There is a painting at the Louvre that few visitors notice.
The Wedding Feast at Cana by Paolo Veronese is the largest canvas in the museum, nearly 22 ft tall and 33 ft wide. In it, 132 figures crowd a scene of such joy that it feels like an artistic miracle.
In theory, it is unignorable, hung floor to ceiling in the museum’s most-visited gallery.
In practice, most visitors simply rush past it, because a few metres away, in the same room, hangs the Mona Lisa.
This is not about which painting is better. It’s about the power of recommendation.
Reading about AI reminds me, time and again, of that room at the Louvre, and The Wedding Feast at Cana, a masterpiece, rendered invisible in plain sight.
Artificial intelligence is set to crowd every room we care about, to a greater degree than we can estimate. Cana vs Mona Lisa feels, to me, like a preview of what AI will bring about in the fields of writing, music, film, advertising… and every other creative endeavour.
What might this world look like?
***
There is a 1975 film called The Irony of Fate, broadcast on Russian TV every New Year’s holiday, whose humour is rooted in the dreary uniformity of state-mandated urban architecture.
In the Soviet Union, apartment blocks were so standardised, street names so interchangeable, that a man, drunk and confused, could board the wrong plane in Moscow, arrive in Leningrad, hail a taxi to what he thought was his address, use his key, find that it worked, and even fall asleep in someone else’s home, without noticing, because even the furniture was the same.
In the creative world, thanks to AI, the Irony of Fate problem has arrived.
On the music-streaming platform Deezer, about 44% of all daily uploads are AI-generated tracks. A blind test found that 97% of surveyed listeners could not reliably distinguish them from human-made recordings.
In film, Dreams of Violets, an AI-generated docudrama set in Iran and screened at this year’s Tribeca Festival, turned out to have been produced for $2,000, without actors, sets or crew.
Last year, the auction house Christie’s hosted an Augmented Intelligence auction dedicated to AI art. In the world of online marketing, major brands have announced that they are using AI to shrink content production cycles from several days to a few hours.
Call it average-garde. Not avant-garde, the experimental, difficult and strange that worked to reorganise one’s thinking after at first baffling the viewer.
Average-garde is technically fluent, aesthetic — and culturally indistinguishable. The technology that promised to democratise creativity may instead be McDonaldising it.
***
There is growing evidence of this flattening McCulture effect.
Researchers at University College London and University of Exeter, in a study published in 2024, found that when writers use AI assistance, individual stories improved — were rated more creative, better-written and more enjoyable — for those who were less skilled to begin with.
Meanwhile, the stories became more similar.
A meta-study by University of Munich researchers found that, among writers, human-AI collaboration improves individual performance but produces a large negative effect on collective idea diversity.
Gone are the chaotic range of ideas, raw emotion, and strange but interesting turns of phrase. All output begins to chime in the same note, bear the same texture, and the same flat but polished confidence.
In an added blow to creative range, researchers have found that AI writing suggestions pushed non-Western participants’ prose toward a generic, Westernised norm, and flattened cultural nuance. A study of music-generation datasets found that these AI models had been trained, overwhelmingly, on music from the global North.
In other words, the machine produces a specific, historically dominant kind of sameness.
An average, with an accent that is not yours.
***
Travel writer Pico Iyer, in his 2001 book Global Soul, observed that all airports have become one airport, endlessly reproduced: the same duty-free perfume counter, the same business lounge, the same news screens.
Going somewhere new had acquired the uncanny texture of going nowhere in particular.
Again, I am reminded of artificial intelligence algorithms, and their articles, songs and images. The degradation is not of the obvious kind. This is not a world of spam, misfiring bots or garbled translation.
The problem is of a competent adequacy that accretes, without adding up to anything.
Most people already feel this, though we don’t have a name for this verbose emptiness yet (“nulloquence”, anyone?).
There is something in the cadence of AI writing, the slightly formal warmth, the rhetorical questions, the conclusion that wraps too neatly around a thesis announced too early, that triggers recognition.
Ironically, content platforms run by the same tech giants investing heavily in AI recognise this too. Meta and Google algorithms actively down-rank AI-generated content in recommendations.
We buy their product, but they don’t want it back.
***
The comparison with earlier disruptions deserves fair hearing.
Photography, it was said, would kill painting. The printing press would destroy the authority of the written word. Each time, the old form reinvented rather than died.
Photography liberated painting from the duty of likeness and drove it toward Impressionism, abstraction, Cubism; forms that could only exist after a machine took on the task of generating likeness.
Should we worry less? I don’t believe we should.
The camera made images cheaper. It did not train itself on a wealth of existing images and begin to churn out art of all kinds. The printing press reproduced text; it did not absorb texts already written and emit new ones within seconds.
Previous technologies replaced one function of creativity.
AI feeds on entire art forms — novels, scripts, music scores, films, images — to encourage us to ask rather than create.
***
What we stand to lose in all this is human specificity: the particular dissonance, perspective, distortion, fear, misunderstanding or obsession that feeds each artist’s art.
What we stand to lose is the kind of art MF Husain created when he arrived for a talk at my alma mater Indian School of Business, years ago, and decided not to speak but to paint.
No brushes were available; no one had planned for this. So, he dipped his fingers in acrylic pigments and painted on a flip chart. The work still hangs near the entrance of the Hyderabad institute.
What it records is irreducible: a particular afternoon, an artist yielding to impulse and adapting to constraint, a mind so unusually talented, and a talent so shaped by decades of practice, that it could fold the absence of its tools into the art itself.
Sure, we can have slickness and fluency with AI. But I believe it will erode, minimise or at best marginalise novelty and genius. It has no use for them.
***
AI is useful across many domains. This deserves to be said plainly, without the hedging that usually follows.
Take language alone. For user manuals, rapid research synthesis, accessibility tools, medical translation, and the thousands of other forms of functional writing that the world needs in vast quantities, it is extraordinary. It is efficient, clear and consistent. These are real virtues.
But creativity with lasting resonance has historically been what we make when we take efficiency out of the equation.
The sweetest songs are about the saddest thoughts. The most remarkable novels are written by people who had to live decades to assimilate what they were seeing and feeling.
What threatens creativity most is not AI output that is obviously bad, but AI output that is fine enough to pass, to fill the space.
Fine enough that The Wedding Feast at Cana sits in the same room, magnificent and ignored, while crowds cluster around the thing the algorithm “created”, and then pre-decided they should see.
(Click here for more on what it will take for humans to fight back.)
(Kashyap Kompella is a tech industry analyst and author of three books on AI)
