Authors:
Daria Loi
Human intention is the issue, not the AI.
— Refik Anadol [1]
Last year, on one of the social media accounts of the art gallery I run, I reposted an illustration that I found aesthetically inspiring. Within minutes, an artist in my network reached out to share their disappointment with seeing "AI art being supported through the gallery." Truth be told, I didn't notice the image was AI generated and my reposting had nothing to do with my or my gallery's take on AI art. The exchange sparked a healthy debate, and as soon as my husband asked, "What is your take on this topic?" it was clear that I had to write this column.
Undoubtedly, there are a number of valuable uses for AI in the art space, from the recovery of lost cultural heritage and enabling scholars via the cross-referencing of datasets, to the provision of more neutral art price estimates with tools like SANG.art and the benefits of AI-powered productivity tools in the commercial art world. That said, though it is important we recognize that AI art has been around for quite some time, it is equally important we recognize that large language models like ChatGPT and image generators like Midjourney, DALL-E, and Stable Diffusion are challenging the meaning of art and the role of artists. As such, the art world's reaction to AI has been and continues to be…complicated and highly inflammable.
The challenges posed by these technologies became mainstream knowledge in 2022, when Jason M. Allen submitted an AI artwork to the Colorado State Fair annual art competition—and won. Then, in early 2023, illustrators Sarah Andersen, Kelly McKernan, and Karla Ortiz filed suit in the Northern District of California against Midjourney Inc., Deviant Art Inc., and Stability A.I. Ltd. because of how their tools violate artists' rights.
Since then, the art world appears to be divided into four groups: pro, against, undecided, and blissfully ignorant, with the first two dominating the debate. Let's focus on the opposing, complex views of these two groups to tease out the key tensions.
In early 2024 on Instagram, art magazine Hi-Fructose posted a commentary [2] on the Midjourney database that is evidence in a lawsuit against the start-up. A read through the numerous responses to the magazine post provides the perfect glimpse of how divided and triggered the art world feels—and how complex and multilayered the issue is.
Mirroring a mainstream view of AI as something that will replace human labor, the fear of creativity and human artistry being replaced is a definite reason behind how incendiary the matter is in the eyes of the "against" crowd. Intertwined with that is a legitimate fear of artists not being in a position to make a living unless they leverage a technology they may have little interest in or that they perceive as cheapening their artistry.
Large language models like ChatGPT and image generators like Midjourney, DALL-E, and Stable Diffusion are challenging the meaning of art and the role of artists.
Another concern is quality and what is felt to be the deprofessionalization of art and the distraction that these tools are bringing to traditional art. After all, with these technologies any amateur can now create art-like imagery, do so very quickly, and then sell cheap prints via multiple online channels. Social media already abounds with shops built on this business model, with photorealistic and complex artworks that appear to be the output of skilled and trained artists yet were created in matter of seconds by typing prompts into a platform.
There is no such thing as an "AI artist", you have to be able to create to be an artist…typing prompts into an AI engine to generate images just makes you a lazy poser.
—@technicolor_creep in [2]
Holy cow, the amount of people in the comments taking the side of an algorithm spewing half-assed images instead of spending their time improving their traditional skills is depressing.
—@xul1349 in [2]
Then there is the issue of attribution. All these platforms are built by scraping massive amounts of images that are freely available—that is, viewable and easily downloadable. What is ingested by these platforms frequently includes images that artists made available to promote themselves and sell their work—artworks that will, however, help train algorithms that competitors may use, hence cannibalizing their market [3]. The attribution issue is a very delicate and complex one, since it includes intricate copyright laws across geographies as well as diverse cultural habits, perceptions, and expectations in relation to notions of copying, leveraging, referencing, collaging, and plagiarizing. Entwined with the attribution problem is an issue that relates to the perceived motivations behind these tools and the strong feeling that these creative platforms were created with one goal in mind only: profits that will benefit large tech players.
Great tool but no one should be opted in without consent to being accessed by the database.
—@smitten_soph in [2]
AI art is not art. AI art is an attempt by a group of tech companies to subvert actual artistic talent for profit. Flag all AI art as copyright infringement when you see it online. AI art is theft.
—@shadowsandstrobes in [2]
Then there are numerous issues related to AI bias creeping into artworks, including colonial [4], racial [5], gender, and demographic biases that inform the platform and dataset and, consequently, are potentially foundational to whatever artistic output these platforms will generate.
