Can AI be used to produce art?

[koraks]

does putting a prompt into GPT and sitting back count as art?

How about standing in a field, pressing the button on a box with a bit of glass attached to the front?
Lifting one end of a tray in a room with a dim red light, then allow it to drop again?
Dragging colorful squares from one end to another in a computer program like Lightroom?
Of course the act of prompting isn't art. That's not the point. But this confusion is at the core of many of the end-all rejections of the potential use of AI for the production of 'good' art (whatever the heck 'good' means here). It ultimately boils down to some form of "you're doing it wrong." Which seems entirely sensible, until some guy or gal pops up who does it "all wrong" and your jaw drops to the floor.

AI is not the same as crafting something in Photoshop.

In some ways it is, in some ways it isn't.

If you're interested, I'd like to invite you to read some primers on how generative Ai works, and in particular the conceptual underpinnings of these models. There seem to be some misconceptions about this in most of your arguments, making them problematic to comment on.

"Good paintings cannot be made with acrylic paints and synthetic brushes. I hate acrylic paints and synthetic brushes unless they're used to paint a door or a wall."
I think much of the problem is that you've not seen or recognized the possibility of an exception to what you perceive as an axiomatic rule. The same happened when photography came around, when CGI first popped up and in plenty of other cases where a new tech emerged, and the first output poured in and it turned out to be lackluster. Da Vinci wasn't the first guy to break a couple of eggs and paint a ceiling with them. Plenty of "meh" ceilings were painted before he performed his miracles.

One of these days, we're going to see art pop up and one day (perhaps soon after, maybe much later) many people will realize it's damn good art - and generative Ai is at the core of producing it. It's a powerful tool and with 8 billion of us running around, there's plenty of talent to exploit it to magnificent ends. The lack of recognition of what capable hands and minds can do with Ai are limited by the same lack of imagination that currently produces the 'slop'. The real challenge at this point of course is to weed out the chaff. But perhaps AI can help in that.

Maybe part of the problem is that AI is so deceptively simple. In that sense, it's a bit like the camera on a modern phone. Point it at something and you walk away with a technically good photo. But how many of those snapshots are 'good art'? A fairly quick session of prompting by a random person can produce technically 'correct' images - but they'll for the most part be 'slop'. Turns out that just like a decent camera doesn't make one an artist, generative AI cannot do the same thing. That, on the other hand, doesn't mean it's inherently impossible.

I feel the core of much of the criticism of many relates to another point you've made, about AI being a potential threat. Many of us are frightened and/or have rapidly grown a dislike for it. It's understandable, and in many cases also justifiable. However, at the risk of gross oversimplification: if something is 90% problematic and 10% promising, you can't just disregard the 10% because of the 90% (or whatever ratio you want to stick to it). Yet, that's what happens. People dislike it, so they disqualify it categorically. It's human nature.

 

[Sean]

Will there come questions of honesty and authenticity (how much fraud)? Here is an ai artwork from my chatgpt, the only prompt to the ai was, "create a work of art with no human involvement". Zero creative input from me, it spat this out in 20 seconds:

I then told the ai, I want people to think I created it, so provide a description for them, I mean.... after all this took me 1,000hrs of painstaking digital illustration (...right???):

----
"After nearly 2 months of illustrating, I'm pleased to show everyone my latest piece. It explores a moment of convergence, where opposing forces collide and radiate outward. The composition is structured around a central ignition point, using color and texture to suggest energy, pressure, and movement.

Warm tones concentrate at the core to imply intensity and emergence, while cooler hues disperse toward the edges, creating a sense of expansion and flow. Layered textures and branching forms reference both organic growth and geological fracture.

The piece is intended to feel dynamic rather than static, inviting the viewer into a field of motion and transformation rather than a fixed image. I hope you all enjoy it!"
----

"OMG that is amazing!"
"I can't believe you spent so long constructing that, truly a marvel!!"
"Where can I order a print?"

