Summary
An exploration of AI's role in art, its creative potential, and its ethical implications.
In 2022 Jason Allen won a prize in the digital art category at the Colorado State Fair. The catch? The submitted piece was AI-generated. Following his controversial win, he also made a provocative statement: “Art is dead, dude. It’s over. A.I. won. Humans lost.” But is it? Will the rise of AI destroy, or revolutionalise, art? Do AI-generated images constitute art, or is it cheating? The answer is more nuanced than a simple yes or no. I doubt many artists, or even art teachers, can confidently define art in simple terms or without making exceptions for their exceptions – even less so what constitutes cheating.
When examining generative AI’s current role and limitations within art it is important to focus on visual arts such as painting and sculpture. AI might, very well, perform as an image generator or sculptor, in conjunction with a 3D printer. It is less likely, however, to threaten performance art – especially that of artists like Marina Abramović for whom the human body is quintessential.
AI has developed exponentially in recent years and may occasionally generate some impressive images. The technology becoming accessible to the general public has challenged traditional understandings of creativity as images and texts can now be produced in seconds. However, AI cannot be considered an artist in itself, and will never be eligible for that title. The most relevant inhibiting factors are also indicative of why generative AI is not, in its current state, a favourable medium. These include but are not limited to integrity, authorship, originality, and expression.
Integrity, in this context, signifies both a respect for history and principles. Art history has consistently been shaped by paradigm shifts and controversies, mirroring changes in other domains of human life. Each innovation has sparked disruptions to the status quo. Digital art, before AI, was responsible for one of these paradigm shifts during the 1960s which saw traditional art forms and new, shiny, devices come together. The rise of computer technology allowed artists to explore new mediums and boundaries, inadvertently causing disruptions to art discourse at the time; Questions were raised about the nature and value of art creation and consumption. While many traditional mediums remain popular today, digital art has solidified its place in modern art. I still prefer to work with acrylics and oils, but the world would look drastically different today without Adobe and Procreate. Disruptions often give way to enrichment, but whether or not this will be the case for AI remains to be seen.
Generative AI is a subcategory of AI technology which is programmed with training data, models and patterns, to generate digital objects, be they textual, visual, or sonic. Recent applications such as DALL-E and Midjourney have gained notoriety for allowing users to simply type a series of words and in turn receive images of any type or style. Human artists are, in response, understandably nervous and fearing their professions may become obsolete. The same concerns are certainly shared by writers and musicians whose work can, supposedly, be outshone by a machine. AI may erode public appreciation for the skill, labour, and thought that goes into creating meaningful art.
While AI may threaten the integrity of art it is important to remember that AI lacks integrity in itself. Many tech corporations pride themselves on their public image of sustainability, and the term ‘’the Cloud’’ may suggest something untethered, but technology has always been deeply rooted in the Earth and its exploitation. The Cloud, in actuality, lives in data centres. The construction of these centres requires immense amounts of labour and resources, including rare minerals. Post construction, the use and maintenance of these centres – which the function of AI depends on – consumes vast amounts of water and energy. Water that could better serve local communities and habitats, and electricity which may be renewable, or more likely generated from coal, gas, or nuclear energy. Few corporations will openly state either or.
One singular response from OpenAI’s ChatGPT consumes enough energy to power a lightbulb for 20 minutes and requires 16 ounces of water per 10 queries. Generating images is likely to consume much more resources than producing text. Jason Allen, upon first discovering Midjourney, admitted to creating hundreds of images before submitting one to the Colorado State Fair. That’s a minimum of 160 gallons of water, and more than 30 hours of powering that lightbulb, all for one final image that still had to be edited in Photoshop before submission. That’s just one person’s project – there are nearly 3,000 data centres in the U.S. alone handling millions of queries every day. The U.K., too, is set to become an ‘’AI superpower’’, with the government investing £14 billion in building data centres. The industry’s ecological footprint is already large enough to raise questions about how to live with AI, and ourselves for indulging in its convenience.
