Weekly Discussion: How does AI "art" impact the furry fandom?
Hello!
This will be the first of our weekly discussions. This thread will be pinned in the local timeline all week long and I hope that everyone feels encouraged to voice their opinion.
This thread will be locked after a week. Make sure to join the discussion in time!
This weeks topic is: AI "art" and its impact on the furry fandom and artists.
Here are some guideline questions that you can use:
What are your opinions on AI generated text and images?
Should furry spaces allow AI generated images?
What do you think are the use cases for these tools?
What long-term impact do you believe it will have on the fandom and the fandom's artists?
So, I'm a programmer and my thoughts here apply to AI generated code and ChatGPT writing as well.
I think these kinds of modern generative AIs create derivative works based on their inputs. Which means that the training data should consist of only public domain materials or other works that give you an explicit right to use them for training purposes. Blindly scraping the internet for training data without regard to the owner's wishes is immoral and IMO should be copyright infringement.
Also, I don't care how legal it is, but if you are a big social media megacorp and start charging money or putting restrictions on "your" training data, I hope you get a special place in hell. Thinking of Reddit, Twitter, DeviantArt and Microsoft here.
I can't think of any justification that isn't emotional to say that AI generated art isn't "art". Writing a prompt is I guess technically an artistic expression. However, I wouldn't classify it as "digital art"; it's its own thing. To me, it'd be like taking a photograph, and then claiming you painted it. It should be marked as "AI generated", and categorized separately.
Ultimately, IMO it shouldn't be banned across the entire fandom (if such a thing were possible), but I would not feel comfortable interacting with a place where AI material was not marked as such, or a place where they reserve the right to use your media for training. Even then, if I were given a choice in a place where I frequented, I would say "no" to allowing AI art.
Personally, I'm very negative about AI generated media, and try to avoid it where possible. Maybe that's because I'm old and grumpy, or maybe it's just because I'm scared of it. But I wouldn't willingly use these tools for anything creative I'm involved in.
I think this AI thing has effects on all of society, not just the furry fandom. The proliferation of AI has its tendrils on many places of society at the moment, and could have drastic changes on many of the ways we live our life. And, of course, nobody really has any idea on how to deal with this, and it's controlled basically by megacorps. I could go into detail about this, but it's... A bit bleak.
This is unironically one of the things that gives me a lot of anxiety and stress.
To answer the original question though, I think in its current form and without massive societal change, not only the furry fandom but society in general would be strongly negatively impacted.
... Of course, I did get a bit curious, and maaay have asked ChatGPT to clean up a first draft of this post, just to see how it all works. I wonder how good people are at noticing things like that?
I agree with pretty much everything you said. I think Ai art is fine as long as it's properly marked as such and the training data can be traced back to the original creater and their consent was given to use their work for Ai purposes.
Once the honeymoon is over I believe Ai will become just another tool that can be used to speed up different tasks. Some will try to fully automate processes with Ai, but especially in art I don't think it will ever come close to replace stuff created by humans. (At least I hope so)
I think AI generated images (as all AI technology used in different areas) would be best used as an extension for an artist's workflow as opposed to a total replacement of the artist. AI can speed up the creative process in multiple ways. I personally occasionally use AI generated images to get inspiration for things like poses, hairstyles, clothing and colours. The middle two especially, since I'm really bad at coming up with stuff sometimes. XD
I don't mind if users post AI generated images as long as they don't do it in a deceptive way, like actively hiding the fact that the image is AI generated, or worse, claiming that they made it. The argument of "I wrote the prompt" doesn't really work because you giving specifications for a commission or a request for an artist doesn't magically mean you made the art that artist creates either. Better yet, some kind of disclaimer about it being AI art would be good so people who do not want to see it know to ignore it or filter it out (if possible).
I'm not as doom and gloom about AI art as some people seem to be. Probably because I'm both a techie and an artist. As long as AI cannot reliably combine separate, possibly never-before-combined contexts in a creative manner, I am not going to be too worried about it becoming a replacement for human creativity. Though I do see that it could be used to amazing gains by someone who knows a thing or two about creating art.
