally, which is a hallmark of general intelligence.

(VL_JEPA) https://arxiv.org/abs/2512.10942
"The release of VL-JEPA (Vision-Language Joint Embedding Predictive Architecture) by Meta AI represents a significant shift in the philosophy of AI development. While it is not "AGI in a box," many researchers—most notably Meta’s Chief AI Scientist Yann LeCun—believe it addresses the fundamental flaws of current Large Language Models (LLMs) that prevent them from reaching human-level intelligence."

AGI may require a convergence of emerging technologies

"It is unlikely that any single model—be it VL-JEPA or the Integral model—will be AGI on its own. Instead, the consensus in 2026 is that AGI will likely emerge from an integrative system:

1. VL-JEPA provides the visual/linguistic world model.
2. The Integral Model provides the ability to plan and act.
3. A Reasoning Engine (like OpenAI’s "o" series or DeepMind’s AlphaProof) provides the formal logic."

Then there are the things going on behind closed doors, with some of the big players hinting at "new science" capable ais in early 2026.

I'd also add I don't think we even need AGI for incredible advances. Have a look at what Google's Isomorphic Labs and AlphaProteo are doing. https://www.isomorphiclabs.com. Their AI designed proteins could solve countless diseases.

The next 2yrs are going to be super interesting.

I will study them carefully since the topic really interests me. As a remark though since I have been in the AI field for some time take with a grain salt what these big companies are saying. Usually they only care about their stock value so they publish ridiculous claims.

For me the road to AGI is really far. We would need total “immersion” a system that could be “embedded” in the world real time and constantly learn from all kind of interactions with it, similar to a newborn

 

[Alan Edward Klein]

I think one could argue that the process you describe is pretty much how many human artists work as well.

There is a saying, the gist of which goes back many years, which says, “Good artists copy. Great artists steal.”

Creativity and innovation are rare in humans, nonexistent in AI.

 

[Sean]

I will study them carefully since the topic really interests me. As a remark though since I have been in the AI field for some time take with a grain salt what these big companies are saying. Usually they only care about their stock value so they publish ridiculous claims.

For me the road to AGI is really far. We would need total “immersion” a system that could be “embedded” in the world real time and constantly learn from all kind of interactions with it, similar to a newborn

Yes, as they say proof is in the pudding. I merely have a hobbyist/futurist/sci-fi like interest in all of this (far from a scientist). I think humans are 20-30yrs away from building AGI, but AI building it is possibly 6 months to 2 years out. My gut feeling is that something will happen with recursive improvement that quickly escalates. Meta claims to have achieved self improving ai already, but let's see it..

 

[Alan Edward Klein]

Well I wouldn’t be sure Alan. They have recently solved mathematical problems humans have struggled for ages.

They don’t really copy but combine ideas, patterns, identifying connections in the most novel ways. Isn’t that what creativity is also about?

As a computer scientist I don’t think AGI is even close yet we already have reached some of their limitations. How many more data can you give them? You already fed them the goddamn internet.

So AI scientists are in the search of novel algorithms and causal representations to move AI research further. But at this point I wouldn’t even worry

Sort of like a dumb kaleidoscope. Where's the innovation?

 

[Chuck1]

NO

Hell no

 

[Alan Edward Klein]

I will study them carefully since the topic really interests me. As a remark though since I have been in the AI field for some time take with a grain salt what these big companies are saying. Usually they only care about their stock value so they publish ridiculous claims.

For me the road to AGI is really far. We would need total “immersion” a system that could be “embedded” in the world real time and constantly learn from all kind of interactions with it, similar to a newborn

My four-year old grandson already laughs at jokes, feels sad at times, all because he has a sense of himself, has an ego. He has feelings. He is a moral being with a heart. The learning process for humans is not only about adding 2+2 or figuring out data and displaying it in some form. But also assessing whether the answer revolves around issues of humaneness, moral clarity, and provides real inspiration. Otherwise, a human would be like a computer, a high-speed idiot processing ones and zeros very quickly, but with no contemplation of the world. Computers, and AI are just extensions of those; they are just that. They're great at processing math and huge amounts of data quickly, better than humans. But they don't know what to do with it unless we program the analysis. It's like looking at your monitor screen and claiming how smart our computers are, that they can show us all this stuff as they do. Does anyone think their TV and monitors are smart, much less innovative?

 

[Alan Edward Klein]

All true. But he is a murderer, not a tradesman. Pehaps his regular hammer was left at the crime scene, and the tintype studio provided this one as a prop?

My point being, some images are created by human photographers to evoke a feeling or create a mood, and strict adherence to factual details is not always a requirement. The movies we love to watch are good examples of how we sometimes like to be fooled by the fantastical -- to suspend our belief and just enjoy the spectacle. Not every movie has to be made to Ken Burns standards, and not every photograph has to be accurate documentation to be successful.

