Online Art Communities Begin Banning AI-Generated Images

As AI-generated art platforms like DALL-E 2, Midjourney, and Stable Diffusion explode in popularity, online communities devoted to sharing human-generated art are forced to make a decision: should AI art be allowed?

Collage of dozens of images made with Stable Diffusion, indexed by Lexica

On Sunday, popular furry art community Fur Affinity announced that AI-generated art was not allowed because it “lacked artistic merit.” (In July, one AI furry porn generator was uploading one image every 40 seconds before it was banned.) Their new guidelines are very clear:

Content created by artificial intelligence is not allowed on Fur Affinity.

AI and machine learning applications (DALL-E, Craiyon) sample other artists’ work to create content. That content generated can reference hundreds, even thousands of pieces of work from other artists to create derivative images.

Our goal is to support artists and their content. We don’t believe it’s in our community’s best interests to allow AI generated content on the site.

Last year, the 27-year-old art/animation portal Newgrounds banned images made with Artbreeder, a tool for “breeding” GAN-generated art. Late last month, Newgrounds rewrote their guidelines to explicitly disallow images generated by new generation of AI art platforms:

AI-generated art is not allowed in the Art Portal. This includes using tools such as Midjourney, Dall-E, and Craiyon, in addition fractal generators and websites like ArtBreeder, where the user selects two images and they are combined into a new image via machine learning.

There are cases where some use of AI is ok, for example if you are primarily showcasing your character art but use an AI-generated background. In these cases, please note any elements where AI was used so that it is clear to users and moderators.

Tracing and coloring over AI-generated art is something best shared on your blog, as it is much like tracing over someone else’s art.

Bottom line: We want to keep the focus on art made by people and not have the Art Portal flooded with computer-generated art.

It’s not just long-running online communities: InkBlot is a budding art platform funded on Kickstarter in 2021 that went into open beta just this week. They’ve already taken a “no tolerance” policy against AI art, and updating their terms of service to exclude it.


Platforms that haven’t taken a stand are now facing public pressure to clarify their policies.

DeviantArt is one of the most popular online art communities, and increasingly, members are complaining that their feeds are getting flooded with AI-generated art. One of the most popular threads in their forums right now asks the staff to “combat AI art” by limiting daily uploads, either by segregating it under a special category or to ban it entirely.

ArtStation has also been quiet as AI-generated images grow in popularity there. “Trending on ArtStation” is one of the most popular prompts for AI art because of the particular aesthetic and quality of work found there, which nudges the AI to generate work scraped from it, leading to a future ouroboros where AI models will be trained on AI-generated art found there.


However you feel about the ethics of AI art, online art communities are facing a very real problem of scale: AI art can be created orders of magnitude faster than traditional human-made art. A powerful GPU can generate thousands of images an hour, even while you sleep.

Lexica, a search engine that solely indexed images from Stable Diffusion’s beta tests in Discord, has over 10 million images in it. It would take a lifetime to explore everything in it, a corpus made by a relatively small group of beta testers in a few weeks.

Left unchecked, it’s not hard to imagine AI art crowding out illustrations that took days or weeks for someone to make.

To keep their communities active, community admins and moderators will have to decide what to do with AI art: allow it, segregate it, or ban it entirely.

Comments

    One could argue that if you are an artist, you have looked at a lot of art to be inspired and base your style on what inspired you. Is it wrong that machines (AI) are faster at doing this?

    It’s a misnomer. Deep learning models are no more “inspired” to make art than Google is inspired to deliver relevant search results.

    Major false equivalency. Humans are inspired by (visual) art, music, emotions, relationships, life events, touches, tastes and smells, and a hundred other things.

    One would have to look at what AIs actually do. AIs are trained to contextual and replicate data. For example, if you train a GAN on only one image, eventually it will be able to recreate that image down to the pixels (or very close, depending on how strict the algo is). That’s what the end goal of these algorithms is: recreating image data in context. This is the fundamental difference in how computers and humans learn things. So human-machine comparisons are largely dubious. And, even if they were valid, there are many ethical rules (both written and unwritten) regarding how humans can use other peoples work. So if AI wants to be equivalent to humans, then it must also be held to similar ethical standards. The issue with AI stans is that they want the human metaphor without the human ethics.

    Of course, we can also reflect on the many changes observed by such art historians and critics such as Ernst Gombrich, observing that cave art gave way to painting stones and mixing novel minerals for colors. As technology advanced, inspiration for technique as well as subject matter evolved as well … and there were plenty of naysayers in the mix until the novelty became commonly accepted.

    AI can and does create art works that people appreciate but, for me, AI art is no more art than teledildonics is an intimate relationship.

