Unordinary

Kyoto, July 1964. A young woman walks onto the stage of the Yamaichi Concert Hall, sits down, and places a pair of scissors on the floor in front of her. Then she stays still.

The instructions are simple. Members of the audience may come up one at a time and cut away a piece of her clothing, which they may keep. ^[1]

At first nothing happens. Then someone stands up. The first cuts are small and hesitant: a corner of a sleeve, a strip of hem. Others follow. Some cut carefully. Some cut boldly, closer to the body. As the piece goes on, more of her is exposed, and she does not stop anyone. She keeps looking ahead.

The woman is Yoko Ono, and the work is Cut Piece. She performed it again in Tokyo, at Carnegie Recital Hall in New York in 1965, in London in 1966, and in Paris in 2003. ^[1]

The Audience Inside the Work

It is easy to describe Cut Piece as shocking and stop there. The shock is real, but it is not the most interesting part.

In a normal performance, the artist decides what happens and the audience watches. Cut Piece reverses this. Ono decides almost nothing after the scissors touch the floor. What happens next depends on strangers.

That changes what the artwork is made of. The material is not only fabric and a body. It is the audience's restraint, curiosity, aggression, embarrassment and choices. A person who cuts a small square from a sleeve and returns to their seat has made a decision in public. So has the person who cuts the strap. So has everyone who stays seated and watches, because watching is also a choice, one Dante punished at the gate of Hell.

The performer, meanwhile, has given up control in the most literal way available. The work only functions because the vulnerability is genuine. If the scissors were fake, or if Ono stopped people whenever she felt uneasy, it would become a different and much weaker piece.

So Cut Piece raises a question that has nothing to do with scissors: does an artwork have to be something the artist makes and the audience receives?

Ono's answer, in practice, was no. But she was not the first person to ask a question like that. The habit of pushing a medium past the edges everyone assumed it had is much older.

The Advance Guard

The word for this habit comes from the army.

In French military language, the avant-garde was the advance guard: the part of an army that goes forward ahead of the main force. ^[2] Its job was to move into ground nobody had secured yet, find out what was there, and meet whatever waited first.

The metaphor moved into culture in the first half of the nineteenth century. It is usually credited to the circle of the social thinker Henri de Saint-Simon, whose 1825 writing has artists say that they will serve as the avant-garde, because the power of the arts is the most immediate way to spread new ideas. ^[2]

The military image is more useful than it first appears, because it contains a warning.

An advance guard is ahead of everyone else. That is the point of it. But being ahead does not mean being right. A scouting party can find the pass the whole army will later march through. It can also walk into a swamp, take a wrong turn in the fog, or simply get lost. From inside the fog, those outcomes can feel exactly the same.

Culturally, the pattern holds. Some avant-garde experiments discover territory that everyone later occupies so comfortably that nobody remembers it was ever contested. Others go nowhere.

Rejected Is Not the Same as Early

This leads to a distinction that is easy to lose.

Many works now celebrated were first dismissed, mocked or refused. From this, it is tempting to conclude that rejection is a sign of being ahead of one's time.

It is not. Being rejected does not mean something is ahead of its time. Most rejected ideas are simply rejected ideas.

The reasoning error is survivorship bias. History preserves the rejected experiments that later won, because those are the ones that got museums, anniversaries and textbooks. The far larger number of rejected experiments that deserved rejection mostly disappeared. Looking back, it seems as if rejection and greatness travel together. They do not. We are looking at a filtered sample, the same filter that makes old appliances seem indestructible in Impermanent Solution.

The real difficulty is that, from the present, genuine innovation and nonsense can look remarkably similar. Both break expectations. Both irritate people. Both are defended by their makers with total conviction. The difference often becomes visible only later, when some experiments turn out to have opened a path others could follow, and most turn out to have been dead ends.

A Six-Minute Single

Not all avant-garde work stays in galleries. Sometimes the experiment happens at the top of the charts.

In 1975, Queen released a single that ran for five minutes and fifty-five seconds. ^[3] Radio singles of the time were expected to be around three minutes, built around a verse and a repeated chorus that listeners could recognise within seconds.

