In 2018, an AI-generated painting sold for $432,500. In 2023, programs like DALL-E and Midjourney made image generation available to millions of people. Today, the tools are everywhere, but the question they raise is still open: is any of this actually art?
This post is inspired by Catrin Misselhorn’s book Künstliche Intelligenz – das Ende der Kunst? (Artificial Intelligence – The End of Art?). Misselhorn is a Philosophy Professor at Göttingen University. I read a Spanish translation by Alberto Ciria I found at a bookstore in Madrid. As far as I know, no English translation is available yet. Throughout this article I paraphrase her ideas extensively, mixing them with my own thoughts. All credit for the philosophical framework goes to her.
A note for the reader: Dear reader, this is Esther. I work as a Data & AI Engineer, and I can call myself a programmer. However, I struggle to call myself an artist. I don’t make money from my art, yet it has been present in my life since I can remember. In this polarized world, Art and Technology seem like complete opposites, but they may also be different faces of the same coin. At the end of the day, both art and technology are things humans make.
This article will be a bit different than what I usually write. I will walk through my personal experience with art, different philosophical theories, how technology relates to art, and I will explore the role of the artist (whether a machine can be considered as one), as well as the key role of connection when experiencing art.
What Happens When I Create
Ever since I was a little girl, when I paint, I lose track of time, space, and my sense of self. But at the same time, I feel more myself than ever. Following one of my idols, Joni Mitchell “If you can paint with a brush, you can paint with your words” I picked up a banjo and started writing songs, not so long ago. This opened up a new dimension of expression for me.
And what I had noticed while painting was confirmed while writing music: sometimes I have no idea what is happening. It’s like a force of nature that takes over. I may start with an intention, but I am not in full control. Even when I’m working from a reference, I notice things about my reality mid-process that I wasn’t aware of when I began.
I, as a creator, am just one factor. The process itself is another. Whatever takes over me (I guess instinct and intuition influenced by my thoughts, my feelings, the weather, etc) is an undeniable part of the finished piece. I think losing control is part of artistic production. Sometimes colors or phrases appear unintentionally. Art pieces often carry semantic layers their authors never consciously intended. It is a mix of conscious and unconscious processes.
What Is Art, Anyway?
In her book, Misselhorn explores different philosophers who said that art was dead, way before the time of machines making art.
Hegel (1770–1831) argued for the “death of art,” claiming it had peaked with Ancient Greece and become obsolete. History proved him wrong: photography, abstract art, and cinema all arrived after his declaration.
Arthur C. Danto (1924–2013) placed the death of art much later, with Andy Warhol’s Brillo Boxes in the second half of the 20th century. In this moment art had became indistinguishable from everyday objects. For Danto, a key requirement is that meaning must be embodied in the piece itself. This means that you cannot separate the meaning from the object without losing something. He also argued that art must be the result of the spiritual activity of an artist. This is exactly what AI calls into question.
Several requirements are discussed for something to count as art:
- A piece should illustrate a meaning.
- That meaning should consist of a higher-order content that goes beyond the literal (like a metaphor, a way of seeing the world).
- Understanding that meaning requires an interpretation, which must take the context of the art world into account.
This creates what philosophers call a hermeneutic circle: to interpret a piece, you need prior understanding; and to build that understanding, you need interpretation. To break this circle we need shared context, which is why museums hang texts about a painter’s life on the wall, and books open with prologues.
Importantly, authors cannot fully determine what their work means. Interpretation must also make sense from the audience’s perspective. Author, audience, and the art piece (with its creative process) are all key.
Art created by machines
GAN: When Machines Paint
The first wave of AI-generated art came from GANs (Generative Adversarial Networks). Two neural networks work together: a generator creates new data, while a discriminator compares it against a training dataset to judge whether it’s real or generated. This loop repeats until the discriminator can no longer tell the difference.
A famous example: Portrait of Edmond de Belamy (70×70 cm) painted in the style of old masters, was created by the art collective Obvious and sold for $432,500.
Edmond de Belamy (2018). Image via Wikimedia Commons. Public domain.
What can be polemic is that Obvious used an algorithm built by Robbie Barrat (b. 1999), a young American AI researcher, without crediting him as requested. And also without compensating him from the sale. Furthermore, they used other artists’ paintings as training data without permission.
We could say Obvious is responsible for the performance. I mean: the conceptual act of presenting this as art, but not for how the painting actually looks. And perhaps, following Warhol’s logic, the performance itself is the artwork, not the painting.
