AI-generated art

[Judge Spear]

One thing I can see happening is an increased demand for artists to stream in order to demonstrate that their art is human-made.

Maybe but I dont see that happening too often in the future really. You can just as easily ask for them to provide a canvas file that shows each layer you've used to make the image.

And if we're talking about commissioners demanding this, artists are supposed to be showing each step of the process from sketch to completion. It would be exceedingly difficult to try and weasel some AI dreck to your clients.

And as of now, theres so many tell tale signs of AI. Despite the "leaps" made in the tech, the best of these machines have consistently had the same exact shortcomings with the exact same frequency since day one.
I think we're deeply underestimating the sheer amount of very human deliberation that goes into the design process no machine is going to account for.

 

[Rayd]

i think AI generated art is a technological marvel and really cool. i don't think anybody creating art using AI actually believes they're a real artist so i don't think there's a problem. if an AI manages to beat real artists in an art competition, i think that just goes to show how far the technology has actually come.

and if it becomes even better over the coming decades, that's great, too. but i don't think it'll ever replace what someone can do by hand. so i don't understand why the art community has been so up in arms about it.

 

[Judge Spear]

i think AI generated art is a technological marvel and really cool. i don't think anybody creating art using AI actually believes they're a real artist so i don't think there's a problem. if an AI manages to beat real artists in an art competition, i think that just goes to show how far the technology has actually come.

and if it becomes even better over the coming decades, that's great, too. but i don't think it'll ever replace what someone can do by hand. so i don't understand why the art community has been so up in arms about it.

'A.I. Should Exclude Living Artists From Its Database,' Says One Painter Whose Works Were Used to Fuel Image Generators | Artnet News

The digital artist Greg Rutkowski has seen his style of painting copied by A.I. image generators, which use databases of millions of images.

news.artnet.com

I think the doom and gloom is a bit excessive, but it definitely poses some cause for concern.

"MIT’s Technology Review reports that Rutkowski’s name ranks among the most used prompts on Midjourney and Stable Diffusion, two popular open-access A.I. image generators, where users have input Rutkowski’s name 93,000 times. That is far more than users’ requests for images similar to the style of Michelangelo or Picasso, whose names have been employed as prompts no more than 2,000 times each."

"At first, Rutkowski considered his newfound popularity on the A.I. platforms an avenue to new audiences. But when ran a web search of his own name for other reasons, works in his style he’d had no hand in making turned up."

This is why corporations are defensive of their own intellectual property and nip illegitimate uses of their work in the bud very early. This is a problem for branding specifically when your name is searched and an alarming amount of results lead to things that other people made using your content but not back to you. And it's especially an issue when you're a not a corporation but a single creator.

 

[TyraWadman]

AI will never be able to create the things I have going on in my brain and I can't imagine I'm the only one that feels this way.
It's going to be a nuisance for sure, but there's no way for it to kill the entire commission community.

 

[Gushousekai195]

I was going to use AI to generate complex backgrounds for my art and have the character(s) interact with it (and I usually prompt it to create stuff made by no particular artist or photographer), but now FA has banned it. It upsets me to see people saying that AI art is just a collage of bits and pieces of stolen artwork by real artists.

I don’t see anything in it that screams “collage.”

 

[Foxridley]

Maybe but I dont see that happening too often in the future really. You can just as easily ask for them to provide a canvas file that shows each layer you've used to make the image.

And if we're talking about commissioners demanding this, artists are supposed to be showing each step of the process from sketch to completion. It would be exceedingly difficult to try and weasel some AI dreck to your clients.

And as of now, theres so many tell tale signs of AI. Despite the "leaps" made in the tech, the best of these machines have consistently had the same exact shortcomings with the exact same frequency since day one.
I think we're deeply underestimating the sheer amount of very human deliberation that goes into the design process no machine is going to account for.

Yeah. I forgot about WIPs. Outside of streams, a lot of artists either don’t offer them or only provide them upon request.

For some reason I was thinking about speed runs, which are recorded to prove no hacks or disallowed strats were used.

 

[Judge Spear]

Yeah. I forgot about WIPs. Outside of streams, a lot of artists either don’t offer them or only provide them upon request.

For some reason I was thinking about speed runs, which are recorded to prove no hacks or disallowed strats were used.

Lmao
Yeah speedrunners better be streaming with as much proof as possible.
But honestly, not giving WIP's (at least 2) is very bad practice. I've recently learned this is not standard myself which shocked me because I figured that was just...an innate thing to do. Like that makes zero sense to not give updates.