Bohemian Rhapsody did almost none of that. It begins with voices alone, becomes a piano ballad, turns into a guitar solo, then collapses into a mock-operatic section of layered vocals and theatrical call-and-response. That section breaks open into hard rock, which in turn slows into a quiet ending. There is no chorus that returns. Freddie Mercury later explained that it was assembled from several songs he had written, some fragments dating back to the band's early days. ^[3]

According to the Official Charts Company's own history, the record company, along with fellow musicians such as Elton John, doubted it would be a hit because it went on too long. ^[3] What changed the situation was the London DJ Kenny Everett, who had an early copy. He teased it on Capital Radio in fragments and played it fourteen times in two days. Listeners went looking for it in record shops the following Monday. ^[3]

The single was released uncut on 31 October 1975. It spent nine weeks at Number 1 in the UK, across Christmas and into 1976, and returned to Number 1 in 1991 after Mercury's death. ^[3]

This is a different kind of avant-garde from Cut Piece. Ono's work stayed, for decades, mostly within art history. Queen's experiment did something harder in one sense: it dragged the mainstream somewhere new. It showed that a commercially successful rock single could be long, structurally unstable, theatrical and chorus-free, and that millions of people would sing all of it anyway.

The contrast is useful.

• Yoko Ono asked whether a performance needs to behave like a traditional artwork.
• Queen asked whether a popular song needs to behave like a popular song.

Both questions exposed an expectation that had seemed like part of the medium itself, and turned out to be a habit.

The Pattern Repeats

Once the pattern is visible, it appears again and again. The examples below differ in form, but each one questioned something that had been treated as a given. The first two columns are historical record; the third is this article's interpretation of what each work exposed.

Sources: ^[4,5,6,7].

Cubism broke the single viewpoint that painting had inherited from the Renaissance, showing objects from several angles at once and flattening the picture into fragments. ^[5]

Dada, which emerged during the First World War, attacked the idea that art must be reasonable at all. Its artists used nonsense, chance and ridicule, partly as a response to a civilisation that had produced the war. ^[6]

Marcel Duchamp's Fountain is the most famous single case. In 1917 he bought a porcelain urinal from a New York showroom, signed it "R. Mutt 1917", and submitted it to the exhibition of the Society of Independent Artists. The show was meant to have no jury. Its organisers voted to reject the work anyway. ^[4] The original was soon lost. What survives are photographs and replicas Duchamp later authorised. ^[4]

A defence published in the small magazine The Blind Man in May 1917 put the argument in two sentences: ^[4]

Whether Mr. Mutt with his own hands made the fountain or not has no importance. He CHOSE it.

John Cage's 4′33″ was premiered by the pianist David Tudor in 1952. Tudor sat at the piano, marked the beginning and end of three movements, and played nothing. ^[7] The piece is not silence. It is whatever the audience hears in that time: the wind, the rain, chairs creaking, people leaving. Cage made the listening frame the composition.

Photography faced a similar objection from the other direction. In 1859, the poet and critic Charles Baudelaire argued that photography should remain a humble servant of the arts and sciences, like printing, rather than claim to be art itself. ^[8] The machine, the argument went, did the work.

Sampling turned existing recordings into raw material. Early hip-hop producers built tracks out of fragments of other people's records, and the courts initially treated this with hostility. In 1991, a New York federal judgment against Biz Markie opened with the words "Thou shalt not steal." ^[9] In 2005, a US appeals court summarised its position as "Get a license or do not sample." ^[10] Sampling survived, but became a licensed craft rather than a free one.

Other movements belong to the same family without needing a full case study here. Punk argued that three chords, cheap equipment and no training could be enough. Electronic music made the machine itself the instrument, and the person programming it the performer. In each case, part of the audience answered the same way: that is not really music.

What Is Art?

Every example above eventually leads to the same question. It has no single authoritative answer, and philosophers have been arguing about it for a very long time. ^[11] It is more useful to look at the competing criteria people use, and at where each one breaks.

The table below lists common candidates. The middle column describes what each criterion captures; the right column is the case that strains it. Both are interpretive, not settled findings.

