Opinion on AI in art?

[BlankyBoyAUS]

I’ve just started learning about AI (artificial intelligence) in art and well I’m seeing a number of pros and cons regarding the ethics of using AI in art. Any artists here, could AI be an ethical method to create art or something?

 

[Edgewater]

Not an Artist, but an active buyer.
Not sure whether AI will obtain a level of originality, but duplicating, likely very quickly.

 

[PadPhilosopher]

So this all comes down to what AI is and does. AI really can't "create" anything in the proper sense of the word; it just rearranges things it already "knows" to satisfy a prompt. It could be argued that human artists usually do essentially the same thing, for example, a painting of a tree will always have the same "parts," with some variations, because a tree is what a tree is. AI can do that, too, but then I think of the work of John Howe, taking descriptions of things no one has ever seen because J.R.R. Tolkien imagined them, and making elaborate paintings from them. AI almost certainly could not do that. It can only rearrange things it has "seen;" it cannot imagine. That is the difference between the creative process and what I'll call the re-creative process.

I see no ethical problem with using AI to create images, some of which doubtless will be pleasing enough that some would call it artistic, but AI cannot reasonably be expected to be able to completely replace the human imagination in art.

 

[Sheogorat]

I’ve just started learning about AI (artificial intelligence) in art and well I’m seeing a number of pros and cons regarding the ethics of using AI in art. Any artists here, could AI be an ethical method to create art or something?

I have an unpopular opinion. I don't mind that the Stable Diffusion base model was trained on even images with restricted rights, but I don't think it's quite ethical to use images with restricted rights as the dataset for fine-tuning these models and selling outputs (or work made of this outputs). Because often author of finetuned model is tuning his model as much as needed to recreate author's original style.

AI is not an artist, AI is a tool. And AI enthusiasts trying to do masterpieces using this kind of technology.

But keep in mind that my opinion may be distorted because I am really bad at drawing, but I have enough knowledge of Stable Diffusion/similar diffusion models and Photoshop.

Mostly AI arts are very repetitive, because newbies (like me a while ago) trying to make many arts that "good enough to get 100 likes on DA", but doesn't even trying to do one really awesome work which will take as much time as it takes to make 10 simple works.

Spoiler: A good example of a good AI work

I find it unethical to use finetuned models to make money, even if you selling good work (it is about selling, not about creating patreon just for receiving donations). It is about the sale of subscriptions on services like patreon by the AI enthusiast, not about sale of services to generate such images using finetuned models.

Spoiler: TL;DL

Stable Diffusion​

Training on copyrighted images​

Selling images created using models that trained on Copyright Free images/images from authors that don't mind to use their arts to train models​

Selling images created using models that trained on Copyrighted images​

 

[foxcub]

im fully down for ai art, its way better than almost everything that people who claim to be artists do.
the majority of art schools spend most of their time making you immitate other artists, which is what ai art bots are doing, and it just floods the art world with crappy immitation stuff. so the way i see it, the ai stuff being way better than 99% of artists should raise the artistic value of truely origional work and force people to try harder to beat the 'baseline' of ai art if they want to stand out as a comercial artist.
Also, i find ai art is a very usefull tool to work with ideas that would usually take a a few weeks of trying to redo and redo a piece to get it to a place im happy with. i can just do some basic sketches, mash in any photos of source material im working with then generate 10 variations of the piece to see how it looks before getting started on a final piece. i work a lot with layering paintings over other paintings, which isnt possible to easily replicate with ai yet so its just a tool for me.

 

[JaysonTheRegressor]

Sometimes it's a good tool

Good for:
Early concept art
Fun/pretty images
Character ideas to be turned into full characters with real art

Not good for:
Real art, it will always look like AI art, following algorithms and not understanding what the shapes mean which creates weird smudging and odd shapes
COMMISSIONS you did NOT draw that yourself, you better not be making a profit for typing up words when any joe shmo with no talent can do it

ADOPTABLES You did NOT draw that yourself, any images generated by AI are public and you have NO ownership if you sell "AI art adoptables" you are scamming the public

That's the biggest point I wanna make, no one should sell AI art images
Because there are REAL artists out there who actually draw things for you, support real artists

 

[foxcub]

That's the biggest point I wanna make, no one should sell AI art images
Because there are REAL artists out there who actually draw things for you, support real artists

in which case people should sell things theyve drawn from photos. thats just copying.