 

[Foxridley]

Lmao
Yeah speedrunners better be streaming with as much proof as possible.
But honestly, not giving WIP's (at least 2) is very bad practice. I've recently learned this is not standard myself which shocked me because I figured that was just...an innate thing to do. Like that makes zero sense to not give updates.

Interesting. I think only one or two artists have given me more than one WIP.

 

[Minerva_Minx]

Are the people using this ai calling themselves artists? I saw comments in that post saying "at this point you're not an AI artist but a multimedia artist" which leads me to believe people think just putting text into a robot is being an artist. Which seems...insulting to people that actually create art themselves.

I know what an artist is may depend on who you ask, but it just doesn't seem like it takes much personal effort or creative input to make ai-generated art in my opinion. Like sure it can look good, but you're not really being an artist imo if the ai is doing all the work for you. The intensive editing is one thing, I can respect that. But direct from the ai creations? That doesn't make someone an artist in my eyes.

No, it doesn't make an artist. I have tried them, even posted works on dA and a couple here. it's interesting, but not ultimately unique. It fills the niche void of you want FaceApp, Adobe, word, powerpoint, amd a powershell script to c4eate a picture based off random math.
Probably for the best it is banned or curbed with a lot of disclaimers.
Some word choices on midjourney and dall-e will expose a hidden watermark.

 

[Frank Gulotta]

That's pretty much what I meant
AI art is fascinating in its own right and has its place around

 

[Minerva_Minx]

It's just as fascinating when we agree...

 

[Judge Spear]

Interesting. I think only one or two artists have given me more than one WIP.

Thats fucking WILD. What the Hell?

 

[Foxridley]

Thats fucking WILD. What the Hell?

And here I was worried about being too demanding if I asked for a WIP.

 

[Minerva_Minx]

Thats fucking WILD. What the Hell?

The higher commission and detailed premium artists - oh god yes I would like two or more WIPs in most cases.

Here is where I piss off friends who are artists. (I love you guys, please don't hurt me!)
My reasoning is this: there is a different between speedpainting and interactive details.
Speedpainting has no focus on detail, is all about mass production, and spends the least time on the most things possible. This is the equivalent, to me, of going to Goodwill to pick out furniture, but only spending $50 and whatever I can scrounge from a couch cushion and taking whatever I get and being happy. if you are the viewer, it seems generic and reproducible:

https://www.deviantart.com/minervaminx/art/Me-the-Succubus-873820375

Detail requires purpose and coordination. everything that can be tuned is, anything else is left alone. Walls are neutral colors, so furniture is still neutral but slightly lighter or darker, so pillows and paintings on the walls contrast so the colors all "pop". this is what grabs you attention and makes you look at the intricacies. These artists usually, but not always, charge $200+.

https://www.deviantart.com/minervaminx/art/Me-Spellsword-854257005

Then you have AI and...

https://www.deviantart.com/minervaminx/art/Craiyon-Raging-Frustration-928360597

and that is a horror show of just WTF meets FML using photos and commissions. Does it make me an artist? No, it males me a sad, depressed woman with too much time on her hands succeeding at being more miserly than Scrooge McDuck.

 

[Judge Spear]

One of the common things I hear people saying as a use of AI is to treat it as reference or inspiration. While on paper that sounds good, here's the problem with this notion as AI currently stands and will almost certainly stand for years...

Novice artists mainly would be the ones to find this method of reference enticing. But it's a bit of a trap. You don't exactly know what to look for at the beginning stages of learning how to structure a drawing. This comes with mileage, critique, and fervent practice to be able to pick out flaws and know what to avoid. The really glaring errors might jump out at someone still learning, but even those, in any work, regularly do not get noticed by beginners. So for sure, they'll be picking up the more numerous, nuanced flaws they simply aren't trained enough yet to spot. Not to imply this will ruin them forever, but their per piece results will be worse than if they studied a master or some photos.

An adept on the other hand simply has no need.
Someone who's been drawing (and taking it seriously) for 5+ years will have books, photos, and hundreds upon hundreds of valuable artist references to pull from. This is on top of their own muscle memory for how things work built over years of studying.
You have to spend some time experimenting with a bunch of parameters to plug into a machine. And what you get in return, at best, will only ever be stock compositions of generic "designs" with dead faces in very basic often cropped poses. None of which are conducive to advanced growth. And for it all to also come with numerous errors on top of that? It's completely worthless.

For both the adept and the novice, even from the angle of just "simple inspiration" as opposed to a learning tool, there's no point. With no exaggeration, there are several hundred quintillions of photos on just the Internet alone. There are are almost 5 billion photos uploaded to the Internet daily. Petabytes of high quality photos from the last 5 years alone and a literal MILLENIA of master art of every genre and style conceivable. All at your disposal to find inspiration from and study.

