Out pours a brilliantly detailed photograph of une vieille grande dame that could easily be someone’s stately grandmother looking sharp in a bowling jacket. Everything from the wisps of hair framing her face to the asymmetrical wrinkles and puffs in the folds of skin around her eyes ring familiar. Her lips are unsmiling and her eyes have the direct gaze common among people not conversant with cameras. It is a compelling and inexorably human portrait with one important distinction — the woman does not exist.
She is the creation of an AI algorithm.
I hesitate to use the word imagination. Yet, looking at the hyper-realistic portraiture of some of AI-generated art, one can’t help but wonder if we are indeed doing the AI a disservice by curtly negating the possibility of any artistic intent. It is not limited to photographic images. AI artworks often have a weird, uncanny, even absurd element about them that artists have explored to fit genres such as horror and sci-fi.
Increasingly sophisticated AI engines are now capable of producing still images, videos, music videos, and even a barely playable video game from something as simple as text prompts, other images, or video. AI art is already being used by mainstream publications such as The Atlantic, which featured a viral AI-generated caricature of Alex Jones. Even TikTok now offers a text-prompt AI-greenscreen for its users. We are left wondering what the ubiquity of AI art could mean for how we perceive artistry, ownership/property values, and narrative meanings associated with traditional art.
Where Does Alex Jones Go From Here? https://t.co/fx4v4cgOTo
— Terry Dresbach (@draiochta14) August 10, 2022
Who Owns the Art and Who Gets Paid?
Polish digital artist Greg Rutkowski recently spoke out against using his name as a prompt in AI engines. Though the engines often miss out on the intended interpretation of text prompts or throw up bizarrely irrelevant details in the art, AI’s capability of recreating stylistic influences of artists, both living and dead, have raised concerns across the board. Artists are concerned their original work will get sidelined by search engines as millions of AI-generated lookalikes flood results.
Many fear that generative AI engines such as DALL-E, Imagen, Crayon, and Midjourney may be exploited by corporations to avoid paying human artists and illustrators. Conversely, it is helping many solo artists and entrepreneurs save significant time and money in creating anything from boardgames, book illustrations, and cover art to branding paraphernalia.
Swedish artist Simon Stålenhag expressed concern about the future of original art in a world flooded by cheap derivatives. Others like AI artist Jason Allen, who recently won an award for his digitally manipulated artwork, contend that those discounting AI art are foregoing the effort involved in fine-tuning the text prompt and envisioning the final product.
— Best MidJourneyAI Generated Art🤖🎨 (@MidJourneyAI_) November 17, 2022
Ethically, we are in a gray area with many arguing that AI artwork, especially those created using particular stylistic influences, are copyright-free. The US Copyright Office has already decreed that “human authorship is a prerequisite to copyright protection,” prompting an appeal at federal court. It is unclear whether the AI developer or the user creating the artwork through prompts should have ownership. Some platforms such as OpenAI explicitly state that images created with DALL-E 2 belong to them and customers need to obtain a license to use them commercially.
Potential for Abuse and Bracing for a Legal Storm
The problem with AI engines infringing on existing copyrights is that some have been specifically designed to erase or copy artist signatures. It is also unclear how far AI companies are willing to remove individual artists from their training dataset at their request. The Have I Been Trained tool lets artists check if their artwork is being used in AI training models but there is no clear process for removal petitions or a timeframe.
Anticipating potential litigation from bigger players, Open AI has already removed some results from its image training set, including celebrity images and Disney/Marvel characters. But the sheer ease and scale of AI image generation makes it impossible to anticipate the range of fallouts, including creating celebrity and machine-generated porn, deep fakes, fake people databases, political disinformation, revenge porn, and more.
DALL-E 2 has created policies preventing users from generating images with contents such as hate, harassment, violence, sex, or nudity. But the range of the obscure data sets are only starting to become obvious as was revealed in the case of an AI artist finding her private medical records in the LION-5B dataset.
What Happens to the Power of Art?
For centuries, the function of art in society has been preserved by the passion for ownership. Art movements such as Abstract Expressionism, L’Art Brut, Pop Art, Auto-Destructive Art, Neo-Dada, and artists such as Francis Bacon and Marcel Duchamp have tried to move beyond the confines of art as property and break its veneer of desirability and value. Newer movements such as Kinetic art have sought to democratize the experience of art by focusing on what the artwork does and not its corporeality.
Open to all environmental influences, the idea is simply to stimulate the spectator and let their visual perception carry through the experience of the artwork.
But all such actions and reactions are the result of a time where the historical meaning of the artwork was contained within its physical existence. With AI art, that meaning, or the locus of the aura of the artwork, is only signaled at with layers of modeling from millions of randomized sources. Does it democratize the experience of art as many artists have intended for years or does it obliterate it?
As far back as 1936, Walter Benjamin had already highlighted that “the adjustment of reality to the masses and of the masses to reality is a process of unlimited scope, as much for thinking as for perception.”
It’s been four weeks that I ask DALL·E and Stable Diffusion to generate art for my newsletter and I really like the results 💖 pic.twitter.com/dLJwJbavhM
— Anne-Laure Le Cunff (@anthilemoon) November 17, 2022
Is AI Art Capable of Meaning?
In his seminal work, “The Work of Art in the Age of Mechanical Reproduction,” Benjamin probed the decay of the aura in art that is subject to mechanical reproduction. He defined aura as a unique function of distance in relation to the person experiencing it. The shadow of a tree branch or the distant view of a mountain contain an aura for the person experiencing it. Benjamin argued that the aura of the original artwork that comprised its unique presence in time and space is lost in mechanical reproduction. This means an erasure of its entire history.
At a time when nearly everyone interacts with the world through a screen, AI images might become ubiquitous. An AI artwork generated with thoughtful, highly detailed prompts that carefully follow the style of a particular artist neither possesses artistic intent nor historical meaning. At best, it serves as a simulacrum of the original work.
MORE FOR YOU
As philosopher Jean Baudrillard pointed out: “Simulation … is no longer a question of imitation, nor duplication, nor even parody. It is a question of substituting the signs of the real for the real.”
The Marxist art critic John Berger believed that the most important function of art is to lead us back from the work to the essential artistic process. With the artistic process now eliminated or at least heavily aided and abetted by artificial image synthesis of all art ever created, our ability to create the world we want just became potentially limitless, or endlessly derivative, depending on how we use the tool. Our reality and visual perception will adjust and evolve accordingly.