No single criterion survives every case. Most serious definitions combine several of them, and most of those combinations still run into counter-examples. ^[11]

One influential response came from the philosopher Arthur Danto. Looking at works that were physically identical to ordinary objects, he argued in 1964 that what makes something art cannot be seen with the eye alone. It depends on an "artworld": an atmosphere of theory and history that lets a work mean something. ^[12] This helps explain Fountain. It also makes art uncomfortably dependent on who happens to be in the room.

The uncomfortable questions follow directly.

If Duchamp did not manufacture the urinal, what exactly did he create? Perhaps not an object at all, but a new relationship to one: a choice, a title, a placement and an argument.

If Yoko Ono did not determine each cut, who authored the final performance of Cut Piece? She wrote the rules and accepted the risk. The audience made every visible decision. Perhaps the answer is that she authored the situation and they authored its contents.

If John Cage allows the sound of rain to become part of a composition, where does the composition end? Cage's answer seems to be that he composed the act of listening, not the sounds.

If a camera mechanically records light, why did people eventually accept photographers as artists? The United States Supreme Court faced a version of this question in 1884, over a portrait of Oscar Wilde. It held that the photograph was protected because the photographer had posed the subject, chosen the costume and setting, arranged light and shade, and drawn out the expression: an original intellectual conception, even though the camera made the image. ^[13]

If a producer samples an existing recording, which part belongs to the new artist? Not the sound itself. Perhaps the choice of fragment, the cut, the loop, the context it is placed in, and what it becomes next to everything else.

Creation is not the same as fabrication

These cases share a direction. Again and again, advances in art separate creation from simple manual fabrication.

The person who makes every physical component of a work is not always the person who authored it. Architects do not lay bricks. Film directors rarely hold the camera. Composers of orchestral music do not play every instrument. In each case, authorship lives in decisions: what to include, what to exclude, how to arrange it, and what it is for.

But this does not lead to the lazy conclusion that anything is art. If authorship lives in decisions, then the quality of those decisions matters. Selection, intention, context and transformation may all be necessary. A urinal in a showroom is not Fountain. A random fragment of a record is not a great sample. A person sitting silently at a piano for no reason is not 4′33″. And many experiments that took the same risks simply did not work.

A useful way to hold this together is the following:

Avant-garde is not necessarily a visual or musical style. It is the act of discovering a rule everyone was following without realising it was a rule.

Before Fountain, almost nobody would have said "an artwork must be made by the artist's own hands". It did not need saying. Before Bohemian Rhapsody, "a hit single needs a chorus" was not a law anyone wrote down. The avant-garde makes these rules visible by breaking them. Afterward, the rule is either abandoned, or it is kept, but now as a choice rather than an assumption.

This framing is an interpretation, not an established theory of art. It does, however, describe the recurring pattern in the cases above.

The Newest Frontier

The same argument is happening now, with a new tool at its centre: generative AI systems that produce images, music, video and writing from text instructions.

This is the unresolved version of the question, and it should not be treated as though history has already answered it.

The case against, taken seriously

Much AI-generated content is extraordinarily shallow.

A person can type a few words and receive a finished-looking image in seconds, with almost no thought behind it. This has produced an enormous volume of disposable images, songs and text, much of it posted because it cost nothing to make. Visual polish that once signalled effort now signals very little.

The outputs also tend to imitate. They reproduce recognisable styles, sometimes those of specific living artists, and they drift toward an averaged look: the most probable version of a request, smoothed and familiar.

There is some evidence for this averaging. In a 2024 experiment, writers who were given story ideas from a large language model produced stories that readers rated as more creative, especially among less creative writers. But those AI-assisted stories were also more similar to each other than stories written without help. ^[14] A large study of more than 50,000 users on an art-sharing platform found that adopting text-to-image tools increased output by about 25% and increased favourites per view by about 50%. Over time, the most novel works became more novel, while the average work became less novel. ^[15] A separate line of research shows that when models are trained on the output of earlier models, the rare parts of the original data tend to disappear. ^[16] These findings do not show that AI art is bad. They do suggest that "more" and "more varied" are not the same thing.

