Thoughts on artificial intelligence in art creation

Adobe has a message for creatives worried about the rise of AI art

1 Like

I have never used pixiv personally but I do find myself increasingly moving away from sites that allow it because now, with people not bothering to tag it and given the sheer output it’s capable of, it becomes less worthwhile to even browse them if you are looking for art made by people.

1 Like

Just today, i got email that the site is trying to mitigate the concern by moving AI Commissioned work on its own category. It’s not perfect, and nothing can be. People still can be dishonest about what they posted.

But this would mitigate my(any many other people’s) concern that actual artist would get drowned (they already are) by people purely using AI to create and post literal hundreds of image in a single day.

2 Likes

Adding to that discussion of why website won’t do much in terms of AI, is because there isn’t a win-win situation for many art websites, in terms of the ai art. There is few option websites could have done/ tried but their not great solutions either.

one option to gatekeep/ask for artist to upload a video every time you wanna upload an artwork, which sounds good in theory, but theoretic speaking how many artist know how to screen record videos and speed it up. Or if an artist finished an artwork and they forget to hit the record button. Or better yet have the camera/equipment they need to actually record their stuff.

This option will potently limit many artists (beginners/hobbyist…etc) from uploading their artwork, because their either do not have equipment/time or knowledge to know how to do it. Or artist could feel ashamed that their video is not 100% perfect; which results that they will not upload because artist feel ashamed/embarrassed even.

Another option website either let bots or people flag the artwork that is consider AI art work. Case study wise look at YouTube. There are many examples of YouTube mishandling their AI bots, better yet mishandling their own system. Their system is basically, If you don’t have enough fame to call out YouTube, Your videos and channel can be erased/deleted wrongfully so.

If we let people handle the flagging, humans are not perfect and they themselves could go on a power trip. Hypothetical speaking, if a person does upload an artwork and for whatever reason someone didn’t like it. They could easily claim that this artwork is made from an AI. If they have enough followers they easily tarnish the artist reputation/bullied the artist off the platform.

Humans could apologize yes yet we see many times before “An apology does not fixed the damage or bring back the reputation that artists once had" Keep in mind it take a long time and trust for artist to build up a good reputation and it can be over by wrongful accusations.

The third option is where many websites are at moment because they themselves don’t have a great solution to this. Is just let people upload art and claim it as art anyways. It is the easiest solution, plus due to AI art they can bring more traction to their website. What I mean by traction people who upload ai art, can upload a lot bountiful of artworks like 10 to 20 within a day. Compared to an artist who spends time a lot of time fixing their artwork before they uploading it. To a lot of websites/apps quantity beat quality every time.

Which is why we will probably see a lot more artist burned out/quitting because they can’t keep up.

Yeah, I get why they are doing it but it seems shortsighted. The easier and faster people can have an AI generate images at-will, presumably the less need they have of gallery sites. Those are useful for artists and photographers, not so much machines. If I could make art at the rate an AI does I would not even need to save it to my hard drive. Why bother when I can just do it again so fast?

If adoption continues to increase I think these sites may genuinely find people are generating and uploading images nobody ever sees for the sheer volume of them being churned out, even without the tech necessarily getting much better. They may end up participating in making themselves obsolete.

2 Likes

I see your point better now. Also didn’t wanted to look rude
Is just that it seemed that you wanted to imply that i didn’t participated the creation at all, and is totally untrue.

We can say that i commissioned the AI to do a work like you say, i have no problem to let it clear that i used it and all.
Is just that, my process on painting normally is me trying to find shapes that works for what i want my paintings to be, and it takes hours. With AI is also me trying to find these shapes, but it’s not me doing them directly and i have way more options in much less time. But It’s still me choosing to work with these shapes or not, molding them if i need, etc. What I already do painting with a lot of Ctrl-Zs sometimes.
What i love more about painting is the problem solving part on finding out what works for the image, and the ai can help with that a lot.

You can say that I’m more of a art director than a painter in this case, that’s ok, although i painted on top a lot. But is me expressing at the end.
I know this is all kind of blurry right now with all these “ai artists” just generating results and calling it a day, and that’s a big problem.

There are two pretty easy tests to check if something is still just a tool or not.

“Can you do your thing without the tool in question?”

If the answer is No, then it’s not just a tool.
For example, take Krita or Photoshop or the entire PC away from digital artists, they can still do art with pencils and ink. I know no digital artist who can’t do at least decent traditional works. It’s maybe not exactly the same and takes longer too but they can keep doing art even without the tool (that is the computer in this case).

Take away the AI image generator and suddenly a lot of people won’t be able to do the thing anymore. So for actual artists it maybe just a tool in some cases.

“Can the tool in question do the thing without you?”

If the answer is Yes, then it’s not a tool.

Example: Robot arms on a conveyor belt can do the assembling themselves. No human needed except maybe to power them on. Image generators can work just the same without anyone writing the prompts. There are even prompt generators and their output could easily be piped into the image generator. No human needed anymore to do the thing.