So experienced artists can make no use of this and novice artists should avoid it.

 

[Foxridley]

One of the common things I hear people saying as a use of AI is to treat it as reference or inspiration. While on paper that sounds good, here's the problem with this notion as AI currently stands and will almost certainly stand for years...

Novice artists mainly would be the ones to find this method of reference enticing. But it's a bit of a trap. You don't exactly know what to look for at the beginning stages of learning how to structure a drawing. This comes with mileage, critique, and fervent practice to be able to pick out flaws and know what to avoid. The really glaring errors might jump out at someone still learning, but even those, in any work, regularly do not get noticed by beginners. So for sure, they'll be picking up the more numerous, nuanced flaws they simply aren't trained enough yet to spot. Not to imply this will ruin them forever, but their per piece results will be worse than if they studied a master or some photos.

An adept on the other hand simply has no need.
Someone who's been drawing (and taking it seriously) for 5+ years will have books, photos, and hundreds upon hundreds of valuable artist references to pull from. This is on top of their own muscle memory for how things work built over years of studying.
You have to spend some time experimenting with a bunch of parameters to plug into a machine. And what you get in return, at best, will only ever be stock compositions of generic "designs" with dead faces in very basic often cropped poses. None of which are conducive to advanced growth. And for it all to also come with numerous errors on top of that? It's completely worthless.

For both the adept and the novice, even from the angle of just "simple inspiration" as opposed to a learning tool, there's no point. With no exaggeration, there are several hundred quintillions of photos on just the Internet alone. There are are almost 5 billion photos uploaded to the Internet daily. Petabytes of high quality photos from the last 5 years alone and a literal MILLENIA of master art of every genre and style conceivable. All at your disposal to find inspiration from and study.

So experienced artists can make no use of this and novice artists should avoid it.

I’ve watched streamers to get an idea, of the drawing process. Not something you can get from AI.
I’ve noticed a lot of midjourney images are portraits with a character completely face on, which is not an angle commonly shown in human made art. The symmetry is a little unnerving when you see a lot of those images together.

 

[Zara the Hork-Bajir]

Having looked through some of the AI generated art and the responses reminds me of the time that Autotune just started getting used in basically all popular songs. Everyone was sure it now took zero skill, song artistry was dead forever. I think over time people will come to accept it more and more as just the way things are, right now people are just playing in the sandbox seeing what they can get away with, where it works and where it doesn't.

What bothers me is how all AI generated art has now been labeled as "lacking artistic merit", simply because a brand new tool was used to make it. Do all autotune songs have no value? While nobody like to admit it some of those songs were more catchy than a plague. Having a seperate tag or category for AI generated/assisted art is a good idea, not sure why a complete ban is necessary.

 

[Green_Brick]

Hear hear, people said the same thing and felt the same about Photoshop when it first came out. It's not going to replace things entirely, rather it will just open up a whole new paradigm. Just because Photoshop exist doesn't mean that traditional art is any less valuable.

 

[TyraWadman]

What bothers me is how all AI generated art has now been labeled as "lacking artistic merit", simply because a brand new tool was used to make it.

While the sensitivity of the wording is debatable, it's different to voluntarily alter your own voice, vs posting an image that plagiarizes existing photos and artwork (to the point where it can sometimes include the artists signature).

I find there are enough image dump websites where things like memes and ai art can thrive. Haven't seen any news about banning AI works on Deviantart. I see nothing wrong with websites wanting to separate themselves/offer different things.

 

[Foxridley]

Having looked through some of the AI generated art and the responses reminds me of the time that Autotune just started getting used in basically all popular songs. Everyone was sure it now took zero skill, song artistry was dead forever. I think over time people will come to accept it more and more as just the way things are, right now people are just playing in the sandbox seeing what they can get away with, where it works and where it doesn't.

What bothers me is how all AI generated art has now been labeled as "lacking artistic merit", simply because a brand new tool was used to make it. Do all autotune songs have no value? While nobody like to admit it some of those songs were more catchy than a plague. Having a seperate tag or category for AI generated/assisted art is a good idea, not sure why a complete ban is necessary.

I guess,
But I think there’s a substantial difference between AI-generated art and AI-assisted art.
The latter is closer to your autotune analogy.
But would people think the same for a song written and sung by a computer?

 

[quoting_mungo]

Having looked through some of the AI generated art and the responses reminds me of the time that Autotune just started getting used in basically all popular songs. Everyone was sure it now took zero skill, song artistry was dead forever. I think over time people will come to accept it more and more as just the way things are, right now people are just playing in the sandbox seeing what they can get away with, where it works and where it doesn't.