    I am glad to see this! I love AI-generated art, but there are also so many talented human artists who don’t get any exposure for their hard work. It’s important to recognize them before allowing robots to take over. Nothing will ever replaces the emotions behind human-made works.

    The desire to ban is natural for marketplaces of human creativity. But this discussion buries the lede — it’s going to be harder and harder to tell AI art apart from human art, and at some point enforcement will become impossible, and I don’t know what a change to terms of service could do about this.

    Maybe a more fruitful long term discussion is “why are we sharing the hard work of humans, when computers can do it with ease”. Is art hoping to go the way of chess — where despite clear computer superiority, there is still room for humans to show ingenuity and competitiveness? Or is it going to go the way of graphic designers, whose jobs completely changed with the advent of desktop publishing tools, but where humans ultimately coexist with technology and produce more output faster? Or is it going to go the way of those rooms of typewriter secretaries, who got completely obsoleted?

    I think that trying to ban AI art is to prioritize creation over consumption. It elevates the process of human creations onto a pedestal where the mere humanity of the creator wins over the product. Some consumers — maybe most? — might not care. Perhaps communities like DeviantArt believe they have audience of creator-preferring users, but at some point in the future they run the risk of becoming a society of bitter out-of-work typewriter specialists.

    A healthier way forward is to embrace AI creation. You can’t pretend this technology doesn’t exist, and that it will not at some point create 95% of all visual imagery that we will see. If in the past we needed 5,000,000 human illustrators in the world to support society’s illustration needs, in the future we might need 250,000. Communities could embrace creating with AI tools as a discipline on the same continuum as pens, paintburshes, and photoshop — and help transition their users over to this new channel of coexistence with computers.

    “But as there is no mind in the photographic picture, so according to some it cannot contain any new idea, pose, light, or expression capable of representing impressions produced on the human mind, and ‘not being the work of man’ it must be, indirectly, the work of the devil–and, since as ‘the work of man is indirectly the work of God,’ as Mr. Sutton has it, where are we to go to?” — O.G. REJLANDER, 1863

    https://cool.culturalheritage.org/albumen/library/c19/rejlander.html

    I was once a member of a panel judging works in a painting competition. Several weeks later we discovered that one of the works to which we awarded a prize was simply a direct copy of an internet meme. This was before the advent of plausible AI generated art but even then it was impossible to keep track of all the needles buried in the limitless haystack that was the internet. Determining and banning AI generated art (AIGA?) seems a quixotic undertaking, noble but doomed to failure. As the previous contributor seems to imply, we have come to terms with the idea that photography can be art. Perhaps we may come to view “computer-assisted” art similarly and we’ll sort the weed from the chaff, as we’ve always done, with the assistance of our hearts and minds.

    Let’s say that we accept the inevitability of AI-assisted art. And let’s say that the sensitive, informed and creative use of prompts might enable artists to express the “hundred other things” that an earlier contributor suggested inspires humans and their non-AI art. There still exists the problem that an already overwhelmed system is being swamped by a new tsunami of digital art. And this further exacerbates a second dilemma – “How do we assess the quality of this flood of material?”.

    In the face of all this, I’m wondering if something like Rotten Tomatoes for digital art might be of assistance. In place of elitist, autocratic curation and censorship, we could have both a Critic’s and an Audience score together with some of the additional features that aggregation sites provide. For one thing, viewers’ feedback might help call out some of the worst cases of plagiarism and unoriginality (though we must remember that these “qualities” often make for popularity). It would. of course, not answer the concern raised by one of our fellow contributors, that there are “so many talented human artists who don’t get any exposure for their hard work”. We must remember, however, that this predicament has existed at least since the time of the Impressionists and the 19th century Paris Salon. An internet based-portal, on the other hand, is at least occasionally capable of turning up a gem even though it might, like America’s Got Talent or The Voice, do so despite itself. What you think?

    As a trained illustrator I feel ambivalent. I find it enormously fascinating to work with AI, but don’t feel anything close to the satisfaction from creating a piece of AI generated artwork that I get from creating something analogically. I think it’s naíve to believe that we can separate AI from all other digital art, but so far it can never replace an actual painting or sculpture etc. It’s a fun tool and the quality of the artwork is still a matter of artistic temperament. There’s a lot of garbage being churned out. Millions of images looking like Artgerm or Craig Mullins. So what sets good AI art apart from bad AI art will ultimately come down to the quality of ideas and the ability of the creator to sort the good from the bad. It’s all very new right now and people are still enjoying creating copies of popular anime art and concept art, but eventually I think that stuff will get tired and give way to something new entirely. At least, that’s my hope.

Leave a Reply

Your email address will not be published.