A reflection on the meaning of the work, on authority and on what no machine can ever know: death, loss, expectation, risk, imperfection. In one word: life.
by Marialuisa Veneziano
4 August 2026
Prelude. Four jokes against ninety thousand songs
In any room, a boy repeats four Bach jokes for the fiftieth time.
The right hand proceeds with a certain safety; the left, however, continues to swallow an internal voice. The master stops him. He tells him to listen to the tenor, not to treat it as a service corridor between the bass and the soprano. The boy starts again. She's still begging. It irritates, breathes, erases a fingering, tries another. The four jokes are always there, intact and inflexible, with that strict courtesy that Bach reserves to those who believe he understood it too soon.
At the same time, without getting tired, without discouraging and above all without having to pay the lesson, an artificial intelligence system can generate a complete song: text, melody, voice, accompaniment, arrangement, production. A few seconds, some well-formed indication and the track is ready. Perhaps it will not be memorable; but, moreover, not all that is made up of human beings has the obligation to be. Humanity has also produced many songs that would have been merciful to entrust to oblivion, and we cannot accuse the algorithm as well as those.
The problem, therefore, is not that the machine can generate music.
He can.
And it begins to do so in quantity that no compositional workshop, no monastery, no conservatory and no multinational record could have imagined. In June 2026, Deezer recorded about ninety thousand pieces of artificial intelligence each day at the platform: more than half of all new daily uploads. It is a data relating to a specific service, not to the entire musical universe; but just describe the size of the phenomenon.
Ninety thousand songs.
Every day.
While the boy is still standing at his four bars.
Disproportion is almost comic. On the one hand, human time, with its slowness, nerves, imperfections and the blessed inner voice that does not want to come out. On the other hand an uninterrupted sound production, capable of filling in a few hours what a composer could not write in many lives.
Yet the decisive question is not how much music a machine can produce.
The demand is older, more uncomfortable and much less technological:
when a succession of sounds becomes a work?
Exhibition. Everything that the algorithm knows
The algorithm can know a lot about Bach.
It can recognize an escape, identify the subject, follow the answers, distinguish a fun from a tight, observe the opposite motion, calculate the counterpoint density, compare thousands of cadences, catalog progressions, imitate figurations, intervals and conducts of the parts.
You can learn that in a polyphonic writing each voice must possess its own direction. It may associate certain harmonic procedures with specific perceptive effects. It can generate a page that has the sound appearance of a coral, a prelude, perhaps even a plausible escape.
It's not very small.
It would be intellectually dishonest to deny it only because it disturbs us. The history of music, moreover, is full of instruments initially looked with suspicion: the piano seemed to someone less noble than the harpsichord; the metronome promised to turn time into a notary practice; the recording seemed destined to embalm the execution; the synthesizer was welcomed as if he had personally intended to fire the orchestra.
The orchestra, fortunately, survived. Even though she keeps paying as if she didn't.
Every technology changes the relationship between man and sound. Sometimes it impoverishes it, sometimes it expands it; almost always it does both, because history does not have the good taste to present itself divided in good and bad with the labels already applied.
Artificial intelligence can help a composer explore timbre combinations, organize materials, transcribe, restore recordings, separate sound sources, improve accessibility or free time from repetitive operations. It can become an extension of the workshop, such as music printing, electronics and notation software.
It is not the tool to create the problem.
The problem begins when the tool does not expand a creative responsibility, but simulates its existence.
A piano does not claim to be the pianist. A diapason does not claim to have plastered the orchestra. The metronome, however authoritarian, has never attributed the interpretation of a sonata.
The algorithm knows all the notes. It can calculate, predict, imitate. But he never waited for a voice that never comes back. And there is a weight, in that note, that no code will ever endure.
The generative system, however, can give us a finished product: a voice that seems to have suffered, a text that tells a separation, a melody that suggests nostalgia. And he can do it without anyone suffering, separated or longed.
Formally, everything is in place.
Only the place where that music was supposed to come from is missing.
Counterpoint. Bach did not know death by hearing
Bach is not great because he possessed many musical information.
It is great because that information, through a human being, became necessary.
She remained a very young orphan: her mother died in 1694 and her father the following year. In 1720, returning from a trip, she learned that the first wife, Maria Barbara, had died after a short illness and had already been buried. Death did not represent for him a poetic concept to be selected in a menu of atmospheres; it was a concrete presence, entered several times in the house, in the family and in the flesh of time.
Of course, we must not commit the romantic mistake of reducing every page of Bach to a encrypted autobiography. An escape is not a personal diary with four voices, and a choral is not necessarily a psychoanalytic session with pedals.
Bach’s greatness lies also in the exact opposite of individual performance: in the ability to transform private experience into a universal form, to subtract pain to the simple rash and deliver it to an order.