Now, I'm sure people will use AI as a tool and be authentic about it, but there is the other dark side to it that will surely have a big impact on the authentic. With something so powerful, how do we know who is and who isn't being genuine? Does it matter? Will it matter? Is there a threshold for authenticity? What would that be? 95% human driven? Who could even prove it? I am expecting no one will care, or give up trying to care. I already know people listening to random ai generated endless music streams and they could not care less if zero human artistry involved.

I could easily get ai to create a random art image set where I have zero creative involvement. Let's say 25 random images in a cohesive art series/style, print at ultra high quality 150" prints, hang in a show and done. Who would know? At that point is merely asking for a random set of images, printing it and hanging it = art? To what extent would such mass produced "slop" eventually eradicate all value in human created art? The "I can tell it's ai" crowd is exponentially shrinking as it improves by the week. Right now is the worst it will ever be. There is no way anyone will be able to detect ai work very soon if not already. A blind test would probably already fool over 90% in many cases and this will be 100% soon.

AI music might be even more painful. I could see some less than capable bands randomly generating full songs in ai, then simply covering the ai songs and releasing those songs as their own.

All this makes me wonder how visual arts, music, etc becomes viewed in the near future when all levels of source or authenticity are unknown, and there are billions of such works flooding the internet every single day.

My answer to this as an artist is to do what I want and love to do, and share in that with others in the same boat. I've never had a desire to sell work, compete or obtain any notoriety. I imagine that aspect of art will quickly evaporate in the next 10yrs as it is lost in a tsunami of content

 

[Hassasin]

NO

 

[Cholentpot]

Will there come questions of honesty and authenticity (how much fraud)? Here is an ai artwork from my chatgpt, the only prompt to the ai was, "create a work of art with no human involvement". Zero creative input from me, it spat this out in 20 seconds:

View attachment 414811

I then told the ai, I want people to think I created it, so provide a description for them, I mean.... after all this took me 1,000hrs of painstaking digital illustration (...right???):

----
"After nearly 2 months of illustrating, I'm pleased to show everyone my latest piece. It explores a moment of convergence, where opposing forces collide and radiate outward. The composition is structured around a central ignition point, using color and texture to suggest energy, pressure, and movement.

Warm tones concentrate at the core to imply intensity and emergence, while cooler hues disperse toward the edges, creating a sense of expansion and flow. Layered textures and branching forms reference both organic growth and geological fracture.

The piece is intended to feel dynamic rather than static, inviting the viewer into a field of motion and transformation rather than a fixed image. I hope you all enjoy it!"
----

"OMG that is amazing!"
"I can't believe you spent so long constructing that, truly a marvel!!"
"Where can I order a print?"

Now, I'm sure people will use AI as a tool and be authentic about it, but there is the other dark side to it that will surely have a big impact on the authentic. With something so powerful, how do we know who is and who isn't being genuine? Does it matter? Will it matter? Is there a threshold for authenticity? What would that be? 95% human driven? Who could even prove it? I am expecting no one will care, or give up trying to care. I already know people listening to random ai generated endless music streams and they could not care less if zero human artistry involved.

I could easily get ai to create a random art image set where I have zero creative involvement. Let's say 25 random images in a cohesive art series/style, print at ultra high quality 150" prints, hang in a show and done. Who would know? At that point is merely asking for a random set of images, printing it and hanging it = art? To what extent would such mass produced "slop" eventually eradicate all value in human created art? The "I can tell it's ai" crowd is exponentially shrinking as it improves by the week. Right now is the worst it will ever be. There is no way anyone will be able to detect ai work very soon if not already. A blind test would probably already fool over 90% in many cases and this will be 100% soon.

AI music might be even more painful. I could see some less than capable bands randomly generating full songs in ai, then simply covering the ai songs and releasing those songs as their own.

All this makes me wonder how visual arts, music, etc becomes viewed in the near future when all levels of source or authenticity are unknown, and there are billions of such works flooding the internet every single day.