The concerns go beyond quality:

• Craft. Skills that took years to develop can appear to be bypassed in seconds, which can make the craft itself look less valuable.
• Labour. Illustrators, designers, musicians and writers face competition from tools that were built partly on work like theirs.
• Training and consent. Many models were trained on large collections of existing work, often without the knowledge or permission of the people who made it. The US Copyright Office has said that copying works into training datasets implicates the copyright owner's right of reproduction, and that whether such use is fair depends on the circumstances. ^[17]
• Authorship. It is often unclear how much of a result any human actually decided.

These are not reactionary objections. Some of them are about fairness and consent, not taste, and they would remain valid even if every output were beautiful.

The other side

It is also true that generative systems introduce a strange new creative instrument.

With it, a person can work through language, selection, iteration, composition and direction, instead of manipulating every pixel, note or sentence by hand. That may produce mountains of garbage. It may also make possible kinds of work that were not possible before, or that previously required skills, budgets or teams most people would never have.

The historical comparisons are tempting, and they should be made carefully, because they are not identical.

Photography did not destroy painting, although it took over much of painting's work as a record of appearances. Synthesizers did not destroy music. Sampling did not make composition meaningless. Digital editing did not eliminate cinema. Each did change what technical ability meant, and each let people create in ways that previously required entirely different skills.

AI may be a similar change. It may not. It differs from those earlier tools in important ways: it was trained on other people's work at an unprecedented scale, and it makes many more decisions on the user's behalf than a camera or a synthesizer does. That uncertainty is part of the subject, not a problem to be solved before the subject can be discussed.

Where the tool ends

Consider two people.

The first has an idea. They generate hundreds of variations, discard almost all of them, direct revisions, combine parts of different outputs, repaint sections by hand, control the composition, and finish with something that moves the people who see it.

The second types six words, accepts the first result, and uploads it.

These are probably not the same thing, even if both used the same tool.

A real case sits between them. In 2022, Jason Allen's Théâtre D'opéra Spatial won the digital art category at the Colorado State Fair. Allen said he had entered revisions and text prompts at least 624 times before selecting an image, which he then retouched in Photoshop and upscaled. ^[18] The US Copyright Office asked him to exclude the AI-generated parts from his copyright claim. He refused, and in 2023 the Office's Review Board declined to register the work, finding that it contained more than a minimal amount of machine-generated material. ^[18]

The Office's broader position, published in 2025, is that prompts alone do not currently give a user enough control over the output to count as authorship, while human-made elements, creative selection and arrangement, and substantial human modification can still be protected. ^[19] In the same year, a US appeals court confirmed that a machine cannot itself be the author of a copyrighted work. ^[20]

These are legal answers, and they are limited to the United States. They decide who can own something. They do not decide what counts as art. But they draw the same line the two people above suggest: authorship follows human decisions, and the question is how many decisions a person actually made.

Several questions disguised as one

Much of the public argument stalls because "Is AI art art?" is not one question. It is several, and they have different answers.

• Is the output an artwork?
• Is the human an artist?
• Can the model itself be an artist?
• Was the process ethical?
• Is the result original?
• Does it require skill?
• Does requiring less skill make something less of an artwork?
• Does wider access reduce the value of the result?
• Is intention enough?
• How much authorship is necessary?

A person can reasonably answer yes to one and no to the next. The output might function as an artwork while its training was unethical. The human might be an artist while the model is not. A work might require little manual skill and still reflect real judgement, or require none and reflect nothing. Collapsing these into a single verdict is the fastest way to stop thinking about any of them.

This article does not decide whether AI art "counts". The evidence does not decide it either.

Walking Ahead

Return to the advance guard.

The people at the front cannot know whether they are finding the next territory or walking in the wrong direction. That is not a failure of courage or intelligence. It is what being at the front means. The map does not exist yet.

History makes the avant-garde look obvious. Of course Cubism mattered. Of course Fountain changed the rules. Of course a six-minute operatic rock single could become one of the best-known songs in the world. But that confidence is produced by the filter. We remember the experiments that led somewhere, and we forget the thousands that did not, including many that were just as daring and just as sincere.

What we know is that the boundaries of art have moved repeatedly, and that they have usually moved toward separating authorship from manual fabrication. What history permits is the possibility that current tools belong to that same pattern. What remains unresolved is whether they do, and which of the things made with them will matter.