I was struck by how many more men's names there seem to be on this list than women's. Also anglo/American and Japanese I guess rather than many other examples, like how many Matts and Daves are there? Fair enough for styles from earlier eras, but for contemporary illustration does this reflect the talent pool, or are the artists getting picked all dudebros making slick images with big tits, reflecting the taste of the dudebros using the tool?
—@mywordsfly in [2]
Another concern relates to the use of personal user-generated content in training algorithms, as that content may include moments of intimacy that can be leveraged by others for very different purposes and even to extract monetary value.
In Salon of 1859, first published in the Révue Française (Paris, June 10–July 20, 1859), French poet and art critic Charles Baudelaire writes, "This industry, by invading the territories of art, has become art's most mortal enemy." His concerns with the latest technology of the time—photography—were shared by many who feared the new technology would be the end of painting. What we now know is that photography not only did not kill art but in fact triggered an unprecedented number of new creative and exploratory endeavors to depict real and imagined worlds, many of which were created using traditional painting methods.
This historical account shows what may occur when a novel and complex technology enters society, becoming mainstream: One may feel suddenly and overwhelmingly scared—uncomfortable with what we perceive as an imminent shift in how things are normally done. Some may opt to voice their dissent as loud as possible, and others may even opt to explore ways to sabotage the new technology and how it may affect their domain (e.g., art).
Everyone interested should look into the Nightshade program…Nightshade will make the AI think it's taking an image of a cat but will actually be using an image of a snake, making the AI useless. Good stuff.
—@zillagrim in [2]
On the other hand, history has also shown us that one may feel empowered, inspired to freely explore and appropriate the new tool with no limitations, when being on the cutting edge of creating futures is a primary focus. In some cases, one may dismiss or even ridicule naysayers' perspectives, pointing out their narrow-minded attitude and showing why the new technology is as benevolent as prior ones.
Every artist has directly copied other artists at some point. Nothing is original, this is just a tool. At the end of the day there is a human telling it what to do. You won't win this fight….You have the ability to create, nurture that.
—@killmek8 in [2]
Every single human artist uses other artists' work they've seen, either consciously or subconsciously when creating their own art. This is no different.
—@zobear2000 in [2]
Hot take: AI machine learning is doing the same as every artist ever; taking from known sources and interpreting it into their work…I understand that, now, capitalism makes us believe that we should own and make the most profit with our work, but the art world has always been incestuous in its learning and creative process.
—@devon.gold.art in [2].
As a monk who has been hand lettering bibles for years, I have spent a lot of time perfecting my traditional skills. These printing presses are blatantly stealing my letter forms. I have created a list of the monks whose work has been copied by this new, obviously "flash in the pan" technology…I am filing a class action lawsuit to stop the wheels of progress. Please support me in my quest!
—@johnlearns2art in [2].
While I strongly believe that all current reactions are human and reasonable, I am somewhat more interested in exploring the interstitial space that exists in the midst of such extremes, as that is where we may find deeper answers to what seems to be the most pressing question: How can art and AI coexist meaningfully, sustainably, and equitably?
I am still learning and deepening my perspective on these complex themes; however, I have isolated a few views that I hope will bring value to those interested in this incendiary topic.
As a practicing artist and gallery owner, I choose to channel my energy into learning before forming inflammatory or blindly optimistic perspectives about fellow artists who are exploring these platforms. We urgently need a healthy, educated discourse on this matter.
There's an important conversation to be had about how models are trained, copyright, permission, etc., but it serves no one to dismiss AI tools and artists using them as "image hungry typists.
—@neekosette in [2]
I propose artists treat these technologies as they would any other medium (i.e., learn how they practically function to see whether there is or isn't value for their craft—and why). In return, AI technologists working to create tools to benefit the art world should learn what art is and, crucially, what artists do when they "do art" (i.e., educate themselves about their users). Also, they should find a better way to explain how these platforms actually work, without jargon or judgments (i.e., communicate to enable folks who live outside the walls of tech bro clubs).
I wonder how many people commenting have even attempted to understand the process of text-to-image generative deep learning…The technology doesn't collage and it doesn't replicate…You look at the output and assume you can infer the process! That's like saying you can just look at the Large Hadron Collider and just "know" how it works, it's laughable.