My answer to this as an artist is to do what I want and love to do, and share in that with others in the same boat. I've never had a desire to sell work, compete or obtain any notoriety. I imagine that aspect of art will quickly evaporate in the next 10yrs as it is lost in a tsunami of content

I don't think all is lost, I think you're going to see a pivot back to classic styles that require human input. Back to music being made without obvious digital processing such as auto tune, AI can't sing off tune for instance. We're going to see random human touches that we've been trying to eradicate come back into style. Racked focus missing in movies, imperfect subjects in photos. Creative work will be passed around by creatives and the vast majority of people who don't care will enjoy their velvet Elvis's.

The growth of film, back to physical audio and kids wrenching on old cars is a good sign. Go onto ebay and check the price of a 4 track cassette studio recorder. People want that again.

How about standing in a field, pressing the button on a box with a bit of glass attached to the front?
Lifting one end of a tray in a room with a dim red light, then allow it to drop again?
Dragging colorful squares from one end to another in a computer program like Lightroom?
Of course the act of prompting isn't art. That's not the point. But this confusion is at the core of many of the end-all rejections of the potential use of AI for the production of 'good' art (whatever the heck 'good' means here). It ultimately boils down to some form of "you're doing it wrong." Which seems entirely sensible, until some guy or gal pops up who does it "all wrong" and your jaw drops to the floor.



In some ways it is, in some ways it isn't.

If you're interested, I'd like to invite you to read some primers on how generative Ai works, and in particular the conceptual underpinnings of these models. There seem to be some misconceptions about this in most of your arguments, making them problematic to comment on.

"Good paintings cannot be made with acrylic paints and synthetic brushes. I hate acrylic paints and synthetic brushes unless they're used to paint a door or a wall."
I think much of the problem is that you've not seen or recognized the possibility of an exception to what you perceive as an axiomatic rule. The same happened when photography came around, when CGI first popped up and in plenty of other cases where a new tech emerged, and the first output poured in and it turned out to be lackluster. Da Vinci wasn't the first guy to break a couple of eggs and paint a ceiling with them. Plenty of "meh" ceilings were painted before he performed his miracles.

One of these days, we're going to see art pop up and one day (perhaps soon after, maybe much later) many people will realize it's damn good art - and generative Ai is at the core of producing it. It's a powerful tool and with 8 billion of us running around, there's plenty of talent to exploit it to magnificent ends. The lack of recognition of what capable hands and minds can do with Ai are limited by the same lack of imagination that currently produces the 'slop'. The real challenge at this point of course is to weed out the chaff. But perhaps AI can help in that.

Maybe part of the problem is that AI is so deceptively simple. In that sense, it's a bit like the camera on a modern phone. Point it at something and you walk away with a technically good photo. But how many of those snapshots are 'good art'? A fairly quick session of prompting by a random person can produce technically 'correct' images - but they'll for the most part be 'slop'. Turns out that just like a decent camera doesn't make one an artist, generative AI cannot do the same thing. That, on the other hand, doesn't mean it's inherently impossible.

I feel the core of much of the criticism of many relates to another point you've made, about AI being a potential threat. Many of us are frightened and/or have rapidly grown a dislike for it. It's understandable, and in many cases also justifiable. However, at the risk of gross oversimplification: if something is 90% problematic and 10% promising, you can't just disregard the 10% because of the 90% (or whatever ratio you want to stick to it). Yet, that's what happens. People dislike it, so they disqualify it categorically. It's human nature.

I'm not arguing against using AI as a tool. I use it all the time. It's a vast and powerful thing. The issue is I'm inundated with slop almost constantly now. I can give an example, I'm part of a group that brainstorms comics. We have one talented artist who takes our ideas, sketches them out and publishes them. They're very niche but it serves a crowd of like minded people. We had one person in that group who would take the punchlines and feed it to AI and about an hour after we'd publish this guy would publish an sloppy AI version of the same punchline and it would make the rounds. It's less about art and acrylics than cheapening creativity. There's nothing to do to stop the tidal wave of slop. I watched this happen with Auto Tune in music and we lost something very special. People that grew up post autotune can't even hear it, let alone know what they've lost.

 

[Alan Edward Klein]

If you entered your works into a prestigious gallery and got bumped by AI slop how would that make you feel? What if it derailed your entire career? What if you now have to compete against data farms powered by nuclear plants harvesting the work of billions with no permission or copyright conditions? Or better you, you've established your work and specific vision and are making traction and you get derailed by someone using your specific body of work via AI?