CAN (Creative Adversarial Network) pushes further: it actively tries to generate art that differs from its training data. But this system is evolving towards abstract art, since it figured out that art develops in this way, and the generated art does not seem very original. The results resemble Mark Rothko or late Gerhard Richter.
Creative Adversarial Networks. Source: medium.com/@abbymilla
Can AI Be an Author?
Jorge Luis Borges illustrated why authorship matters in “Pierre Menard, Author of the Quixote”: a fictional French writer reproduces Cervantes’ Don Quijote word for word. The texts are identical, yet they are completely different works. Context, intention, and the reader’s knowledge of the author change everything.
Computer pioneer Ada Lovelace (1815–1852) already asked whether machines could ever be truly creative. The creative outcomes of AI algorithms cannot be predicted even by their own programmers. But sometimes I wonder, is there a parallel with that “inner fire” I described earlier in my own creative process?
If we follow hypothetical intentionalism, which is the idea that what matters is not the real conditions of creation, but whether the piece itself provides enough information to construct a hypothetical author, then AI could be considered an author. But Misselhorn argues that AI cannot take responsibility: not aesthetically, not morally, not individually, not collectively. You cannot attribute aesthetic responsibility to a system that only simulates or acts in a functional way.
She also mentions that the German Constitution defines art as “the immediate expression of the individual personality of the artist,” a definition shaped by both conscious and unconscious processes. AI has neither.
What If AI Becomes an Agent?
So far, AI art systems react: they respond to inputs, whether a training dataset or a text prompt. But what happens when AI becomes an agent: a system that can act autonomously, adapt to its environment, and respond to the world around it?
This opens new possibilities, but also new problems. The unpredictability we already see in AI outputs could be seen as some kind of “inner fire”. But there’s a crucial difference: an agent can act, yet it still cannot take responsibility for its actions. And this is crucial to what makes someone an author.
Programs like DALL-E, Stable Diffusion, and Midjourney work differently, translating text prompts into images. Here, the programmers are responsible for the algorithm, but the person writing the prompt carries the aesthetic responsibility for what gets generated. The prompt itself could be considered part of the artwork. Misselhorn used the parallelism of how Old Masters directed workshops of assistants to realize their vision.
Art as Human Connection
David Hume (1711–1776) wrote: “We choose our favourite author the same way we choose our favourite friends, by similarity of temperament and character.”
Wayne Booth (1925–2005) took this further. For him, the aesthetic experience of art involves relating to its author as someone you would (or wouldn’t) want as a friend. Not a transactional friendship, but something closer to Aristotle’s virtuous friendship: a mutual appreciation of personalities. Even if the relationship isn’t symmetrical, given that artist and audience play different roles, it is mutual in the sense that the work can only fully manifest when it reaches a receptor with a compatible temperament.
One of the most rewarding things about art is not the encounter with yourself, but with another person, often from a distant time and place, who suddenly feels incredibly close. Art can cure solitude. It connects humans across centuries and kilometers.
One of these experiences, which I remember clearly, happened to me while reading Madame Bovary. The main character is reading a book, and she describes this very specific feeling of connection with an author. She was reading about a feeling she thought only she had ever experienced, until she found it on the page. And I, reading about her reading, felt exactly the same thing. Like a golden thread connecting strangers across time.
When I paint someone or something, I feel tenderness toward them. When I gift my art, it is an act of love. I can experience connection when I’m on both sides, creating or receiving.
Conclusion
I feel like experiencing art is a very human experience. AI can replicate. It can generate outputs that are aesthetically beautiful, conceptually interesting or commercially valuable. But the kind of connection I’ve described, that inter-dimensional, human-to-human understanding, cannot be faked. This is more noticeable when talking about our favorite artist, musician, painter, writer, whoever. We may even detect flaws in their work, but we will feel a connection to their art piece regardless. There will always be something ineffably real about encountering another person’s consciousness through their work.
I think circumstances are everything. Art is a symbiosis between the artist (and their creative process), the artwork, and the interpreter.
The ultimate goal of art should not be to paint a proper portrait or write a commercial love song. The goal is to express yourself, to connect with yourself, with others, with your instincts and your inner fire. A vision is defined along the creative process, not before it.
As humans, we want to be in touch with reality. Art manifests a real human person, not a simulation. Or maybe the question isn’t whether AI can make art, but whether we can truly connect with it.