What bothers me is how all AI generated art has now been labeled as "lacking artistic merit", simply because a brand new tool was used to make it. Do all autotune songs have no value? While nobody like to admit it some of those songs were more catchy than a plague. Having a seperate tag or category for AI generated/assisted art is a good idea, not sure why a complete ban is necessary.

These tools aren’t like autotune, though. Autotune is like a Photoshop filter. You go out and snap a photo and slap a filter on it people might call you lazy, but the work (the photo and by extension the modified one) is still yours.

When they make synthetic voices for people who need to use text-to-speech in their regular life (think Stephen Hawking), the people whose voices are sampled are aware and have consented to that use. No one has a problem with that.

If you were to go out and grab the audio of a large body of TikTok clips, YouTube videos, and podcasts, and use those to teach a computer speech, that might be acceptable as a research project, either in Linguistics and related fields or in AI/computer parsing research. But if you go on to then release your computer voice maker for anyone to use? You’ve crossed a line. You used the work of others, which they own the rights to, without their consent. It doesn’t matter if the computer doesn’t use any one person’s voice straight. Yes, it’s a problem if this hypothetical computer voice maker undercuts and replaces voice actors, but it’s much more of a problem how it’s using the labor of people without permission or acknowledgment.

 

[Faustus]

One thought I've been having recently is this. AI isn't programmed. It's trained. You show it stuff, indicate a portion of the image and say, for example, 'this is a boat' or 'this is a fish' or 'this is a surrealist commentary on contemporary values in Sweden using the medium of pig faeces', and the AI learns.

Human artists learn in a very similar way.

So, while I can understand arguments against AI art because (for example) it makes it more difficult for human artists to find work, I don't entirely agree with arguments that say it's copyright theft to use images to train a model. Every artist uses reference images at some point. It only becomes a problem if they are recognisable in the artist's own work.

Besides, if you make that specific argument, you also have to ban all those human artists who draw 'fan art'.

 

[quoting_mungo]

One thought I've been having recently is this. AI isn't programmed. It's trained. You show it stuff, indicate a portion of the image and say, for example, 'this is a boat' or 'this is a fish' or 'this is a surrealist commentary on contemporary values in Sweden using the medium of pig faeces', and the AI learns.

Human artists learn in a very similar way.

So, while I can understand arguments against AI art because (for example) it makes it more difficult for human artists to find work, I don't entirely agree with arguments that say it's copyright theft to use images to train a model. Every artist uses reference images at some point. It only becomes a problem if they are recognisable in the artist's own work.

Besides, if you make that specific argument, you also have to ban all those human artists who draw 'fan art'.

But you can ask those humans about their influences and they’ll tell you. They will (generally) correctly attribute fanart and homages. (And will get in hot water if they try to make money off of anything owned by The Mouse. ) They have learned not only from pictures, but from watching the world. And human memory isn’t static. We interpret, fill in blanks, and so on. Drawing a giraffe from memory a couple days after you sat at the zoo doing gesture drawing by the giraffe enclosure is going to yield a different result than doing so months or years down the line.

Trained computer applications that generate images directly derive their value from the work they were trained on. We don’t allow people to use art they don’t own the rights to (and that aren’t in public domain) in order to create a product in any other sphere. When it’s done, there is justified backlash.

The conversation would be very different if publicly released generators only used public domain work to train their models. Hell, it would be different if there weren’t already people monetizing (ie charging to use) their particular generators. But using work you don’t own the rights to in order to generate value and quite possibly undercut those same rights holders is unethical. Allowing prompters to request art in the style of artists whose work has not yet passed into public domain is unethical.

Hell, there’s training data used for some of these generators that wasn’t supposed to be public at all in the first place. And there’s not really any way to “untrain” specific things from the algorithm.

Artist finds private medical record photos in popular AI training data set

LAION scraped medical photos for AI research use. Who's responsible for taking them down?

arstechnica.com

 

[TrishaCat]

I did this in about 2 hours in NovelAI and Photoshop.

This particular genie is out the bottle. I would advice against a blanket ban, just create a new category for it.

Spoiler: Vore

https://forums.furaffinity.net/threads/how-to-properly-handle-adult-content.1655576/

Hiding NSFW content within spoiler: BAD (Anyone can click it)

 

[Firuthi Dragovic]

I did this in about 2 hours in NovelAI and Photoshop.

This particular genie is out the bottle. I would advice against a blanket ban, just create a new category for it.

((image removed from post))

You probably could have picked different subject matter for a forum post... but let me give this a try.

So... what was with the knee, shoulder, and lower jaw of that dragon? Were those markings the AI's idea, or yours?