Does 'Let there be Art!' still sound like a good thing?

Lately, on YouTube, I;ve seen complete blogs, newly created by AI that are fakes. They show real people from other real blogs speaking words they never said. I can't believe this is legal. It's to the point that I don;t know if I;m watching the real person or their double. I;m losing interest in even watching these blogs which is a signal to what's going to happen to photography as AI starts dumping their photos out to the public.

 

[Alan Edward Klein]

If you don't mind me asking, how are you involved with AI? Have you spent much time using prompts to generate AI images?

I have generated only a few AI images, maybe a dozen, so I am only very briefly familiar with the process. But it seems to me, creating AI images can involve a certain amount of creativity and skill.

First, one must conceive the image. It requires a certain amount of imagination to come up with an idea -- a concept -- and then create a mental image of that concept. Someone with a vivid imagination will "see" a lot of details -- what particular objects will be in the image, their colors and textures, composition (how the objects should be arranged), framing, and lighting. This part of the process would be no different from a painter visualizing the concept for a new painting, and it is inherently a creative process.

Next, one must come up with the right prompts which lead to a generative image which matches the image in your head. In my experience (very limited), it can take some time and multiple additional promts before an acceptable result is obtained. Each modified prompt creates a different version of the scene; sometimes the changes are favorable towards your goal, sometimes not. I am guessing that with practice, more carefully chosen prompts will get better results?

Certainly, the process of creating an AI generated image is quicker and easier than making a photograph or a painting. Maybe it is not art. Maybe it's more of a craft? I don't know. But I do think some creativity is required to make an AI image, and I do think it is possible to make AI images which are compelling, enjoyable, and/or thought provoking in ways that are similar to other visual arts.

As you explain it, I think you're right about how a person can be creative using AI. At the end of the day, however, it's going to marginalize photography as a hobby. Primarily, it will be used by people taking family and vacation shots, as well as capturing other real-time moments of their lives. Artistic photography will be side-tracked to a minor role as film photography is today. I can also imagine AI creating oil paintings with a device that applies paint in selected strokes based on your ideas. I wonder what painters are going to say when this starts happening? How about sculpting?

PS Grammarly AI app just did a nice job updating this post to something more readable.

 

[Alan Edward Klein]

Will there come questions of honesty and authenticity (how much fraud)? Here is an ai artwork from my chatgpt, the only prompt to the ai was, "create a work of art with no human involvement". Zero creative input from me, it spat this out in 20 seconds:

View attachment 414811

I then told the ai, I want people to think I created it, so provide a description for them, I mean.... after all this took me 1,000hrs of painstaking digital illustration (...right???):

----
"After nearly 2 months of illustrating, I'm pleased to show everyone my latest piece. It explores a moment of convergence, where opposing forces collide and radiate outward. The composition is structured around a central ignition point, using color and texture to suggest energy, pressure, and movement.

Warm tones concentrate at the core to imply intensity and emergence, while cooler hues disperse toward the edges, creating a sense of expansion and flow. Layered textures and branching forms reference both organic growth and geological fracture.

The piece is intended to feel dynamic rather than static, inviting the viewer into a field of motion and transformation rather than a fixed image. I hope you all enjoy it!"
----

"OMG that is amazing!"
"I can't believe you spent so long constructing that, truly a marvel!!"
"Where can I order a print?"

Now, I'm sure people will use AI as a tool and be authentic about it, but there is the other dark side to it that will surely have a big impact on the authentic. With something so powerful, how do we know who is and who isn't being genuine? Does it matter? Will it matter? Is there a threshold for authenticity? What would that be? 95% human driven? Who could even prove it? I am expecting no one will care, or give up trying to care. I already know people listening to random ai generated endless music streams and they could not care less if zero human artistry involved.