His counterpoint does not cry because he accumulates musical sighs. He is crying because each voice is free and tied to others; because the disorder is crossed without being denied; because dissonance is not an accident to be erased, but a tension that demands fulfillment.
In Bach, suffering does not abolish architecture.
It makes it necessary.
This is what we often forget when we talk about musical emotion. We believe that emotion consists in making everything slower, stronger or more tearful. It is the old strategy of those who, not knowing how to open a door, try to move it on shoulders.
Authentic emotion is not a seasoning poured over form. It comes from the form itself: the expectation of a resolution, the memory of a theme, the distance between a hue and its return, the late entry of a voice, the silence that prepares an agreement.
A machine can analyze these reports.
Can reproduce them.
It can even combine them in ways that surprise the listener.
But he didn't wait for them.
And it is waiting, perhaps, that the human being begins to become author.
Interlude. The Great Food Supermarket
We have ahead of a technology capable of generating, on request, "a melancholic piece in baroque style, with piano, strings and atmosphere of loss".
The pain finally became customizable.
You can choose the duration, gender, type of voice and perhaps, one day, even the exact number of tears per line. "I would like an elegant mourning, thank you, but not too tragic: I have to use it as a background to a video of thirty seconds."
It's easy to iron. The harder it is to admit that the machine did not invent this trade: it has already found it ready. For years a part of the music industry has not asked what to say, but in which category it can be placed. Music to relax, to concentrate, to fall asleep, to cook, to forget to be cooking. The song is no longer met: it is used.
Artificial intelligence then arrives in a world that had already begun to consider music not as a language, but as an acoustic tapestry.
The real danger is that the machine becomes similar to Bach. It is that we become listeners unable to distinguish Bach from a sufficiently pleasant sound surface.
A research published in June 2026 examined the spread of artificially generated music on platforms, describing a model based on massive production and statistical hope that some traces gain visibility. According to the study, 93 percent of the songs identified as generated by AI received very few listenings or none; the authors also experienced how simple it was to distribute music produced in series through policies often inconsistent or poorly applied.
So we are not only before a creative revolution.
We are before a quantitative revolution.
And the quantity, when it becomes absurd, does not necessarily widen freedom. You can submerge it.
An infinite library where no one can find a book is no longer the paradise of knowledge. It is a warehouse with excellent ambitions.
Escape. Who is the author?
The author’s question is not resolved by asking who pressed the button.
Even the pianist presses keys; but no one, at least so far, has attributed to the keyboard the interpretation of Goldberg variations.
The author is not simply the one who initiates a process. He is the one who takes responsibility for the choices: he accepts, rejects, modifies, orders, renunciation. Above all, it recognizes the difference between what works and what must exist.
Here the decisive word is need.
A work is not born only because it can be produced. It was born because, for someone, not producing it would have been a form of untolerable silence.
This does not mean that the author should be tormented, sick or unhappy. Let us be free once and for all from the superstition that to write a good adage it is necessary to have at least one family tragedy, three creditors and a precarious health. Mozart knew how to laugh; Rossini knew the pleasure; Haydn had irony and discipline.
The artistic necessity does not coincide with suffering. It fits with the presence.
Someone's there. Decide. Listen. It's a failure. He exhibits a part of himself in the judgment of time.
The generating system risks nothing. He does not feel shame on a mediocre page, he does not fear having repeated, he does not remain awake wondering whether a modulation is inevitable or only elegant. You should not defend a choice in front of a director, editor or particularly ruthless court that is your ear the next morning.
It can generate a thousand solutions and abandon them all without loss.
The human being, however, also chooses because it does not have an infinite time.
Every authentic work contains consciousness, declared or secret, that we cannot write everything, play everything, love everything, live everything.
The shape is also born from the limit. And it is perhaps for this reason that Bach knew death not only as theological argument, but as a musical principle: every note must cease because the next one can buy sense; every sentence must end; every cadence is together arrival and leave.
The algorithm can continue indefinitely.
Man composes knowing that he will have to stop.
Cadence of the interpreter. The sound that happens once
There is then a figure often forgotten when talking about generative music: the interpreter.
Our time seems to believe that the score already contains everything and that the execution is a kind of home delivery service. The composer prepares the package, the interpreter rings the bell and the audience signs the receipt.
But a score is not yet music. It's a promise of music.
Between the sign and the sound there is a territory made of weight, breath, joint, stamp, agogical, space and silence. Two pianists can respect the same heights, the same values and the same indications, however producing two completely different truths. Not because one arbitrarily adds his biography, but because each interpretation is a form of embodied knowledge.
