Marie Dernoncourt · 2026
ART AND THOUGHT
This reflection is also available in French. Read the French version.
Artificial intelligence can now produce, in a matter of seconds, an image capable of making us stop. In a world where images keep pouring in, that speed intrigues me as much as it unsettles me.
It can compose melancholy music without having known sadness, or produce a moving image without our being able to establish that it feels what it evokes.
The viewer’s emotion is real; it tells us nothing about the machine’s own feelings.
The temptation, then, would be to draw an immediate boundary: if it does not feel what it produces, it cannot truly be art.
But that boundary does not hold for long.
A composer does not need to be sad in order to write sad music. A painter can work on solitude without being alone.
The human artist, however, creates from within a lived history. Emotions belong to a memory, sometimes to a body that retains their trace, even when they are not felt at the moment of making.
Is that indispensable to making art?
We have not established that.
In my studio, I work almost against that flood of images: material, a few cubes, lines of passage or tension. I take things away to let the eye breathe, to quiet some of the world’s noise. This slowness tells you where I speak from; it does not, by itself, prove what makes art.
Intention and autonomy
Perhaps, then, we should look to intention.
The artist looks, hesitates, discards, begins again. An accident may be welcomed, or the artist may discover, in the course of working, something that was never planned.
Artificial intelligence, too, when instructed to do so, can explore several possibilities, compare them, evaluate them according to certain criteria, and pursue an objective without a human specifying every intermediate step.
In 2026, Google DeepMind evaluated its agent Aletheia semi-autonomously on 700 open problems related to Erdős conjectures. It generated solutions to four of them autonomously.
These results demonstrate neither consciousness nor an intention of its own.
They do show that a human need not dictate every step.
Autonomy in solving a problem does not imply autonomy of intention.
Creating has never meant simply making
Since the twentieth century, the collages of Picasso and Braque, Duchamp’s acts of selection, and Kosuth’s conceptual work have shifted the boundary between making, choosing, and conceiving.
Joseph Kosuth’s One and Three Chairs (1965) offers a particularly clear example: the installation brings together a chair, a photograph of that chair, and a dictionary definition. The artist’s act lies, in part, in selecting and arranging these elements.
We therefore cannot dismiss something produced through AI solely because a human did not make it by hand.
That does not mean that AI has become an artist.
Neither human nor machine begins with a blank page
Generative models learn relationships from datasets containing, depending on the system, texts, images, and other human productions.
The human artist also inherits forms and a culture, but transforms them within an embodied and biographical existence.
To say that the human creates while the machine recombines would be too simple.
Yet their ways of learning are not equivalent.
And then comes the viewer
Imagine an image that makes us stop. We then learn that it was generated by AI. Its appearance has not changed; perhaps our judgment has.
A 2023 study randomly labelled AI-generated paintings as either human-made or AI-made. Those presented as human-made received higher average ratings; the story attributed to the work and the effort believed to lie behind it also mattered to viewers’ judgments.
This experiment does not speak for every viewer; it shows that what we believe about an image’s origin can affect our judgment.
When an image that moved me turns out to have come from AI, what really changes?
The emotion remains. What changes is the story I thought I was reading in it.
Emotion alone, however, does not make a work of art. A landscape or a quality of light in nature can move us deeply without thereby becoming a work of art.
Aesthetic experience and artistic status are not the same thing.
What a signature tells us
I do not fear the tool itself: selecting, revising, and presenting AI-generated images can be part of an artist’s work. What troubles me is presenting an image largely generated by a system as the exclusive product of one’s own labour. Then the signature conceals a decisive part of how the work was made.
I worry that galleries may one day fill with images whose making is left unexplained. I do not know whether this will become the norm. Viewers should be able to distinguish human choices, what was generated, and the work done between the two.
The object, the work, the artist
Perhaps we have been asking one question where we should have been asking three:
Can AI produce an object with aesthetic value?
Can that object become a work of art?
Can the AI that produced it, for that reason, be called an artist?
These three questions may not require the same answer.
AI can produce, compare, select, and pursue certain objectives. But capacity, consciousness, subjective experience, will, and intention are not synonyms.
In 2026, philosopher Tom McClelland argued that consciousness need not be required for all creativity but might be required for creativity pursued with genuinely aesthetic aims. This is a philosophical position, not an established fact. Whether artificial consciousness is possible remains an open question.