Authorship in art is intrinsically tied to craftsmanship, encompassing the skill, time, and effort invested in creating a work. The notion of authenticity derives from this connection, valuing the artist’s direct involvement. With the rise of generative AI, authorship may be reduced to simply typing a prompt for the computer to process, or it may become obsolete altogether. AI generates imagery from trained algorithms and based on the user’s given text – vastly different from the hands-on, skill-based, labour of human artists. A computer will never get paint under its nails. Generative AI removes the bulk of the creative process and challenges our understanding of authorship – but it is not the first to do so. In the early 20th century, Marcel Duchamp exhibited mass-produced objects after changing their identity by simply naming them. His infamous Fountain, a provocative pseudonym-signed urinal, questioned the value of craftsmanship by distancing art from the artist. Through his ‘‘readymades,’’ Duchamp shifted focus from the artist as a creator to the concept behind the artwork, paving the way for contemporary conceptual art.
Andy Warhol, only decades later, challenged the concept of authenticity. Warhol repeatedly admitted that his paintings were made by his assistants, under supervision but often without direct involvement or getting paint under his nails. His signature was, in fact, a stamp that could have been used by anyone in his employ – some of Warhol’s most famous works might have been created in his absence, effectively severing the link between art and artist. Despite these challenges, authorship and authenticity remain integral to how art is understood. Warhol eliminated the authenticity but retained authorship through the conceptual framework of his work. Duchamp rejected the authorship of objects but inadvertently claimed it through his ideas – it is in combination with his message that it becomes an artwork, his art. Authorship, within art, is a prerequisite. It can be challenged but not erased, as in the case of Duchamp’s Fountain.
In the case of AI-generated images, the absence of a human creator’s physical touch removes any possibility of authenticity. Authorship is a more complicated matter. Generative AI, past the consumption of training data, does not have a process where it contemplates the next step. It does not stop, mid-brush-stroke, to contemplate colour or hue, or even the motif or significance of the work. AI generates images in seconds, prioritizing speed over aesthetics, often producing something colourful but devoid of shapes, texture, or meaning. AI cannot have an intent of its own.
On that note, more often than not I find AI-generated art easily distinguished from human-made, even digital, art. Perhaps I am not alone. This lack of creative process also increases the demand for the software. If the generated image is unsatisfactory the user cannot make changes without independently manipulating the image in Photoshop. They can only generate a second, third, and perhaps a hundredth image until one suits their needs, which only adds to that large ecological footprint.
The human artist, on the other hand, is an imperfect, living specimen. Human authorship is marked by crucial flaws and anxiety throughout the creative process, which may be lengthy and better for it. It is by your process that you develop your style – you may ask AI to replicate Van Gogh and his expressive brushstrokes but it will never have a signature style of its own. Human involvement remains integral to authorship. Jan Svenungsson wrote that ‘’For a work of art to emerge, there must be an independent actor who claims that whatever-it-is is art and assumes the role of the artist, with all the responsibilities that entails.’’
AI challenges traditional notions of authorship by removing human craftsmanship, yet it will never generate ideas of its own and, therefore, cannot claim authorship. But what about the user – can they claim authorship? If Duchamp and Warhol could, by emphasizing the concepts behind their works, then perhaps Jason Allen can as well – if only for his Photoshop alterations and the seemingly intentional controversy surrounding his work. At its core, it’s conceptual art, avant-garde rage-bait.
Originality has always been an ambition for artists; From exploring new themes to revolutionising the use of a medium, originality connotes innovation and brilliance. It is, also, the topic of the most recurrent criticisms against AI-generated imagery. Generative AI is fundamentally limited to the recombination of existing data. Unlike human artists, who may draw inspiration from mundane acts, objects, and emotional insights, AI produces works by referencing and recombining patterns in the data they’ve been trained on or found on the web. However, these tools cannot think beyond these patterns or imagine new forms that have not already been presented to them. AI can only remix, alter, or synthesize the data provided to it, works that already exist. Tools like DALL-E 2 and Midjourney produce nothing more than sophisticated collages.
Still, the concept of originality in art has always been somewhat fluid. Even human artists often create works by borrowing from cultural influences, other artists, and historical styles. Claude Monet famously drew inspiration from Japanese art during the late 19th century – which greatly influenced his work and that of fellow Impressionist artists. The poet Audre Lorde believed that there are no new ideas ”only new ways of making them felt,” which exemplifies the ironic undertones of originality as an ideal, as it is not entirely attainable. Human creativity involves recombination but it is shaped by the individual artist’s subjective experiences, by joy, fear, and grief.