I think it’s most reasonable to compare AI use in the production of art to something like a Camera Lucida or photography itself as an art, because there are many levels of which AI could be involved in the production of art.
Is it reasonable to exclude such in a situation like an exhibition/contest of drawing skill? Yes, absolutely.
Is it reasonable to exclude them in other situations, like a general furry space where you’re allowing non-artist commissioners to upload art they had no involvement in other than money and prompting? No.
I believe that the fandom is better off for more of its members being able to express themselves and their ideas visually. Some people hold the view that a non-artist should have to pay an artist or become one in order to realize their characters and ideas. I think being discriminatory about what tools people are allowed to pick up should they wish to become an artist is not fair.
One last angle that I really wish wasn’t a thing, but I feel I have to bring up: 3D art using existing models and AI art are among the last-ditch pressure reliefs to facilitate visual expression of ideas that are unpopular. No artist should have to draw subject matter they don’t want to, but many are bullied out of drawing things like feral or feral/anthro art even though they might identify with or want to serve those interests. While in an ideal world that might be partially compensated for by some of the braver members of the fandom learning how to draw/paint/etc for themselves in order to fill those niches, it isn’t enough. Intersectional discrimination against those who choose to learn how to use AI tools to create art has a particularly disproportionate impact on unpopular demographics within the fandom.
Opinion: I personally have no problems with AI art. The methods AI uses to draw are almost verbatim the same techniques humans have used for centuries to learn the techniques or even copy the techniques of the masters that came before them. The difference is that the AI can learn at a much quicker pace due to the models it has to work from.
I have no problems with this. Is AI perfect? Absolutely not. It still requires, besides a prompt, human intervention (AKA inpainting) to fix the flaws of the piece the AI has created to give it the perfection cannot due the first time. A flaw most artists also suffer from (some fix, some don't.)
The good thing about AI art is that it isn't copyrightable. As proven with the AI comic Zarya of the Dawn.
AI's are not humans. Meaning the works they create are in the public domain. And companies that want to use it, will not care enough to even pay someone to perform the inpainting needed to fix the extra limbs flaw that AI produces now. So such games like bloodlines, which uses AI and does not do inpainting, can easily be spotted and avoided because you can easily spot the flaws of AI art in it.
Now I do feel their is the problem that people could try and pass AI art as an artists actual work. Which is a legitimate concern. But AI companies like Stability AI or individuals building their own AI software could easily resolve this, by including in the software a watermark that is placed in the bottom right or top right corner and shows either the company logo and generated "DD/MM/YYYY" or clearly states "AI generated art, Generated on DD/MM/YYYY"
This way it can easily help separate the fakes from originals. Only way for anyone to fake an AI work as an original after this is to manually go in and edit out the watermark. Which a more talented individual can do, but why bother if they got the talent to remove it?
On the point of AI writing, I don't think anyone has much to actually fear. AI writing is still pretty bad even by modern and ancient author standards. The wiriting can sound convincingly human, but their is still nuance in what is written by humans that AI still cannot capture and easy to spot. Emotions are a good example of this. What AI writes sounds wooden and stilted and when it tries to include something like a joke, it falls flat. This is because it doesn't understand what makes a joke funny. It can copy something someone else has written that is funny but when it attempts to create something in that style and keep the humor, it doesn't work because again, AI does not understand this.
AI music is another thing, after having listened to AI neco arc, Freddie Mercury, Frank Sinatra, and Johnny Cash, the only two AI even got anywhere close to good was Cash and Sinatra.
But then the people who generated those carefully selected titles that would work well with the voice models they had on hand. And probably did extra work to perfect and make Cash and Sinatra's performance better.
Where as Mercury and Neco Arc suffer from AI with a very limited models and their is garbling within the audio leaving some lyrics unintelligible. And Freddie in particular has moments where it sounds like him then moments where it sounds like someone playing around in audacity to make the original singer mimic Mercury.
Should AI Art be allowed in furry spaces?: Yes it should. However I do believe it must be limited down to a single Community or thread if this were an image board, fully dedicated to AI imagery. None of the AI art should leak out into the other community or threads.