 

[JaysonTheRegressor]

in which case people should sell things theyve drawn from photos. thats just copying.

Outed yourself as a nonartist i see lol
Using a reference image is FAR FAR different. It takes a hefty amount of skill and all skilled artists who want to know what they're doing do.

Meanwhile, AI-generated images (it's not art, it's generated images) is created by typing words. Which anyone with enough fingers can do
It's a different skillset yes but it's NOT art. Why don't YOU try to "draw something from a photo" lol

 

[Deleted member 58849]

in which case people should sell things theyve drawn from photos. thats just copying.

 

[JaysonTheRegressor]

Photography is also it's own art form but it's not competing in the same space like AI art is.

 

[JaysonTheRegressor]

My partner made an edit with a better take

 

[LilByte]

Outed yourself as a nonartist i see lol
Using a reference image is FAR FAR different. It takes a hefty amount of skill and all skilled artists who want to know what they're doing do.

Meanwhile, AI-generated images (it's not art, it's generated images) is created by typing words. Which anyone with enough fingers can do
It's a different skillset yes but it's NOT art. Why don't YOU try to "draw something from a photo" lol

AI generated art with no help looks like trash still in most cases unless it's something simple.
It takes allot of skill using sketched or generated control nets, thousands of generated images with slight changes each time till you get what you want, tuns of inpainting, sketching and photo merging. To get good AI art.
You also need to many times train your own custom LoRA's if you want something more specific like for example AB art.
.
In a few years when you can get fully everything perfectly done by the AI instantly and it's exactly what you want first try with no artifacting. Then I'd consider it to be no skill.
.
Till then I'd argue that unless your using fully hand drawn free hand no trace, without computer tools like copy paste and such computer assisted design, it's not "True Art"
.
As far as legality, it's not legal to sell images for more then the cost to generate them if generated with many AI's that you can download for free. However you could train your own AI from the ground up on open non copy written datasets and then sell images. But again your still going to have to do all the things from above to have art that looks good. And the more specific your looking for the harder it gets.

 

[JaysonTheRegressor]

It takes allot of skill

Time and patience maybe, but skill not so much
AI art isnt art. Down and out.

Till then I'd argue that unless your using fully hand drawn free hand no trace, without computer tools like copy paste and such computer assisted design, it's not "True Art"

This is just a bad take. Like saying a movie isnt a movie because they used CGI instead of practical effects
Art is an evolving form. Digital art is another medium like painting, sculpting, pencil work, etc.

Two artisit follow the same art idea. Each is unique. Two artists use the same reference image. Each is unique

Two people running the exact same word-for-word copy paste text prompt in an AI image generator will spit out the exact same image.


Your definition of "true art" isn't based off anything but your own opinion.

 

[LilByte]

Time and patience maybe, but skill not so much
AI art isnt art. Down and out.
View attachment 110849

This is just a bad take. Like saying a movie isnt a movie because they used CGI instead oif practical effects
Art is an evolving form. Digital art is another medium like painting, sculpting, pencil work, etc.

Two artisit follow the same art idea. Each is unique. Two artists use the same reference image. Each is unique

Two people running the exact same word-for-word copy paste text prompt in an AI image generator will spit out the exact same image.


Your definition of "true art" isn't based off anything but your own opinion.

Again almost no one is talking about 1 click run image generation. that's like making a box in paint.exe and saving it as the image and saying it's art. it will look like garbage with tuns of artifacts.
If you want to make anything that looks good and looks like what you have in mind when you start. Then you have to spend hours if not days on a single image even when using AI doing tuns of different techniques. There is a tun of human skill and creativity involved. AI is not anywhere near the level yet to do the entire pipeline.


This for example took hours and still needs a tun of work to remove artifacts.
And it wasn't just changing text2img either, this was thousands of different settings changes and tweaks, control nets, generating images and saving them in paint changing parts of them and putting them in as img2img and a tun other things.
I used two custom LoRA's that i trained.
I never finished the image because my main goal was learning more about control nets. but even after a few hours I only got this far with it. and it would take several more hours easily to clean it up to look good.