AI generates amalgamations of pre-existing art without giving credit to the original work, which is intellectual property and often copyrighted. This practice has led to legal disputes, such as Getty Images’ lawsuit against Stability AI, alleging unlawful use of copyrighted materials for training algorithms. Artists have protested the appropriation of their creative efforts in various lawsuits against OpenAI, Meta, Google, and others but are often unsuccessful. Every brushstroke or pixel in AI art is, at best, derivative and, at worst, stolen. Allowing this practice undermines the work of human artists who rely on fair compensation for their creativity. The legitimacy of AI-generated art has become a topic of debate in the art community, with lacking originality and ethical concerns positioning AI as a tool rather than a medium for genuine artistry.
Expression comes to mind. Art, widely perceived as a uniquely human endeavour, is practised across the world as a hobby and a form of expression. A crucial limitation for AI-generated imagery, then, is the lack of human experience and emotion – humans create art from emotion and identity, while AI is bound by predefined instructions, incapable of genuine expression or creative autonomy. More often than not art emerges as a response to feeling, lived experiences, and the complexities of existence. Although a physical installation, which AI may struggle to produce, Tracey Emin’s My Bed is a profound example of this. Created after a severe personal crisis, her cluttered, dirty, bed captures the raw vulnerability and chaos of human emotion. Emin described how the piece represented a “time capsule” of her life, embedding her pain and reflection into the artwork itself. Such works transcend physical depiction, serving instead as deeply personal and communicative mediums.
AI has, by its artificial nature, nothing to express. AI functions as an external observer, creating an impression of emotional expression without truly experiencing it. I am not alone, nor original, in painting a portrait of my family pet, but the final painting is unique in the specific cat, technique, and grief that inspired me to paint it in the first place. AI may be able to replicate artistic techniques but it cannot produce works that reflect internal states, struggles, or experiences – for it has none of its own. AI can only be compared to an actor as it mimics human expression but does not understand it.
The capacity for expression in human art fundamentally contrasts with AI’s sole ability to depict. Spontaneity, too, defines much of human expression and further distinguishes it from AI-generated works. Emin’s My Bed arose organically from her lived experiences, embodying a visceral need to process her emotions and confront personal despair. This urge, or spontaneity, to create and express oneself contrasts sharply with AI’s dependence on pre-programmed algorithms and user instructions. AI lacks autonomy and cannot initiate or modify its creative direction without external input, reducing its role to that of a tool rather than a creator. Continuing with the analogy of acting – AI cannot function without a director.
Indeed, AI is neither artificial nor intelligent; It is, at best, a tool for executing human directives, devoid of true creativity or understanding. A tool, rather than a medium of any merit. AI-made images may very well be considered art if altered by secondary means or given context and conceptuality – even if just to provoke. Defining art is near impossible but much of it, historically, has challenged our preconceptions and the status quo. Allen’s award-winning piece is no different.
Perhaps calling AI-generated objects lazy is brutal, but I must confess I find the use of generative AI unambitious and uninspired. Artists using traditional mediums don’t always produce new, inspired, work – how many times haven’t I returned to the same motif when lacking inspiration? Nor do we always challenge ourselves, but it is still a process of creation that cannot be replicated. Small changes from grip to brand of pencil produce unique results. AI is instant and thoughtless, there is no effort spent besides electricity and water. It is as convenient as it is lazy. Allen edited his art in Photoshop and exhibited it with the intent to challenge and provoke, just like Duchamp’s urinal and Warhol’s conveyor belt approach. It may all be art, for it is conceptual, but it is lazy.
In truth, AI will not destroy nor revolutionise artmaking any time soon. In fact, it would have a much greater appeal to me, personally, if it could do my dishes or washing, if it would free time for creative purists rather than paint my portraits and write my poems for me, in the instantaneous, joyless, manner it performs. Art is not dead. I will hold on to my paint and easel and support real, human, artistry, local and worldwide.
Featured Image Credit: Pexels.
Sleep-deprived fourth-year Literature, Film, and Media student. Cat enthusiast, Fleetwood Mac devotee, and avid collector of hobbies and obscure facts. Occasionally finds the time to paint, crochet, and write stories.