This way it doesn't overflow to all the other communities or threads dedicated to work created by actual artists.
Use cases: the actual AI creating generated art? Not so much. The inpainting tool however? I feel that is a game changer for lazy artists. It can be used to fix certain flaws that exist in their own works. Like say they drew a left foot, but meant to draw a right foot, the inpainting tool and with a model built around their own art style, can be used to allow AI to redraw the foot correctly. Fixing the artists actual mistake in the piece.
Long-term impact?: Their won't be much of a long term impact. AI is nice right now, but among furries, we are going to use and abuse it to the point we will burn ourselves out on it. It will drive many back to having greater appreciation for actual art made by real artists and not a machine. Simply because the machine has oversaturated the fandom with it's knock off art.
People have created knock off artwork for years that mimic a masters style. But people don't want art that mimics a masters style. They want an original piece from a well known artist or atleast a print of a well known artists work.
I have a simple shoe analogy that sums this up quite well "People want Nike Jordans, not Mike Jongdonz"
Granted, I don't know how popular Jordan's are in Europe or the UK, I know in the US they are quite expensive quality basketball shoes, and if never worn, retain their value. And prices for them simply go up because they are collectable.
Whereas if you have a knock off like Mike Jongdonz, they are not as well made, they contain no value and are simply their to rip off the real brand. Not to mention you'll be thoroughly embarrassed when people point out what you got are not real Jordans.
The same will be true of furry artists and AI art. Furry artists are the real deal and AI is the cheap knock off.
The knock off can be neat and interesting, but it will never compare to the real deal.
For me AI generated images are double-edged sword. As a non-artist I find that these are an amazing tool that allows me to create the specific type of Yiff that I really enjoy. However, I can definitely empathize with the concerns voiced by artists.
To be honest, I have to say that I don't believe that learning or training from publicly available artwork is immoral, especially if it's not for profit, I am worried that many of the popular models seem to be clearly 'over-fitted'. Over-fitting happens when you train a model so much that it ends up reproducing content that seems almost copy/pasted instead of merely being able to reproduce certain concepts present in the prompt.
The second concern I have is related to mediocrity and saturation of AI imagery. I believe that in upcoming years many of us will grow tired of seeing AI imagery as a replacement of stock images. The truth is that most images generated aren't really that good and people will quickly grow tired of seeing them pop up everywhere. It will become a sign of mediocrity or poor quality.
However, I also think there's a silver lining: AI images have the chance to drastically increase the amount of visual support that any type of content has. This could lead to us growing into a society in which images and other forms of visual representation such as mascots might become more and more popular (as it is in Japan, for example). And, as a consequence, this could increase the demand of quality artwork produced by real artists.
In regards to the furry fandom, I think it's a good idea to separate AI images from artist-drawn artwork. Not only because of the controversy, but also because of the potential for massive amount of low-quality an low-effort content.
I do see potential to use these image generators to quickly try out concepts and scribbles when i have no idea how something should look, but try out how it might look
Then later i could give it to the real artists as the refsheet
I simply don't think images created by AI count as art. Art requires meaning, emotion and soul. These things are not possessed by AI nor are they imparted onto things created by AI. They create images. Images are no replacement for actual art.
I do think it will cut down on commissions in many places, though. Like porn and generic fursona reference sheets/images would most certainly be impacted negatively. Though I don't think it will be devastating.
One additional thing to consider is that not all art is either entirely AI-generated or entirely devoid of AI-generated material. Rick Griffin, for example, has used AI assistance to fill in darker parts of the background for one of his pieces before, falling afoul of Fur Affinity's policies. This was a work in which the vast majority was regular digital art created by Rick Griffin himself.
It's definitely useful for exploring ideas, I recently used AI-generated images as reference for an artist and it helped me get my thoughts across. I can't wait to see the final result of it!
It's probably also good for adding illustration to text where someone otherwise wouldn't be able to afford it or spend the time, adding detail to your own art and so on.