I could easily get ai to create a random art image set where I have zero creative involvement. Let's say 25 random images in a cohesive art series/style, print at ultra high quality 150" prints, hang in a show and done. Who would know? At that point is merely asking for a random set of images, printing it and hanging it = art? To what extent would such mass produced "slop" eventually eradicate all value in human created art? The "I can tell it's ai" crowd is exponentially shrinking as it improves by the week. Right now is the worst it will ever be. There is no way anyone will be able to detect ai work very soon if not already. A blind test would probably already fool over 90% in many cases and this will be 100% soon.

AI music might be even more painful. I could see some less than capable bands randomly generating full songs in ai, then simply covering the ai songs and releasing those songs as their own.

All this makes me wonder how visual arts, music, etc becomes viewed in the near future when all levels of source or authenticity are unknown, and there are billions of such works flooding the internet every single day.

My answer to this as an artist is to do what I want and love to do, and share in that with others in the same boat. I've never had a desire to sell work, compete or obtain any notoriety. I imagine that aspect of art will quickly evaporate in the next 10yrs as it is lost in a tsunami of content

What you did is similar to Photoshop Elements when it first came out 25 years ago. It's still available on my 2020 version. Under the Filter tab, you can select various artistic forms, such as oil, or sketching, or charcoal, and then apply different strokes, converting one of your photos to "art". I was excited about it initially and even printed a couple of "artistic" results. But then it quickly got boring. Seen that; done that. I suspect that after people do a couple of AI photos, they too will be bored and move on. The real use for AI will be by professionals who need to create images for commercial use quickly and cheaply. The rest of us will continue to snapshot photos of our children and ourselves with a camera.

 

[Sean]

Here is some playing with Google's new nanobanana pro.

prompt:

"create a fine art photograph, it will need to be shot with a large format view camera, contact printed to platinum palladium, the landscape should be death valley, you decide on the creative aspects"

"do another variation, maybe from the desert floor":

"Do a tintype":

 

[Alan Edward Klein]

I;m curious about how the AI program selects what to do. Obviously, AI is not a sentient being. It doesn;t really think on it;s own. So there has to be a program behind it written by some actual human being. How does that program work? How does it select the features? IS it polling the entire web for ideas and then selecting and integrating what applies? Anyone know?

 

[Sean]

I;m curious about how the AI program selects what to do. Obviously, AI is not a sentient being. It doesn;t really think on it;s own. So there has to be a program behind it written by some actual human being. How does that program work? How does it select the features? IS it polling the entire web for ideas and then selecting and integrating what applies? Anyone know?

I asked it:

--------

You are spot on that there is no "thinking" or sentience involved. The process is much more like a very complex, high-speed mathematical prediction than an act of creative will.

To answer your specific questions: No, the AI is not polling the web in real-time. It isn't going out to Google Images or Wikipedia to look up "tintype" or "Death Valley" the moment you ask.

Here is a look at the actual mechanics of how the program "selects" features and generates the image.

1. The "Library" in the Box (Training)​

Imagine if a human spent 1,000 years sitting in a room looking at billions of photographs. Every time they saw a photo, a teacher told them what it was: "This is a tintype," "This is a platinum print," "This is Death Valley," "This is a large format view camera."

Eventually, that human wouldn't need to look up those terms anymore. If you said "draw a tintype," they would rely on their memory of the patterns that make up a tintype (grainy texture, metal edges, high contrast, chemical swirls).

• The Model: The "program" you are talking about is a Neural Network. It has already "read" the internet (during its training phase). That training is now finished, and the knowledge is compressed into a file (the model).
• Frozen Knowledge: When you ask a question, the AI accesses this frozen internal model, not the live internet.

2. How it "Selects" Features (Text-to-Image)​

When you gave the prompt "shot with a large format view camera, contact printed to platinum palladium," the AI broke that sentence down into mathematical vectors (lists of numbers).

• Association: It knows from its training that the words "platinum palladium" are mathematically associated with specific visual traits: warm tones, soft focus, broad tonal range, and matte paper texture.
• Diffusion: To create the image, the AI starts with a canvas of pure static (random digital noise). It then slowly removes the noise, step-by-step, refining the chaos into an image that matches the mathematical "address" of your prompt.
• The Selection: It didn't consciously "decide" to put a hammer in the man's hand or make his beard grey. It selected those features because, statistically, in the billions of images it studied, 19th-century laborers in tintypes often had tools, beards, and rugged clothing. It is playing the odds of what matches the pattern "authentic tintype."