Personally, as a viewer, I can enjoy the tintype of hammer man for what it is, without paying too much attention to the details.

Now if it had been me asking Nanobanana to create this image, I probably would have added more prompts to make the hammer bigger, or whatever. If the hammer is inappropriate for this image, is it AI's fault for not getting it right? Or is it Sean's fault for not asking the right prompts?

But that raises a question. If the AI program can't tell the difference between a knife and a hammer, or when one or the other is appropriate, how can you expect it to be innovative?

Regarding the second point about the correct prompts, if it requires the operator to ask the correct prompts to get a result that makes sense or maybe even be innovative, then it's the operator who's being innovative, not the AI program. AI is no more than a paintbrush. But we have to point the way to creativity.

 

[nikos79]

My four-year old grandson already laughs at jokes, feels sad at times, all because he has a sense of himself, has an ego. He has feelings. He is a moral being with a heart. The learning process for humans is not only about adding 2+2 or figuring out data and displaying it in some form. But also assessing whether the answer revolves around issues of humaneness, moral clarity, and provides real inspiration. Otherwise, a human would be like a computer, a high-speed idiot processing ones and zeros very quickly, but with no contemplation of the world. Computers, and AI are just extensions of those; they are just that. They're great at processing math and huge amounts of data quickly, better than humans. But they don't know what to do with it unless we program the analysis. It's like looking at your monitor screen and claiming how smart our computers are, that they can show us all this stuff as they do. Does anyone think their TV and monitors are smart, much less innovative?

You are coming to the core of the problem that our human brains are "optimised" for this world we are living in and for our species. I wouldn't be as optimistic as Sean for the timeframe for AGI but I would call it a different kind of intelligence that we could reach in perhaps 2-4 years. When AI can diagnose X-Rays better than any radiologist, produce jaw-dropping moves in chess of sheer creativity and solve 100 year old math problem there is definitely some sort of intelligence there. But very specific to the tasks so we cannot call it AGI

 

[nikos79]

But that raises a question. If the AI program can't tell the difference between a knife and a hammer, or when one or the other is appropriate, how can you expect it to be innovative?

Regarding the second point about the correct prompts, if it requires the operator to ask the correct prompts to get a result that makes sense or maybe even be innovative, then it's the operator who's being innovative, not the AI program. AI is no more than a paintbrush. But we have to point the way to creativity.

Alan, if we set aside any philosophical, religious, or metaphysical considerations, which I’ve deliberately avoided here, there are today AI researchers who believe that some behaviors observed in large scale AI models resemble processes associated with human sentience!

I’m not presenting this as a fact, but rather to show that the skepticism is genuine, and that many experts believe we may be approaching something truly groundbreaking.

 

[Alan Edward Klein]

I agree.
6-12 months down the road, or perhaps even today with a little bit attention to the prompting, that would probably disappear and neither you or I would recognize these images as AI-made.

We are then left with the inconvenient realization that much of the photography we're used to isn't very original, at least as viewed on a computer screen. Which might feel a little uncomfortable at first, but I think in the end, it'll bring a couple of possibly more comforting ideas.

Firstly, as we all know on this forum, a computer screen is a poor facsimile of a real print - and there's still a degree of magic to seeing, fondling and admiring a real, physical print. Of course, there's no real print underlying @Sean's examples; they're visually pleasing images as such, but no tangible artifact underlies them.

Secondly, once we have acknowledge that the vast majority of the photography that's made (including by us and certainly by myself) isn't particularly original. The creative aspect and perhaps also the intrinsic value lies in the hands-on nature of the process, and/or the fact that we have shaped the end result on every step along the way. I.e. it's not just the end result that represents the value, it's also and to a large extent the road that leads up to it.

AI can't touch that.
It's a different question, though, from the one this thread started with.

Also, as to your example of the comic group you're part of - your complaint the way I see it is with this one guy who doesn't want to play by your rules. The fact that he uses AI is of lesser importance. Had he made a poor ripoff version with his own hands, or had he outsourced it to Pakistan, you might have felt quite the same. It's not the car's fault that the dog was run over. The driver had a lot to do with it.

So it's the human who's creative, not AI. Without the correct human prompts, AI is lost and creates nonsense. Just like you need a human hand for a paintbrush to make a painting.

 

[Alan Edward Klein]

Alan, if we set aside any philosophical, religious, or metaphysical considerations, which I’ve deliberately avoided here, there are today AI researchers who believe that some behaviors observed in large scale AI models resemble processes associated with human sentience!

I’m not presenting this as a fact, but rather to show that the skepticism is genuine, and that many experts believe we may be approaching something truly groundbreaking.