We should therefore remain cautious about attributing to current systems a subjective experience comparable to that of human beings.
But recognizing this difference does not yet demonstrate that it is a necessary condition for being an artist.
The body and the gesture
On my canvases, I move a cube, revisit a line of passage, and wait for the material to respond. Sometimes taking something away finally gives the eye room to breathe.
This work engages the body, takes time, and sometimes welcomes accident: thinking and making cannot always be separated.
But manual gesture alone does not define the artist. Photography, conceptual art, delegation, and digital creation are enough to weaken that criterion.
Imagine that tomorrow a machine has a robotic body, able to handle a brush and respond to the resistance of the canvas. It would perform a gesture. Would it experience that gesture?
Turning our criteria back toward the human
We ask the machine: do you feel? Do you understand? Do you have an intention, an intuition, lived experience? Let us apply the same demands to the human.
The artist does not always fully understand what they create. They may welcome accident, delegate part of the fabrication, transform what they have inherited, and provoke an emotion they did not foresee.
Even intuition must be handled carefully: it cannot become the mysterious word we reach for whenever we can no longer explain human creation.
One major difference nevertheless remains: human beings possess a subjective, embodied, and biographical experience of the world that we cannot currently attribute in the same way to AI systems.
A real difference is not necessarily a definition of the artist.
What if creation is sometimes distributed?
Who creates: the human or the machine? Some practices complicate that opposition.
A person sets a direction, the AI generates; the person chooses, revises, and decides what to show. Some of these operations can also be automated.
Creation can then become a process distributed across human intentions, technical systems, successive choices, and context.
That does not mean that this ensemble automatically constitutes a new “collective artist.”
Where does the artistic act lie when creation is distributed?
A revealing lens and a limit case
Emotion. Intention. Body. Gesture. Lived experience.
Each time we look for a definitive boundary, we return to profoundly human characteristics. Perhaps they are necessary to art; perhaps we associate them with art because every artist we have known until now has been human. We have demonstrated neither claim.
AI acts as a revealing lens: it brings to light what we assume when we speak of creating, choosing, understanding, or being an artist.
It is also a limit case: it pushes these criteria to their boundaries and forces us to ask which ones hold.
It does not oblige us to call it an artist.
Nor does it entitle us to decide today what it could never become.
A boundary still open
Producing something with aesthetic value may not be the same as producing a work of art. Producing a work of art may still not be the same as being an artist.
Some of these differences are profound. Yet a difference between human and machine is not, by itself, a definition of art.
The opposite conclusion would be equally premature: the growing capabilities of artificial systems demonstrate neither subjective experience, nor a will of their own, nor the status of artist.
So the boundary remains open.
Not as a defeat of thought, but as its point of departure.
I do not have the answer. I want to preserve the time to look, honesty about what is shown, and room for silence.
Artificial intelligence may not yet tell us what an artificial artist is.
It forces us to look again at what we thought we knew about the human artist.
And perhaps that is where the question truly begins.
Sources and works consulted
For further reading:
Tom McClelland, Does artificial creativity require artificial consciousness?, AI & Society, 2026.
Alice C. Helliwell, Can AI Create with Flair? Artistic Creativity and Artificial Intelligence, The British Journal of Aesthetics, 2026.
Dominic McIver Lopes, AI Art and Artists: What They Are, What They Could Be, What They Should Be, The Journal of Aesthetics and Art Criticism, 2026.
Sofie Vlaad & Elliot Samuel Paul, Appreciating AI Art: Aesthetic Vices, Virtues, and Values, The Journal of Aesthetics and Art Criticism, 2026.
Lucas Bellaiche et al., Humans versus AI: whether and why we prefer human-created compared to AI-created artwork, Cognitive Research: Principles and Implications, 2023.
Thang Luong & Vahab Mirrokni, Accelerating Mathematical and Scientific Discovery with Gemini Deep Think, Google DeepMind, 2026.
Stanford Encyclopedia of Philosophy, The Definition of Art, revision 2024.
Museum of Modern Art (MoMA), Joseph Kosuth, One and Three Chairs, 1965.
© 2026 Marie Dernoncourt — touscreateurs.eu — All rights reserved.
This reflection is also available in French. Read the French version.