Part 2. What You Can't Hear in a Song

The Life Behind a Song

What lies behind music in the age of AI

Imagine two songs.

I like both of them. They evoke roughly the same emotions. Listening to them, I can't tell which one was written by a human and which one emerged after a few prompts fed into a music generator.

Then someone tells me the story behind them.

The first song took six months to write. Its creator threw away verses, changed the melody, argued with other musicians, once decided to delete everything, then came back a week later. Maybe the song was born out of heartbreak. Maybe none of that happened, and the songwriter simply invented a compelling story.

The second one appeared almost instantly. Forty seconds.

If I listened to both songs blind, I might not notice any difference. But the moment I learned how each of them came into existence, the difference would suddenly feel enormous. Even though the sound stayed exactly the same.

Would that change how I feel about them?

It's an uncomfortable question because there is no right answer. Yet the music industry is trying to find one right now.

On September 25, British producer Fred again.. shared several Instagram Stories about AI music. One line has stayed with me ever since: "The opposite of AI music maybe isn't human music. It's making music."

For Fred, what matters is the process itself: the agonizing attempt to catch up with an idea that so far exists only in your head. That's why he has no interest in handing that part over to a machine. "The process is the magic," he wrote. DJ Mag

Fred isn't saying a machine can't make a good track. His argument lies outside the finished recording. It's about what happens to a person while they are making it.

We are used to judging creativity by its results: the painting hangs on the wall, the novel sits on the table, the song plays in our headphones. Everything that came before is usually hidden from view. Generative technologies have suddenly made that hidden layer a matter of public argument.

How Much Human Is Enough for a Grammy?

A few days later came a case that looks almost lab-made.

French DJ and producer HUGEL said that his track Movin' To The Sun had been removed from Grammy consideration because of questions about AI use. What interests me most is the argument about where the line should be drawn.

HUGEL's team does not deny that AI was involved in the track's early stages. But according to the producer, he, Maximilian Riehl, and Loris Cimino substantially rebuilt the final recording in Ableton, reworking instrumental parts, programming drums, and changing the arrangement. The Academy has not publicly given its reasons. Now HUGEL is asking the Recording Academy to explain which specific elements of the final recording raised the issue. DJ Mag France

The Academy's rules look reasonable enough. Only humans can receive a Grammy. Using AI-generated material is not prohibited. But the human contribution must be "meaningful and more than de minimis": substantial rather than merely symbolic, and relevant to the category in which the work is submitted. Recording Academy Rulebook

The problem begins when someone tries to apply them.

What exactly is a "meaningful human contribution"? The boundary slips away as soon as you try to draw it.

A person writes all the lyrics, but the music is generated. Is that enough? A person comes up with the harmony, but AI creates the arrangement?

Someone receives a generated track and spends three days rebuilding it? Chooses one result out of a hundred generations? Changes two bars?

And then there is the most uncomfortable scenario of all:

A person plays nothing and writes nothing in the traditional sense, but spends a week revising prompts hundreds of times, rejecting outputs, clarifying the idea, and eventually arrives at precisely the work they envisioned.

Are they the author?

Or have we simply not invented a better word for that role yet?

Rick Rubin: The Idea Matters More Than the Execution

Yet where Fred sees the magic itself, Rick Rubin sees only one possible path toward it.

In a recent appearance on the Diary of a CEO podcast, Rubin said that AI could let people create art even if they can't paint or play an instrument. In his view, the creative act lies primarily in the idea, and a prompt followed by careful iteration may simply be another way of bringing that idea into the world. He compares it to filmmaking: a director is not expected to personally perform every task required to make a film. Business Insider

It's hard to wave that away. We have never demanded that creators do everything themselves. Architects don't lay bricks. Directors don't play every role. Producers may never touch an instrument. Modern pop songs often pass through dozens of hands: songwriters, arrangers, session musicians, engineers.

Why should a machine strip a human idea of its creative power?

The answer, "because the person can't play an instrument," feels weak. The history of music is full of technologies that let people do things that once required a different level of skill: synthesizers, sequencers, drum machines, samplers, Auto-Tune. Every time, someone declared that it was no longer "real." Music survived.

But with generative AI, one thing is new. In the past, a tool extended what a musician could do. Now it can perform the action itself. The line between "I made it with a tool" and "the tool made it and I chose the result" is becoming very thin.

Does the Listener Even Need to Care?

There is a simpler response: I like the song. Nothing else matters.

After all, I rarely know what happened in the studio. I don't verify whether a singer truly lived through the heartbreak they sing about. A novelist can write convincingly about murder without killing anyone. An actor doesn't need to be the Prince of Denmark to play Hamlet. Here the argument about "authentic lived experience" starts to wobble: we don't demand biographical authenticity from every emotion in art.

And yet the origin of art somehow continues to matter. I wrote about this in the previous essay in the series: U2 returning to their old school, and a playlist for a stepdaughter. The sound stayed the same.

The meaning did not.

Two Different Kinds of Value

Perhaps we have been blending two different values: the value of the work itself and the value of the human journey that led to it.

They need not coincide. An AI-generated song may be brilliant. A human-made song may be mediocre. Declaring the second one better merely because its creator suffered longer would be strange. That would become a new kind of artistic snobbery, where "real" artists are expected to spend seven years learning their craft, starving in an attic, and ideally living through at least one heartbreak, and anything else is cheating.

But the opposite position, that only the final result matters, feels too simple as well. Because then we would have to accept that the creative process never had any value of its own.

That rehearsals were just an inefficient way to get to a concert. That ten failed versions of a song were simply production costs. That everything important happens only when we press Play.

Fred again.. argues the opposite: those failed attempts are part of the reason he makes music in the first place.

I understand him.

We May Soon Need Provenance

When photography went digital, nobody felt the need to label images "taken by a human." When synthesizers arrived, records didn't carry stickers announcing electronic instruments.

With generative art it may go differently. Its output isn't necessarily worse, but the way a work comes into existence becomes information that changes how we perceive it.

The first attempts to formally define "meaningful human authorship" are already here; the Grammys are having to draw that line right now. Grammy.com

But rules solve only part of the problem.

I can listen to a song and only later discover that nobody wrote it. Not a bad musician. Not someone without formal training. Nobody, in the usual sense. A model built the melody, a voice performed the lyrics, an algorithm mastered the track, another model drew the cover. A human typed a few lines of prompt and picked version number seventeen.

If I loved the song, should that fact change anything?

I don't know yet. And honestly, that uncertainty bothers me. But I do know this. Until now, we have almost automatically assumed that there was someone on the other side of a work of art. Someone who wanted to say exactly this. Someone who made choices, made mistakes, had doubts, began again. Someone who spent a day or a year of their life bringing that thing into existence.

The generative era makes that assumption optional. And the question that remains is this: if the journey to the song is no longer necessary, what do we lose with it?

Perhaps we won't lose the music. But we may lose the feeling that there was someone behind it.

Maybe what was always scarce was the life behind the song.

After login, everything only begins.