I think AI is being overhyped, either out of greed by corporations and investors or out of the exciting belief we've discovered a new god, or at least a new superhero. I think we're going to find out its just another tool to be used by humans to allow us to work quicker and more effectively, creating things as an extension of our own minds.

 

[Cholentpot]

I think AI is being overhyped, either out of greed by corporations and investors or out of the exciting belief we've discovered a new god, or at least a new superhero. I think we're going to find out its just another tool to be used by humans to allow us to work quicker and more effectively, creating things as an extension of our own minds.

It's the current dot.com bubble. They're shoving it into everything and always. No I don't need AI in my fridge.

 

[VinceInMT]

Most of my artistic output over the last year havsbeen my drawings rather than my photography. I did complete a series of cyanotypes and one, along with a drawing, were recently accepted into a contemporary art museum's annual auction. That said, I've mostly been concentrating on abstract drawings but had a thing/image in my mind of an object that I wanted to draw. My abstracts come from my imaginations but other work is assisted with reference material, usually my own photographs. I went through my photos looking for a suitable reference for this idea and didn't have one. I did a Google search and couldn't find one. Then I went to AI and described what I was looking for and it produced an image. I put that image in a search and couldn't find anything that it might have copied it from.

I ended up using the AI image as reference, but I still completed the drawing myself including lots of little details not in the AI reference. I have some mixed feelings about what I did. However, I frequently use things for reference that I didn't create be they a building, an object, etc. The resulting drawing of mine definitely shows the "artist's hand" something that even a photograph might not necessary have. I am leaning toward AI being just a tool.

 

[Alan Edward Klein]

Could you show us any of that work at this point?

 

[VinceInMT]

Could you show us any of that work at this point?

15”x22”, graphite on paper

 

[Sean]

So it's the human who's creative, not AI. Without the correct human prompts, AI is lost and creates nonsense. Just like you need a human hand for a paintbrush to make a painting.

I don't think most grasp that this is merely temporary. AI as it is now is the worst it will ever be. AI as it is now is not the final iteration or a framework set in stone, it is evolving by the hour. It's like I've posted before on how some view AI...

"Well it still can't do A, B or C"
"Ok, now it can do A, B, and C, but it will never do D, and E"
"Ok wow, it is doing D and E, but it'll be 100 yrs before it even gets close to F, G and H"
"What the heck? It's doing F, G and H? But it'll never achieve I, J agh ok forget it!"

This is what I've seen in the past 3 yrs and it is accelerating in capability. I have been working with Gemini 3 Pro to do a range of server and code improvements and my mind is so blown I don't know how to explain how good this thing is. Just a year ago, trying to get ai to help me with server configurations and xenforo code work was an exercise in frustration, it just couldn't understand what I wanted. However, this new Gemini 3 Pro truly seems to understand exactly what I want and provides me perfect instructions, code and more to do whatever I ask it to.

There is also the issue of people using the lesser free ai models or the models that are not ideal for what they need to accomplish, then passing judgement. Google seemed way behind so I was stunned that Gemini 3 Pro was so insanely good (almost seeming like AGI at times). So it is already at an incredible state and I can't wrap my head around what Gemini 4 or 5 would be like as the capability is increasing by orders of magnitude.

 

[runswithsizzers]

But that raises a question. If the AI program can't tell the difference between a knife and a hammer, or when one or the other is appropriate, how can you expect it to be innovative?

What makes you think AI cannot tell the difference between a knife and a hammer? I'm pretty sure if Sean had asked for a knife or a hammer in the prompt, he would have got what he asked for.

If I understand correctly, Sean was giving examples of what AI comes up with if given only minimal prompts(?) The prompt for the tintype was simply, "Do a tintype" - which suggests a style, but it is very nonspecific about subject matter.

Many historical tintypes were portraits, so the subject of the AI image is not surprising. But the fact that a hammer appears in this AI image -- without being prompted -- is interesting! If I didn't know better, I might think the AI was having a little fun with us. ;-)

 

[Cholentpot]

What makes you think AI cannot tell the difference between a knife and a hammer? I'm pretty sure if Sean had asked for a knife or a hammer in the prompt, he would have got what he asked for.

If I understand correctly, Sean was giving examples of what AI comes up with if given only minimal prompts(?) The prompt for the tintype was simply, "Do a tintype" - which suggests a style, but it is very nonspecific about subject matter.

Many historical tintypes were portraits, so the subject of the AI image is not surprising. But the fact that a hammer appears in this AI image -- without being prompted -- is interesting! If I didn't know better, I might think the AI was having a little fun with us. ;-)

There's a hammer because I suspect part of the coding of free publicly available AI is to keep away from violent imagery.

If you look up portraits from that era of people holding hammers they're never held in that position. The position that AI gave was popular for settlers or soldiers holding their most prized possession, a six gun. AI won't put in a gun unprompted so it put a general hammer in, which I might add is not an era correct hammer.