The Work of Art in the Age of Artificial Reproduction: Proust, Vermeer, and Cézanne, and GPT-4
Top image: Hobbs, Tyler, and Wist, Dandelion. QQL #43.
With the release of GPT-4, I am finally ready to concede that A.I. can perform most tasks better than humans. From acing the SAT to finding potential exploits in Ethereum Smart Contracts, GPT-4 feels like it’s not that far from AGI—and any Harvard Professor not worried about how this technology can be used to “assist” students in their assignments assigns too much integrity to college undergraduates.
But can A.I. truly create art? The answer is a resounding yes—if by art we mean objects that elicit aesthetic pleasure. Setting aside old Kant’s assertion that there can be no objective principle of beauty and that aesthetic experience is confined to the pleasure derived from the free play of the imagination and understanding, I find myself dissatisfied with this answer. Hence, in this essay, I will challenge this view by emphasizing the limitations of such a perspective on art. At the core of my contention lies the notion that A.I., as a non-subjective agent, lacks perspective—I believe that a work of art is nothing less than the personal experience of the artist and that the artist’s past when remembered and reexpressed transforms into a work. It is this very process that elevates aesthetic pleasure into the realm of true art.
Hilary Putnam’s “Brains in a Vat” essay offers an insightful starting point. Putnam imagines a scenario in which an ant, crawling on a beach, creates an image that humans would recognize as an accurate portrait of Sir Winston Churchill by sheer chance. Putnam contends that since the ant has never seen Churchill or even a picture of him, it can have no intention of depicting him. The ant simply traced a pattern that we interpret as a representation of Churchill. This example suggests that intention is a crucial component of representation. The ant is entirely incapable of recognizing Churchill—in fact, the ant would have drawn the same curve even if the prime minister had never existed.
In my opinion, Chat GPT is analogous to the unintentional ant. We merely perceive intention because we, as humans, assign meaning to its creations. Specifically, Chat GPT employs a transformer neural network to generate text-based content in response to input prompts. When provided with a directive like “write a Shakespeare sonnet about Tiktok,” it tokenizes the text, processes it into numerical representations, and produces an output based on relational patterns learned from its training data. Similarly, when DALL-E 2 generates visual art, it employs a comparable method to construct the piece pixel by pixel. In this sensory-deprived world—the domain of machines—A.I. is incapable of genuinely recognizing or evaluating the art it outputs. This idea is tangentially connected to John Searle’s Chinese Room intuition pump, which demonstrates that even if A.I. can mimic human mental processes, it fails to comprehend the meaning behind them. In other words, such intelligence merely follows a set algorithm procedurally—so what qualifies as art if mechanical production doesn’t?
I believe the answer lies in Marcel Proust’s assertion that there is nothing in a work of art other than the personal experience of the artist. Proust offers us a profound insight into the nature of artistic creation. This notion goes beyond merely emphasizing the significance of the creator’s private memories, which may not always be effectively communicated, but rather underscores the transformative power of his interpretation. Proust is not merely concerned with the ineffable—the aspects of experience that are too elusive or subtle to be expressed. Instead, he highlights the importance of how the artist’s past is rediscovered and metamorphosized into art through their unique perspective in a process of intense remembering. This is the key to what I am calling art: the artist engages in a dialectical dialogue with their memories, imbuing their work with an intense human meaning from a uniquely human perspective that transcends the sum of its parts. This process of transformation is akin to a searchlight scanning the depths of the artist’s experiences, seeking out implications and signs that can only be revealed to the remembering mind.
The act of artistic creation, in Proust’s view, is not merely the result of evoking a sensory experience, such as the taste of herbal tea and petites madeleines that forcefully stir a forgotten memory from slumber. Rather, it is the construction of an entire world in the present moment, composed of remembered fragments that resonate with the artist’s inner life. The art that emerges from this process has the capacity to create a vivid and cohesive whole that speaks to the human experience. The words, gestures, and symbols employed by the artist are not isolated elements; they gain their power and meaning through the artist’s unique lens. It is this creative vision, born from the depths of the artist’s experience, that gives art its transformative potential.
In Proust’s understanding of art, the artist’s interiority and personal journey are essential components of the creative process. Art is not merely a product of technical skill or aesthetic sensibilities; it is the expression of the artist’s soul, a reflection of their unique experiences and perspectives. This intricate relationship between the artist’s past experiences that make up their inner life and their creations is what sets true art apart from the mere imitation of form or style begot from mechanical reproduction. As such, Proust’s view of art emphasizes the indispensable role of human experience and subjectivity with all its embedded scaffolding, an aspect that cannot be replicated by A.I.