3. The Text Response (LLM)​

When I (the language model) answer you, a similar process happens. I am not thinking. I am predicting the next word.

• Probability: If you type "The cat sat on the...", my internal statistics know that "mat" is 90% likely, "floor" is 5% likely, and "sandwich" is 0.001% likely.
• Context: Because you asked about "authenticity" and "indistinguishable" traits in your prompt, I biased my predictions toward words that sound professional, artistic, and historically accurate.

Summary​

There is no ghost in the machine. It is a predictive engine.

1. Humans wrote the learning algorithms (the code that tells the AI how to learn).
2. Humans curated the dataset (the billions of images and text).
3. The AI found the patterns in that data.
4. You provided the trigger (the prompt), and the AI completed the pattern.
   
   
   
5. 
   
   Getty Images

Does that explanation help clarify the "magic" behind the curtain? Would you like to try "breaking" the pattern by asking for something that doesn't statistically belong in a tintype, like a modern smartphone, to see how it handles the conflict?

------

 

[Cholentpot]

As you explain it, I think you're right about how a person can be creative using AI. At the end of the day, however, it's going to marginalize photography as a hobby. Primarily, it will be used by people taking family and vacation shots, as well as capturing other real-time moments of their lives. Artistic photography will be side-tracked to a minor role as film photography is today. I can also imagine AI creating oil paintings with a device that applies paint in selected strokes based on your ideas. I wonder what painters are going to say when this starts happening? How about sculpting?

PS Grammarly AI app just did a nice job updating this post to something more readable.

I agree with this. Photography will be left to capturing moments rather than for creative artistic uses for most people.

Here is some playing with Google's new nanobanana pro.

prompt:

"create a fine art photograph, it will need to be shot with a large format view camera, contact printed to platinum palladium, the landscape should be death valley, you decide on the creative aspects"

View attachment 414818

"do another variation, maybe from the desert floor":

View attachment 414819

"Do a tintype":

View attachment 414821

I'm not saying I'm an expert but 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. Be it the lighting, the depth of field or perfect sharpness. The tintype messed up by sticking a modern hammer in there. And the tintype got the depth of field wrong. But to a causal viewer everything looks fine.

 

[Alan Edward Klein]

Thanks for that explanation Sean. So basically, it's not thinking out of the box. So it can't be really innovative. It's a copy expert. It's basically polling what others have done, looking at huge data fields of original thinking and work, selecting favored probabilities, and spitting out the same stuff rehashed so you don;t recognize the original parts from the new whole. It;s not really imaginative or intuitive. New ideas will escape it.

It seems we humans still have a chance.

 

[Sean]

Thanks for that explanation Sean. So basically, it's not thinking out of the box. So it can't be really innovative. It's a copy expert. It's basically polling what others have done, looking at huge data fields of original thinking and work, selecting favored probabilities, and spitting out the same stuff rehashed so you don;t recognize the original parts from the new whole. It;s not really imaginative or intuitive. New ideas will escape it.

It seems we humans still have a chance.

Limitation of LLMs, but LLMs are not the end game of AI nor should be considered the future of AI. I'm already seeing 2 non LLM AI systems that might send LLMs to the junk pile and lead to AGI, full creativity and new science.

 

[Sean]

I agree with this. Photography will be left to capturing moments rather than for creative artistic uses for most people.



I'm not saying I'm an expert but 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. Be it the lighting, the depth of field or perfect sharpness. The tintype messed up by sticking a modern hammer in there. And the tintype got the depth of field wrong. But to a causal viewer everything looks fine.

Oh definitely, but I suspect in another year these red flags will evaporate. I think the hardest ones to tell right now are product based photography images.

 

[Alan Edward Klein]

Limitation of LLMs, but LLMs are not the end game of AI nor should be considered the future of AI. I'm already seeing 2 non LLM AI systems that might send LLMs to the junk pile and lead to AGI, full creativity and new science.