This isn’t exact science of course, there can be overlaps but it can give someone a sense of what is a tool or a replacement.

5 Likes

Tools are intended to ease the labourious grunt work. Is art a grunt work to you if your answer is yes then you have bigger question to answer.

The answer may vary for people. For those who think art is repeatitive grunt work and want only the end result AI looks like a tool. For other it is not.

6 Likes

Just a a random thought:
I’ve been looking in a lot of AI generations lately because I’m following this scene very close and because something new happens every day and in trying to implement it all in my workflow.
It was funny how, i got back to Twitter and to see work of real artists in my feed and it became very clear to me the difference immediately. Not in the technical side, AI is already doing beautifull things, but they are just beauty for beauty. Real art has intent, it’s made to make you feel something, AI art don’t do that. This immediately teach me how to be a better artist and became clear to me how the technicalities of a image are not what really matters.
For me it’s just that i used to give a lot of praise to artists that know how to render a image good, and them all of suddenly i started to praise the ones that got a feeling right out of their works, and these feelings also started to be much more clear to me. Really interesting stuff.

5 Likes

For me personally this debate has ended. When it arrived, it had its novelty, it had its share of panic and fear inducing phase. Now I am just concerned with my own work. I do not have good graphic card so I am left out of the race and thus it is not yet “democratized” for me as they say, yes I could use google colab but it is for research only not for commercial work.
I tried the web demo of stable diffusion and with few tries it failed miserably to give me what I imagine in my mind. Some time it errored out with the description that I gave it, May be I am a bad “prompt engineer” (I never intended to be an engineer to begin with otherwise I would have chosen that stream when i was in college), May be my imagination power is beyond AI :stuck_out_tongue:

This technology if it is 100% effective is direct competitor to me, since I produce commercial art for clients in different style, medium depending on what the problem at hand is. And when I tried this as a client it failed. Clients are really demanding in their requirements, there are countless iterations, brand guidelines, nuances of each brand etc, which the dumb AI tool doesn’t even come close to solving. There is also the problem of client themselves not knowing what they want sometimes, we have to solve the problem of coming up with a message, thinking about the target audience etc etc. AI falls flat in executing things which are specific. After that experience I know that AI can’t do what I do yet. Even if it does I am not worried as it can’t visualise and solve a problem yet, client can’t visualise and solve the problem themselves, so I am not worried with this. I have even muted the social media about this, which are mostly techbros trolling artists since they have a regret that they couldn’t choose art as a career :slight_smile: I would advice others too to move on and spend time in something productive :slight_smile: Go make a sketchbook thread in here or pick up that unfinished piece you left two years back :slight_smile:

For the curious one of my prompt in the grotesque category was “A white slimy giant Millipede which has human hands as its legs crawling on a barren black soiled planet with green sky filled with red crows with human head without eyes flying above to hunt it by spewing acid from their mouth” As you can see this is not some white girl standing with mangled hands type of prompt :slight_smile: it is very specific and AI falls flat in understanding any of it. May be if I break down and slowly build the image or if I use image to image it may be able to get what I imagine but at that point I’ll juts paint it.

4 Likes

Here’s a simple concept i’ve made around your idea with it’s help just for fun.
It can’t do a lot of things and i don’t think it will be able to do in the future because it takes artistic sense and experience to visually solve artworks around an idea. and there’s all the Client thing you mentioned, so unless it goes full AGI i don’t see it capable to do it on the next years. So i don’t think you need to worry really.
It’s a tool at the end of the day. But it will take good artists to use it right, and the “prompt engineer” knowledge will be needed for it. It’s all too new, keep changing, and i think it will take a while for it all to settle down. But i can see people that don’t learn to use it falling behind on the industry in the future, and that’s a point not to be ignored for people working on the field.
For the hobbyists that are the majority in this forum i believe, i’d say don’t worry about it that much, just focus on your expression.

I notice it that with all this “AI art” discussion lately something happened to me: I started to gain a LOT more of appreciation of the efforts made by young people that are starting to learn how to draw. I don’t know why.

Maybe because “drawing” is now a thing that you can do effortless and I appreciate the struggle? idk, just a thought.

Oh Thank you very much for trying.

I know you did this for fun and not to be rude to you and your effort in coming up with an image for my insane idea thought on a whim to test the demo of AI. The feedback I acting as one of my client would give is this (This is just give a gist how clients would give feedback and how the AI doesn’t give you what you imagine)

Hey Raghu, thanks for working on this in such a incredible speed.

The image looks good but it is no where what I had imagined. The millipede’s legs do not look like human hands at all, there should be more details in it. There is no feel of giantness in the millipede. the planet should have black soil. The crows don’t look like crows or a bird for that matter. The crows should have human head. There is one big head of something but this was not in my description. Instead the crows should have human head and they should be spitting acid. Overall this image strays away from what I had in my mind and doesn’t work out for our message yet but it is a good first starting point. I am sure you’ll be able to come up with more convincing sketches based on this feedback. Really excited and looking forward to what you come up with next.

