What Art History Teaches Us About the Question of AI Creativity
- walkerinthecloud
- Jan 14
- 7 min read
On November 12, 2025, a new country tune titled Walk My Walk by the artist Breaking Rust reached number one on Billboard’s Country Digital Song Sales Chart, surpassing three million streams on Spotify in under a month. This would not be especially remarkable except for the fact that the song was generated entirely by artificial intelligence. The public reaction was divided. Some listeners, like myself, found it impressive that generative AI has reached a tipping point where it is creating chart-topping music. Others saw it as a threat to human creativity or as a gimmick, fueled more by novelty than by musical merit. TIME magazine dismissed the phenomenon as overblown, arguing that the song’s success reflected digital hype rather than genuine achievement (Caramanica, 2024). Yet regardless of the critique, Walk My Walk functions as a cultural signal. AI is no longer producing the cringeworthy, obviously synthetic melodies that once defined machine-generated music. It is now capable of creating competent, chart-reaching songs that are genuinely catchy. This is just one example of AI producing creative work that many consider artistically worthy, part of a growing set of cases that force us to revisit the age-old questions of what counts as art and what constitutes creativity. It also raises a relatively new and pressing question: must creativity remain exclusively human? Art history can teach us valuable lessons about these questions. Creativity has never been fixed, and from Duchamp’s urinal to Pollock’s drip paintings to the award winning AI generated Théâtre D’opéra Spatial, the definition of creativity has consistently been challenged throughout history.
Before we dive deeper, we need to briefly understand who Margaret Boden is, what her criteria for creativity entail, and why they are central to this paper. Margaret Boden is a leading cognitive scientist and philosopher of creativity whose work has been foundational in discussions of both human and artificial creativity. In The Creative Mind (2003), Boden defines creativity through three core criteria: novelty, surprise, and value. According to Boden, a creative product must be new in some meaningful way, unexpected within its context, and valuable or worthwhile to a particular audience. These criteria are especially important for this paper because they provide a structured framework for evaluating creativity without relying on assumptions about intention, consciousness, or emotional experience. While this essay uses Boden’s criteria to assess both human- and AI-generated artworks, it does not attempt to determine whether these criteria themselves should be revised; rather, they serve as an analytical lens through which changing forms of creativity can be examined.
The history of art is filled with moments where new forms of expression arrived before the public was ready to understand them. Jackson Pollock’s drip paintings, now revered, were once mocked by the same institution who is now mocking Walk by Walk, citing that Pollack’s art was merely random splashes masquerading as art. TIME magazine itself coined the nickname “Jack the Dripper” in 1956, suggesting Pollock was little more than a man flinging paint at a canvas. Today, those same paintings are highly sought after by prestigious collectors with resale prices upwards of $200 million. Whether Pollock intended to or not, he redefined what the public viewed as art, a shift that led to his art seen as novel, surprising, and eventually deeply valuable, aligning closely with Margaret's Boden's (2003) three criteria for creativity.
Fig. 1 Number 17A, 1948 by Jackson Pollock, courtesy of jackson-pollack.org.
Another example is Marcel Duchamp’s Fountain, the infamous urinal signed “R. Mutt”, which sparked an even deeper crisis to what constitutes art when it was submitted to the American Society of Independent Artists exhibition in 1917. The piece was rejected despite the exhibition's stated rule of accepting all works that submitted the entrance fee. William A. Camfield, a leading scholar and professor of art history with a PhD from Yale, highlights the ongoing disagreement about the piece, “Some deny that Fountain is art but believe it is significant for the history of art and aesthetics. Others accept it grudgingly as art but deny its value. To complete the circle, some insist Fountain is neither art nor historically consequential, while a few assert that it is both, though for incompatible reasons” (Camfield, 1917). This conflict itself is why Fountain remains central in art history courses. It forces enduring questions: is art defined by the object, the artist’s intentions, or the viewer's interpretation? It also meets two, possibly three, of Boden’s criteria: it was certainly novel and surprising, and it would be extraordinarily valuable today had it not been lost.
Fig. 2 Fountain, 1917 by Marcel Duchamp, courtesy of Wikimedia Commons.