—@darkling_hearts in [2]
Deep insights into how these technologies operate provide unique opportunities for the art world to appreciate and evaluate the nuances that exist in these systems, and if/where we should draw some hard lines, without compromises or exceptions. For example, blindly rejecting these platforms based on a one-size-fits-all perception that they "use data without consent" is myopic, while questioning how and why datasets are created is not. Similarly, discounting our discomfort or rage when our data has been used in ways that do not reflect our understanding of what we consented to is shortsighted, and, frankly, demonstrates a lack of empathy.
AI technologists working to create tools to benefit the art world should learn what art is and, crucially, what artists do when they "do art."
Another aspect to consider is time. While all can benefit from well-articulated codes of conduct and laws describing what plagiarism looks like in academic environments and what copyright infringement means in a pre-AI world, regulation surrounding these new technologies is, at best, in its infancy at this point.
So, how can art and AI coexist meaningfully, sustainably, and equitably?
A number of interesting answers can be discovered in the work of artists who are seriously engaging with these platforms and their craft. I am thinking about Sougwen Chung's collaborative paintings with AI-guided robots, Mario Klingemann's portraits, Sofia Crespo and her imaginary sea creatures, Sara Ludy's edges, Ivona Tau's curations, Anna Ridler's narratives, and Claire Silver's models' explorations.
My favorite examples of artist-AI collaborations come from the work of Refik Anadol, who describes the purpose of AI as "finding the language of humanity by using collective memories to create collective dreams and eventually collective consciousness" [6]. In his famous Unsupervised artwork, the artist trained a machine learning model on the MOMA database collection, which includes 200 years of artworks. Watching Anadol explain how he created the piece and maneuvering the tools of his craft is a mesmerizing experience. Venturing way beyond the act of leveraging AI, he created new ways of making and interpreting what art means and could become—and he does that by leveraging datasets that his team creates or data provided by its rightful owners. While Anadol refers to data and his thinking brush as his two primary tools, he is clear on the role of technology: "I don't think that the machine is thinking; it's helping to think in a different way…We should think of AI as a mirror, a mirror that will reflect whatever we are training…it is on our hands to train this mind to dream what we want" [1].
While I have no doubts that these new platforms will continue shifting, morphing, and advancing our understanding of art, a full prediction of how art and artistic practice will look like in the future is excitingly impossible. That said, I welcome millions of educated explorers like Refik Anadol and would prefer a decrease in irrationally anxious artists, superficial amateurs, and art techno-exploiters.
Unlike Baudelaire, I do not believe that AI, by invading the territories of art, has become "art's most mortal enemy." That said, the space between AI and art is still far from being meaningful, sustainable, and equitable. Therefore, I applaud all those who are cautiously experimenting, fearfully exploring, open-mindedly learning, and respectfully challenging, as they will help advance our collective understanding of what it means to be an artist.
References
1. PBS NewsHour. Use of Artificial intelligence generates questions about the future of art. CANVAS. Mar. 2, 2023; https://www.youtube.com/watch?v=2HYemjTSxp8
2. Hi-Fructose Magazine. Instagram post. Jan. 6, 2024; https://www.instagram.com/p/C1x_p5SRLie/?igsh=MWxkemJid3phMjd5aA%3D%3D
3. Roose, K. An A.I.-generated picture won an art prize. artists aren't happy. The New York Times. Sep. 2, 2002; https://www.nytimes.com/2022/09/02/technology/ai-artificial-intelligence-artists.html
4. Dhanesha, N. AI art looks way too European. Vox. Oct. 19, 2022; https://www.vox.com/recode/23405149/ai-art-dall-e-colonialism-artificial-intelligence
5. Murphy, H.T. Artificial intelligence is dreaming up a very white world. Slate. Feb. 8, 2023; https://slate.com/technology/2023/02/dalle2-stable-diffusion-ai-art-race-bias.html
6. The Museum of Modern Art. AI Art: How artists are using and confronting machine learning | How to see like a machine. Mar. 15, 2023; https://www.youtube.com/watch?v=G2XdZIC3AM8
Author
Daria Loi combines design strategy with experience research and innovation to enrich people's lives and humanize technology. She is chief experience officer at Astral and owner/founder of Imperfecta, an art and design gallery focused on women artists and minority creatives. Loi serves on the DemocracyLab board of directors, and is an honorary professor at the University of Newcastle, Australia. [email protected]
Copyright held by author
The Digital Library is published by the Association for Computing Machinery. Copyright © 2024 ACM, Inc.
Post Comment
No Comments Found