This was just how clients would give feedback, it will be more brutal to be honest. And some times clients would say something vague like “it doesn’t have the feel to it”. it is the our job to find a solution and AI doesn’t help me in that yet. Again this is not to be rude to your image or anything I just want to give a perspective of what client will probably say in my experience.

Yeah I am not worried at all, that is why I started my message with “For me the debate has ended” I don’t worry at all.

Well it depends on people and what they want to do. Some people will work like a factory and churn out image after images in short time some will take days, it happens now too. With AI this will get more intensified with AI. I don’t think any artists will shun this technology if it is created using ethical terms and if it helps them. After all artists are the group which embraces innovation and technical things. For example in olden days artists would try to find new pigments colours, some used a new technique called camera obscura to capture nature as it is, which was like a primitive version of camera. Artists are known to embrace new technology. While nobody will ignore this there is also no need to create a sense and fear of missing out if we don’t use this. Like saying you will loose your job if you don’t learn this. it is just a stupid tool after all, some people do traditional some do digital nobody is saying traditional people are left out. it is their concious choice :slight_smile:

Ultimately this tech will be used by corporates and moulded into subscription model, recently adobe released some more features with AI, it will be milked and artists will be paying for this. And it will not be some sort of revolution which “democratizes” art, after all to do art you just need to think, you can do art with a stone too, nobody is gatekeeping anything.

2 Likes

AI artwork isn’t going anywhere, you can download, share, gather, and train models on your own hardware at home with no internet. It’s as unstoppable as torrenting.

I believe art sites should quit it with the defensive approach ‘We will ban artists who use AI’. Why? Because it’s not future-proof. It’s a knee-jerk reaction to the sudden appearance of this threat used to soothe themselves and their users. AI will get to the point where it is absolutely indistinguishable from real art very soon and this will not work anymore.

If any gallery sites are smart and realize this, they will take a proactive approach to their content moderation. They don’t necessarily have to accept AI art, but they will have to have a verification process for artists to prove that they are actually creating the art they’re making. This can be done through speedpaints, streaming, and having an archive of your older works so people can see your skill progressing naturally.

Ah yes, if it was a work for a client my approach would be totally different, is part of my job as a Illustrator to help the client to visualize what they want so I’m kind of used to this.

About the fear of mission out you mentioned, This technology launched 2 months ago, my PC could not run it at first because it need 12 gb of vram, now it needs 4 GB. My PC took 50 seconds per image at first, now it takes 25 on the same settings. There’s already possible to train the AI how your face looks and it actually can do generations with you now, When this feature appeared you needed 24 gb of VRAM to train it, 5 days later and you need just 10 GB. This is all powered by the open source community who is developing this at a speed rate like i never see anything before. Literally ever day there’s a new optimization or a new feature and it have been like this for a month.
Imagine all this in years time.
I do think that people who work in this field need to have eyes opened for all this.

Yeah I am not denying the progress at all, in future may be we would not be using any painting software at all, may be client won’t need artists they would learn to write prompts etc. we can keep on guessing and waste time. I am also not saying anyone should not use this. I am not giving a counter point to whatever you said. Whoever wants to use it they can. I will use it may be or may be not who knows future. Right now it is not for me. And I am bit tired by the hype and evangelist selling it to the face just for the benefit of the corporates is all my opinion.

Take this thread for example we are going on and on and on in circles telling the same thing again and again. There is no end to speculations and other stuff. Just get over it use it if you want or don’t use if you feel. We can’t do anything to stop, improve or shape it from this forum thread. It is a giant waste of time.

I think we should get back to what the forum is about, that is krita software. Hence I said earlier “for me the debate is dead and over”

6 Likes

Except maybe for streaming this only proves you managed to train your image generator in different styles or quality. And streaming isn’t an option for many. You would also have to do this for every artwork.

1 Like

Video recording should work fairly well, considering the AIs do not show their process at all. I don’t imagine most sites would want to be as drastic as needing a video for every upload, because it could be quite a pain to do on their end as well, but verifying that someone at least can make what they’re uploading would be more feasible and cut down on a lot of people who are just posting unedited AI images.

1 Like

I doubt this is going to happen on a large or medium scale. Most popular art websites already have a lot of trouble managing their current moderation workload. I guess we have to wait how they’re going to solve this, if they try at all.

I also kinda fear that there will be a general mistrust especially for digital artworks and artists. Just today I saw generated furry artworks that looked indistinguishable to human made ones. I only knew it because the account was flagged as AI image account.

2 Likes

Not long ago, a test was conducted on a social networking forum in China. More than 400000 people participated in the test, and 85% of them could not recognize the works drawn by AI well……

1 Like