Somewhere today, people are using tools and making things that much of society regards as ugly, lazy, fake, meaningless, or not art at all.

Most of it may deserve to be forgotten.

Some of it may not.

The frustrating thing about living in the present is that we do not yet know which is which.

References

[1] The Museum of Modern Art. "Yoko Ono. Cut Piece. 1964." MoMA Learning.
https://www.moma.org/learn/moma_learning/yoko-ono-cut-piece-1964/

[2] Tate. "Avant-garde." Art Terms, Tate.
https://www.tate.org.uk/art/art-terms/a/avant-garde

[3] Official Charts Company. (2025). "Celebrating 50 years of Queen's Bohemian Rhapsody." Official Charts.
https://www.officialcharts.com/chart-news/queen-bohemian-rhapsody-50-years/

[4] Philadelphia Museum of Art. "Marcel Duchamp and the Fountain Scandal." Philadelphia Museum of Art Press Room.
https://press.philamuseum.org/marcel-duchamp-and-the-fountain-scandal/

[5] Tate. "Cubism." Art Terms, Tate.
https://www.tate.org.uk/art/art-terms/c/cubism

[6] Tate. "Dada." Art Terms, Tate.
https://www.tate.org.uk/art/art-terms/d/dada

[7] The Museum of Modern Art. "John Cage. 4′33″ (In Proportional Notation). 1952/53." MoMA.
https://www.moma.org/slideshows/24/389

[8] Baudelaire, C. (1859/2018). "The Salon of 1859: The Modern Public and Photography." In Modern Art and Modernism: A Critical Anthology, 19–22. Routledge.
https://doi.org/10.4324/9780429498909-6

[9] United States District Court, S.D.N.Y. (1991). Grand Upright Music Ltd. v. Warner Bros. Records, Inc., 780 F. Supp. 182.
https://law.justia.com/cases/federal/district-courts/FSupp/780/182/1445286/

[10] United States Court of Appeals, Sixth Circuit. (2005). Bridgeport Music, Inc. v. Dimension Films, 410 F.3d 792. Music Copyright Infringement Resource, George Washington University Law School.
https://blogs.law.gwu.edu/mcir/case/bridgeport-music-v-dimension-films-et-al/

[11] Adajian, T. "The Definition of Art." Stanford Encyclopedia of Philosophy.
https://plato.stanford.edu/entries/art-definition/

[12] Danto, A. (1964). "The Artworld." The Journal of Philosophy, 61(19), 571–584.
https://doi.org/10.2307/2022937

[13] Supreme Court of the United States. (1884). Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53.
https://supreme.justia.com/cases/federal/us/111/53/

[14] Doshi, A. R., & Hauser, O. P. (2024). "Generative AI enhances individual creativity but reduces the collective diversity of novel content." Science Advances, 10(28), eadn5290.
https://doi.org/10.1126/sciadv.adn5290

[15] Zhou, E., & Lee, D. (2024). "Generative artificial intelligence, human creativity, and art." PNAS Nexus, 3(3), pgae052.
https://doi.org/10.1093/pnasnexus/pgae052

[16] Shumailov, I., Shumaylov, Z., Zhao, Y., Papernot, N., Anderson, R., & Gal, Y. (2024). "AI models collapse when trained on recursively generated data." Nature, 631, 755–759.
https://doi.org/10.1038/s41586-024-07566-y

[17] U.S. Copyright Office. (2025). Copyright and Artificial Intelligence, Part 3: Generative AI Training (pre-publication version).
https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf

[18] U.S. Copyright Office Review Board. (2023). "Re: Second Request for Reconsideration for Refusal to Register Théâtre D'opéra Spatial." U.S. Copyright Office.
https://www.copyright.gov/rulings-filings/review-board/docs/Theatre-Dopera-Spatial.pdf

[19] U.S. Copyright Office. (2025). Copyright and Artificial Intelligence, Part 2: Copyrightability.
https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf

[20] United States Court of Appeals, D.C. Circuit. (2025). Thaler v. Perlmutter, No. 23-5233.
https://media.cadc.uscourts.gov/opinions/docs/2025/03/23-5233.pdf