When an organist prepares a Bach escape, it should not only identify the most suitable registers. It must understand the acoustics, the delay of sound, the response of mechanics, the breathing of phrases. It must know when an internal voice risks disappearing, when the pedal supports and when, instead, it becomes a respectable liturgical hippopotamus.
A pianist must decide how much weight to entrust to a finger without breaking the line; a singer must turn his breath into a word before the body complains; a director must let many people breathe as if they were one organism, without reducing them to a well synchronized pack.
All this is not a fault of notation. It's his miracle.
The score does not completely prescribe music because it leaves the interpreter with the task and responsibility to make it exist.
Artificial intelligence can simulate stolen, hesitation, dynamic variations and even those small irregularities that we associate with human presence. But one thing is to plan imperfection, another is to risk it.
The human interpreter does not produce a difference every time so that he has been asked for a percentage of variability.
He produces it because he's alive.
Because that night the acoustics are different. Because the public breathes in another way. Because a sentence, crossed a hundred times, suddenly reveals a wound that was not audible before.
And sometimes because, between one note and the other, it reminds someone.
Nobody notices. Yet time changes.
An attack is held for an almost imperceptible fraction; a voice is accompanied longer; a cadence does not close, but seems healthy.
That instant is not written in the score.
It cannot be proved.
But that's where music stops being correct and becomes true.
recapitulation. Not against the machine, but against the waiver
It would be reassuring to conclude that artificial intelligence produces only soulless imitations and that human beings, instead, create invariably palpitating masterpieces.
Unfortunately, reality, with the usual lack of respect for our thesis, is more complex.
A machine can produce an interesting piece. A human being can compose a frighteningly empty one. Organic origin does not guarantee quality, as well as the use of technology does not automatically erase autorality.
The important distinction does not pass between natural and artificial.
Pass through liability and waiver.
AI can be used by an author as material, interlocutor, research tool or device capable of opening unexpected possibilities. In this case, it does not necessarily replace creativity: it can force it to redefine its boundaries.
But when man delegates to the machine not only the execution of a task, but the very weight of choice, he gets perhaps an effective product; it is not said that he obtains a work.
Institutions are also trying to trace this border. In July 2026, IFPI introduced global principles to establish eligibility in official record rankings developed with generative systems. In the United States, the Copyright Office stated that expressive elements determined by a human being can receive protection, while the simple formulation of a prompt is not, alone, sufficient to found authority.
The right always comes after art, with the dignified affection of those who enter the theatre when the overture has already begun. But the legal question contains an essential philosophical question:
How much human presence is necessary so that someone may say: is this work mine?
The answer will not only depend on the number of commands given. It will have to cover the quality of control, the transformation of the material, the awareness of the choices and the responsibility of the result.
It is not about protecting a romantic privilege of the artist.
It is a matter of preventing the word "creation" to be reduced to an elegant synonym of "generation".
Coral. What remains after the last note
Perhaps one day the algorithm will imitate Bach so well that we will no longer be able to distinguish the copy from an authentic composition. Perhaps it will generate impeccable escapes, singing touches, corals capable of lowering our eyes. Maybe he'll know us enough to build the exact music for all our nostalgia.
But even then there will be a difference. Not necessarily audible difference.
A difference of origin.
Bach did not compose because he had all the necessary data to compose Bach. He composed because he was a man thrown over time, forced to measure himself with faith and doubt, discipline and fatigue, birth and loss. His music is not true because we can reconstruct every episode of his life in every joke. That is because a mortal being managed to limit a form that still surpasses us today.
The algorithm can know all the notes.
It can calculate relationships between them, predict our reactions, imitate the sound of consolation.
But he never waited next to a bed.
She's never heard too late.
He never gave a name for the last time without knowing it would be the last.
He never heard a voice continue within himself when in the world he had already silenced.
That is why we must not defend human music by claiming that the machine will never be skillful enough. It could be. And perhaps, in some fields, it is already more than we are willing to admit.
We have to defend her for a deeper reason.
Because art is not only the result we reach.
This is what happens to us while we try to reach him.
He is the boy who repeats four jokes and slowly learns to listen to a voice that he ignored before. He is the composer who deletes a technically perfect page because he does not recognize it as true. It is the interpreter who, in front of a cadence, suddenly understands that concluding does not simply mean ending.
Artificial intelligence will force us to wonder what is really human in music.
It's not a threat recently.
But it could also be a gift.
He will force us to stop calling art all that is only well packed. It will force us to distinguish competence from necessity, emotion simulated by transformed experience, production from presence.
And maybe, after generating billions of songs, we will find that one note played by someone who really has something to say keeps a different weight.
Not because it's perfect, but because it can break.
Not because he knows the pain, but because he crossed it.
Not so he can imitate the crying.
But because, before he became music, he had to learn to cry.
Leave a comment