What are LLM and AGI?

 

[nikos79]

Thanks for that explanation Sean. So basically, it's not thinking out of the box. So it can't be really innovative. It's a copy expert. It's basically polling what others have done, looking at huge data fields of original thinking and work, selecting favored probabilities, and spitting out the same stuff rehashed so you don;t recognize the original parts from the new whole. It;s not really imaginative or intuitive. New ideas will escape it.

It seems we humans still have a chance.

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

 

[nikos79]

Limitation of LLMs, but LLMs are not the end game of AI nor should be considered the future of AI. I'm already seeing 2 non LLM AI systems that might send LLMs to the junk pile and lead to AGI, full creativity and new science.

Can you share them with us?

 

[runswithsizzers]

Now, I'm sure people will use AI as a tool and be authentic about it, but there is the other dark side to it that will surely have a big impact on the authentic. With something so powerful, how do we know who is and who isn't being genuine? Does it matter? Will it matter? Is there a threshold for authenticity? What would that be? 95% human driven? Who could even prove it? I am expecting no one will care, or give up trying to care. I already know people listening to random ai generated endless music streams and they could not care less if zero human artistry involved.

This whole concept of authenticity fascinates me.

Imagine two wooden writing desks that are identical in every detail. Same dimensions, same wood, same construction details, same workmanship, same glue and varnish — same everything down to the most minute detail.

But one of these two desks has documentation showing it to be the exact same desk used by Thomas Jefferson to write America's Declaration of Independence, and the the other desk does not. One desk is an American national treasure, an heirloom of great value; and the other one is just some old desk. Yet, without knowing which is which, the closest examination and most rigorous testing will show the two desks are virtually identical.

The authenticity that makes the one desk so much more valuable than the other is something ghostly that exists only in the minds of men. Yes, there may be some paperwork to document the authenticity, but the value of the object itself becomes secondary to the back story, the narrative, the metadata.

When the last man dies who knows which desk is which -- on that day the authenticity of the Jefferson desk evaporates, and two desks become equal in value.

I am having a hard time understanding authenticity. It seems as if it might be somewhat delusional?

 

[nikos79]

What are LLM and AGI?

Large Language Models and
Artificial General Intelligence (the holy grail of the whole AI race for the next 20 years)

 

[runswithsizzers]

It's basically polling what others have done, looking at huge data fields of original thinking and work, selecting favored probabilities, and spitting out the same stuff rehashed so you don;t recognize the original parts from the new whole.

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.”

 

[nikos79]

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.”

This is also not true at all. It is not polling or stitching. The generated stuff may have never appeared in the data. It is in that way very novel

 

[runswithsizzers]

"Do a tintype":

So this is how AI views humans, murderous brutes with hammers?

Actually, I really like this image. Looks like it could be a wanted poster for Maxwell the silver hammer murderer.

 

[Cholentpot]

Oh definitely, but I suspect in another year these red flags will evaporate. I think the hardest ones to tell right now are product based photography images.

I'm not sure the red flags will disappear. The programs see what humans seem to like and copy it. AI loves the idea of shallow depth of field or deep depth of field. It sees that the most popular works online has these elements. However it just shoves it into photographs that have no business having it. It also struggles with real world scale and time. It can't figure out eras, they're not binary enough. AI will always err this way because it's reading what it interprets as what humans wants. Subtly is lost on AI, but to be honest it's also lost on a lot of humans.

Product based was nailed years ago. AI is still struggling with food and I think it always will. My field is portraiture and event photography. AI will assist but never replace what I do. Phones will eventually have AI assist portrait mode that will eat into my field but for the customer who knows that they want they'll still call a professional. AI can't line people up in groups no matter how good it gets.

 

[runswithsizzers]

This is also not true at all. It is not polling or stitching.

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

The generated stuff may have never appeared in the data. It is in that way very novel

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

 

[Cholentpot]

So this is how AI views humans, murderous brutes with hammers?

Actually, I really like this image. Looks like it could be a wanted poster for Maxwell the silver hammer murderer.

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.