Study Argues AI Can Make The Art, But Can’t Impersonate The Artist

Polished Art Takes Minutes Now. Artists Still Take Years.

A pianist spends twenty years building calloused fingertips and training an ear to hear music the way a composer does. An AI program can now produce a piece that sounds impressively polished after a few typed instructions. That gap, between the years a human needs and the minutes a machine takes, is the starting point for a new academic paper asking a question most AI art debates have missed: what happens to the point of making art when the hard part disappears?

Amit Raphael Zoran of Hebrew University of Jerusalem wrote the paper, published in the journal AI & Society. Debates over generative AI keep circling who owns an AI-generated image or whether a chatbot’s poem counts as real art. Zoran says all of that fixates on the finished product and ignores the personal transformation an artist goes through while learning to make something good.

Learning to paint, sculpt, or compose has long done double duty. Practice produced finished pieces, and it also changed the artist, sharpening perception, judgment, and skill. Renaissance perspective changed how people perceived space. Zoran calls this kind of transformation “artistic becoming,” and the paper’s central claim is that generative AI loosens its link to the finished artwork.

A Philosophy Paper About AI Art, Not a Lab Study

No test subjects, lab experiments, or surveys were involved. The paper is an argument built from aesthetics (the philosophy of art), craft theory, anthropology, and research on how people interact with computers. Zoran traces artists who have long worked alongside computers, from early computer art pioneers to contemporary artists using robots and AI software.

Across that history, computers mostly handled precision and repetition while humans held onto interpretation and meaning. A drawing program could execute detailed shapes, but a person decided what they meant. Generative AI shifts that balance by working on the meaningful part too. Its raw material is “the human symbolic and cognitive world itself,” meaning the writing, images, and art people have produced over centuries.

From there, the paper builds what it calls the “decoupling thesis,” a fancy way of saying that making sophisticated art no longer requires going through the training that used to shape the maker. Zoran notes AI-made work does not appear from nowhere, since prompting and editing remain real activities and the systems rest on enormous amounts of past human labor.

Where Human Creativity Goes Once AI Takes Over the Hard Part

Zoran argues the meaningful part of art relocates rather than vanishes. Artistic becoming operates on three levels: inside the person making the work, inside the person experiencing it, and across a culture over time, as when a new art movement changes how people see. Generative AI mainly disrupts the first level, the maker’s development, and leaves the other two open questions.

Creative effort shifts toward prompting, curating, selecting, and holding an ongoing back-and-forth with an AI system. Zoran calls these “genuine developmental practices,” though they differ from older training, since many can be entered without a long apprenticeship and operate on ideas rather than physical materials.

To show what this redirection can look like, Zoran describes three projects from the author’s own practice. In one, a live lecture on perception reveals that parts of the speaker were prerecorded, showing the audience how easily confidence in a digital presence can be manipulated. In another, participants taste an essence reconstructed from an old oak using ideas from historical traditions such as alchemy, along with botany and chemistry. In a third, participants smell and taste an essence derived from a tree and hear sounds recorded from it, which AI transforms into music performed by a human musician. No measurement of how these affected participants was attempted, so they show only that the approach is practical.

In the Age of AI Art, Deciding What Is Worth Making Matters Most

Because sophisticated-looking output is now so easy to generate, Zoran argues the central question has shifted from how to produce something impressive to what, among everything possible, is worth making.

Zoran’s answer is specific: practices worth pursuing take AI’s knack for turning out words, images, and music and route it back into physical, lived experience, such as the push-back of real materials, place, the body, and other people. Work that stays entirely inside remixed digital content, however polished, lacks the weight art has historically carried, in this view, because it multiplies forms without deepening anyone’s connection to anything. The paper sketches provisional recommendations. Artists should treat the effort AI saves as a budget to reinvest in hands-on experience. Schools, museums, and funding bodies should value process and documented engagement alongside finished work, and AI builders should avoid pushing everyone toward the same results.

AI can hand over the polished piece in minutes, but the change an artist undergoes cannot be handed over. It can only be lived through. Whether a machine can produce something that looks like art is the wrong test. The better one is whether people using these tools are still willing to let the experience change them.

---

Disclaimer: This article is for informational purposes only and is not professional advice. It summarizes a published academic paper, and the paper’s conclusions are the author’s interpretation rather than experimentally tested findings.

---