The debate resurfaced in 2022, when Jason Allen’s AI-generated work Théâtre D’opéra Spatial won first place in a fine-arts competition. The judges, unaware of its computational origin, awarded it the top prize. When it was later revealed as having been AI generated, criticism followed. Some argued that Allen had done nothing worthy of an artistic prize because he used Midjourney– a text-to-image AI generated tool. Others claimed the artwork should not have been included in the competition because the AI model had been trained on existing human made artwork. Yet assessed under Boden’s criteria, the piece was novel (it had never existed before), surprising (the judges did not identify it as AI generated), and valuable (it was deemed the winning submission). As Vincent (2022) noted, the backlash stemmed not from the artwork itself but from the discomfort with the idea of machines creating art.
Fig. 3 Théâtre D’opéra Spatial, 2022 by Jason Allen, courtesy of The New York Times.
This discomfort echoes long-standing anxieties about whether machines can “originate” anything. Ada Lovelace famously wrote in 1843 that the Analytical Engine “has no pretensions to originate anything. It can do whatever we know how to order it to perform” (as cited in Franceschelli & Musolesi, 2024). Many still treat this sentiment as definitive, yet I would tend to disagree as I believe that with generative AIs continuous improvement it will be able to create original and artworthy content without detailed prompts. We have already seen great strides in this area. An example of its evolution was presented at the Hinton Lectures in Toronto in November of 2025. Owain Evans, a Berkeley AI researcher and founder of Truthful AI, presented a compelling demonstration of just how far generative AI models have come. Using the same vague prompt in 2022, “create a vintage image of a woman smoking”, he showed the evolution of AI image generation across three years. The 2022 images were abstract and arguably closer to an expressionist experiment. The 2025 images, however, looked like an authentic vintage photograph - rich in detail, contrast, and texture. Watching the progression, I found the 2025 results realistic and beautiful. Evan’s presentation revealed something important around AI’s creativity: AI is not static. It is rapidly improving in its ability to produce images of high quality and of originality.
Fig. 4 Presentation slides generative AI image of Vintage Image of Woman Smoking, 2022 by Owain Evans courtesy of The Hinton Lectures.
Fig. 5 Presentation slides generative AI image of Vintage Image of Woman Smoking, 2025 by Owain Evans courtesy of The Hinton Lectures.
This raises another key issue–critics often claim that AI merely imitates because its output derives from training data created by human artists. Yet this mirrors how humans learn. Artists often study art history or art fundamentals in college, absorbing stylistic influences and gaining inspiration from earlier generations. No creator, human, or machine works free of influence. Much of creativity emerges from the inspiration of the world around us. If we dismiss AI on the grounds that it draws from famous artworks or the world we created for it, we must also confront the reality that human originality is similarly entangled with these notions as well.
Prompting complicates this even further. When a detailed prompt shapes the output, is the creativity in the instruction or in the artifact? And what happens when the prompt provides almost no direction? To test this, I asked ChatGPT to “create a piece of art any way you would like”. The resulting image depicted a vivid field of multicoloured flowers with trees and a swirling sunny sky subtly reminiscent of Van Gogh’s Starry Night. ChatGPT even named the artwork, Whispers of the Sun. Judged by Boden’s criteria, the image displayed novelty in its composition and surprise in its aesthetic choices. Its value, however, remains low unless it were posted online and happened to go viral, or were turned into an NFT that someone purchased. This aligns with Franceshelli and Musolesi’s (2024) argument that while LLMs typically take inspiration from training data, their outputs can still meet established criteria for what humans recognize as creative.
Fig. 6 Whispers of the Sun, as named and created by ChatGPT, 2025
Ultimately, the history of art teaches us that our definition of creativity evolves. Each generation confronts new forms of expression that initially appear unsettling, illegitimate, or even absurd, only to later recognize them as pivotal contributions to art itself. Today, AI compels us to confront yet another shift. As generative systems produce music, images, and writing that satisfy many of the same criteria we use to evaluate human creativity, we must ask whether our understanding of creativity needs to expand. Are we prepared to acknowledge that AI can participate in creative processes, or must we redraw the boundaries of what creativity means? Or do we take a position similar to the one Camfield documented around Fountain: refusing to call it art, yet still acknowledging its undeniable significance in the history of art and aesthetics?
Comments