 

[ade]

I was wondering when photography would get a mention (I was going to do it, but this becomes such a big subject once you begin to peer beneath the veneer).
(Incidentally, in the above cartoon, the lens would obscure the light from the flash at the shown subject distance, and the glint on the lens doesn't match with the casted shadows )

I think it's safe to say that the debate about photography being 'art' continues to rage, as it has done since it's early days.
Indeed, the similar debates of 'is it art?' 'what is art?' also continue.
The reason why it's all so debatable goes back to, and is thus built upon, the origin of 'art' and it's 'dismissable' counterpart, 'craft'. Simply put, they're the same thing, it's just that 'craft' is English and 'art' isn't. And just because I'm saying this from an English-speaking/British perspective, doesn't mean that such such debates, based upon prejudices and discrimination, didn't also occur in France (from where we get 'art' and continue to seek inspirational and ideological affirmations of our likes and dislikes).

That being so, we've inherited a living cultural legacy (for all it's racist and classist bent) and sort of know how to use 'art' and 'craft', which is why it occurred to me that the shown processes of creating Ai images looked very crafted (could we thus also say 'arted'?), be that at the coding or imaging level.

Will most of the results be art/craft?
That sounds like more the real debate [maybe yet to come?].
Like, we're all familiar with somebody doing something as if were 'art', even if it doesn't meet any of the accepted pretentious definitions of 'art' (machine operating, race/rally driving, motorbike trialling, gaming, etc; we'll even describe systems operations, autonomous or not, as 'balletic' in their motion) and where, for the most part by most people, it's just a basic level of functionality in use, devoid of any real mastery or graceful motion, but a few will, by skillfulness and mastery, elevate the activity above the base into something which can only be described as 'art'.

 

[JaysonTheRegressor]

Again almost no one is talking about 1 click run image generation. that's like making a box in paint.exe and saving it as the image and saying it's art. it will look like garbage with tuns of artifacts.

Did I not just say "time and patience"
Whatever this is pointless

 

[foxkits]

What's making the furrie artests up set is they're taking parts of there art to train AI as any of the artest work be it any venue of art .and they feel it's stealing from artists to train AI.

 

[PadPhilosopher]

Some illustrations from "this person does not exist" of just how the fact that AI doesn't really understand the meaning of things makes it weird sometimes. This algorithm is very trained to make faces, and does a beautiful job. But anything else, it starts to get weird quickly.


Lovely lady........but that hat. Clear illustration of AI not understanding something, in this case flowers, and maybe a brim.


Again, this is a lovely face, but that neck is horrifying. Indescribably weird.


A lovely lady with a lovely neck, but.......that thing in the background, right side of the frame, is one of the most disturbing things I've ever seen.

I'm not at all sure what I'm arguing here; I guess I'm just illustrating.

 

[lilSorcerer]

When asking a question of ethics, I always think it's a question of what should I/we do? There are many ways to approach this question such as from moral views, or practical views.

Personally, I think it would be morally/legally questionable to profit off of ai art without paying dues to all the art and content used to train the model, as without consent this falls under copyright violation. Feeding, and abstractly storing "copies" of art is definitely a violation.

Practically, I personally am afraid of generative ai, but not because it is too good, but rather because it is too poor at making works of art, especially when it comes to writing. While it produces ok content in some areas and applications, it is generally much poorer than what even a moderately skilled writer can produce; however, any butthole can quickly and cheaply produce and distribute their poor quality works, choking out works of good quality and evermore contributing exponentially to the information overload that characterizes our "digital age".

 

[LilByte]

Did I not just say "time and patience"
Whatever this is pointless

so then when you drag a stick around on a paper i guess that's time and patience. if the skill of how you do things while spending the time doesn't matter to you.

I didn't just click a button and walk away to do the above image. I spend hours changing different settings, words, percents of LoRA's, traced out different areas for the model to focus on, merged images together and all kinds of other things. I created control nets and had the model overlay data using them. There are a tun of steps i was working every moment of it, my computer only takes a few seconds to do the generation parts. but i had to do it thousands of times with many steps that take far longer in between that vastly effect the results.