We also see this definition of art at play in paintings. Last weekend I toured the Vermeer exhibit at the MET and was captivated by his A Maid Asleep. As I stood before the oeuvre, I couldn’t help but wonder what thoughts swirled in the sleeping girl, her dreams suspended in time. Vermeer masterfully captured the essence of his subjects in their most intimate moments, their inner worlds tantalizingly elusive and entirely their own. These works of art beckon us to peer deeper, seeking the truth behind the eyes of the subjects even though they remain ultimately self-concealing. It is through this very inaccessibility of their thoughts and emotions that Vermeer unveils the essence of art, a tribute to the splendor of the human experience.
Vermeer, Johannes. A Maid Sleep. 1656, The Metropolitan Museum of Art
I climbed up the stairs and made my way from Tiepolo to Bouguereau—slowly, Rococo and academic neoclassicism gave way to impressionism. I briefly paused to admire the world disclosed in Monet’s Seine before making my way to my favorite artist: Paul Cézanne. The French father of modernism was fond of portraying apples and Saint Mont-Victoire, yet his art’s distinction lies in the way he reimagined everyday objects for us, offering studies devoid of extraneous symbolism and contrived intellectual depth. With an apple, Cézanne astonished Paris and made love to his world. He went to the things themselves by depicting the world as it is experienced rather than as it is objectively known. He takes the world into his hands, gives it his touch, and returns it to all of us except for the fact that it was he who was there to experience it.
Cézanne, Paul. Still Life with Apples and Pears. 1891, The Metropolitan Museum of Art
In sum, art cannot be produced by A.I. because it lacks the unique experience or situation in space and time that breathes life into the interiority of the artist. Viewed from another angle, if all that was beautiful and aesthetically pleasing qualified as art, then a waterfall, a meadow of wildflowers, and the rosy hues of dawn would all be considered artistic creations. However, they are not deemed art because they are not born from human intention (except for those who attribute their existence to a divine maker), nor are they purposed as vehicles to express and convey the unique human experience. My definition of art demands a human perspective, stemming from the artist’s individual interiority and experience, situated in a specific context within space, time, and personal narrative. To claim otherwise is to misunderstand what it is to be and reject our defining subjectivity.
If art only means aesthetically pleasing objects of experience, then, yes, Chat GPT can write songs and plays, and DALL-E 2 can make good-looking surrealist collages. But this is to take an incomplete view of what art is—in other words, A.I.-generated content may allow for human interpretation and criticism by viewers, but it lacks the opposing or intrinsic point of view from which the artist can be interpreted or critiqued from. It misses the moon for sixpence. It is like yelling at the end of a deaf telephone—half the conversation is muted. It disregards aesthetic creation and allows for the tyrannical free reign of art as pleasure after la mort de l’auteur. Walter Benjamin believed that even “the most perfect reproduction of a work of art is lacking in one element: Its presence in time and space, its unique existence at the place where it happens to be.” In other words, the artist has a world, and his remembrance of things draws it out and discloses something for the viewer. A.I. has no such world, only a vector space—so it mechanically takes from this and that artist and makes a surrealist collage that is now real enough to fool us.
Ultimately, I acknowledge that Proust, Vermeer, and Cézanne developed their theories in a vastly different historical context, devoid of advanced A.I. technology. I even admit that my definition of art is anthropocentric and narrow. In the end, however, this is all to encourage the use of tools like A.I. to create more art in new innovative ways befitting our own present historical moment—something more tailored to our digital era (I.E. the metaverse on NFT rails). Long-form generative art immediately springs to mind.
Fidenzas and QQLs by Tyler Hobbs have taken the crypto world by storm, selling for many millions for rare pieces. QQL is the first project to fully leverage the product market fit of generative art because it uses a generative algorithm that is solely the birthchild of the artist and is imbued with his lived experience. On the other hand, it also allows full customization by the collector, who serves as curator and can make the piece her very own by continuously generating outputs to her exact wishes. In spring 2023, Hobbs will become the first artist in Pace Gallery’s program to present an IRL exhibition of an NFT project at one of its global locations, underscoring the gallery’s support of crypto art.
It is my hope that through our recognition that what A.I. creates is not art in the human sense, we can better leverage A.I. as a tool that complements, enhances, and even educates human creativity, instead of subjugating and replacing it. In fact, there’s a real argument to be made that A.I. art right now is essentially art theft because it takes a mass amount of real creations and then scrambles them without thought or care. The way that AI art can actually become art is by serving as a tool through which an artist expresses himself, perhaps by helping him with writing generative algorithms like QQL that don’t require the remixing of existing pieces. Feeding a particular and curated set of prompts into AI adds an element of human perspective—and choosing which prompt and iteration of the prompt best illustrates an artist’s vision is also part of the process of creating art. We have to put more emphasis on the prompt engineer as the artist rather than the program itself, which lacks the interiority and human perspective necessary to my definition of art qua art.