The worrying thing to me is how fast it can create all sorts of images, and combined with LLMs and other tools even be automated. I could totally see entire feeds being populated by personalized AI art, stories, and music, let's not forget that's largely dictated by algorithms already if someone uses things like Spotify.
So one day it might get extremely difficult to tell if what you're looking at was even created by someone on their own, or if AI had any significant role in it. This would make it impossible for regular artists to compete under the current economic model, which to be fair needs to go, but it will still cause suffering in the meantime. Even beyond capitalism, people want share their work for others to appreciate, and if every channel of communication is flooded it's going to be pretty difficult to get noticed among the noise. If AI can create stunning images it might also depreciate art in the eyes of others, because to get to a very similar result using AI could take a lot less effort if the technology further improves, so they might see it as just a commodity or "pretty picture" and not think about artists at all.
Its a fun thing to use for visualizing new ideas I dont think it should be taken that seriously. The stuff is here whether we like it or not, I think there should be a specific section in all places for AI art. I dont see it as a threat at all.
I think AI art is best when its a collaboration between human and AI. Look at Stelfie and how his workflow involves planning, photoshop, iteration, and intent of the artist to create the final work. Absolutely this is a form of art. However, I don't actually place a ton of weight onto that word. A child's crayon scribble is art too, but who really cares.
AI models that have been trained on public domain work is legally problematic, yep for sure,I think calling the OUTPUT plagiarism is a bit of a stretch but I'm not going to defend that position here because ultimately the right thing is to move towards models that have been trained with consent (which Adobe has at least made steps to do)
I think the legal outcome that AI art is not copyrightable unless substantial transformative work has been done to/with it is a pretty good legal take and makes it limiting on how it will effect traditional artists.
A lot of AI art can be classified as "low effort" and USUALLY you can tell, because it'll have artifacts, flaws, and issues that the person hasn't bothered to edit with in-painting, or make a manual edit that they then use AI to refine. It does not require a ton of skill, no, but it can require some skill and some effort to do well. I think these low-effort submissions should be looked down on and deterred.
There absolutely is the danger that we will get flooded with low effort AI art, though the community backlash against it has seemed to largely stifle that reality. I think people SHOULD be careful to post something unless they have actually spent effort in refining or at least very heavily cherry picking their generations and spamming it should not be allowed. I think people also need to be TRANSPARENT they use AI tools, but that ultimately, it should be something that people are ALLOWED to use (in the right circumstances, with transparency), without immediate derision.
Well, I think it’s not how it’s made. It’s how much time you put into it that matters. If someone used AI as a base, vs someone who made it from scratch, but they both took 8 hours and were the same amounts of effort, is it really that different? not defending anyone here, just adding my perspective.
I've heard a lot of perspectives ranging from it's evil, to it's great to kind of... Concerned silence?
I'm honestly surprised how fast it came to the point that ai can generate some pretty good art on its own? I feel it kind of lowers the point of entry to simply obtaining art but also I'm left puzzled on whether it negatively impacts some artists? I can appreciate some artists having concerns someone would just go to an ai of some type and get a picture there of their fursona instead of commissioning an artist but I wonder if that would ever realistically happen considering the detail needed for some sonas. However with the various tools available... I do think this is a legitimate valid concern.
I'm on more patreons than I should be, and some are honestly worried what it will do to their ability to earn an income. I'd say there are other factors at play in general (basic cost, quality, communication, competition, etc.) But in the end an ai becomes another player/factor in the challenges that some artists face. Enter the idea of ais trained off specific artists to the point where you can get 85% similar style to that artist... For some people why wouldn't you just use the ai? It would be faster, cheaper, and if something is wrong you can just have it try again.
Also considering some artists just use mainly templates for what they do (telegram stickers mainly) I figure that might be an area where an ai would shine.
I'd say for me in the end it depends on which perspective you're going to and how long if a gaze do you want to take. For an artist, I would be terrified. Ai can kind of hold your feet to the fire constantly to churn out great product but also potentially serve as just data for an ai to learn off of. For a consumer I would be somewhat excited. Ai can provide a cheap source for art that would otherwise cost 100s and take weeks :/