Dear readers,
A quick note before I hand over.
This week’s issue is not mine comes to us from my good friend Fred, of Fred & Yang, who writes songs for a living and has been thinking hard about what AI is doing to his craft. Last week’s issue of The Compute ended with “taste” as the one thing that cannot be automate; it felt right to let someone who works in taste all day pick up the thread. Over to Fred.
A song starts playing at the Swedish bar where I am working. It gives everybody a nostalgic trip. Swedish music at its best. It sets the atmosphere perfectly for all of the seniors eating meatballs and enjoying a midday glass of wine. The singer is singing in beautiful Swedish, bringing everybody back to midsummer in the countryside, youthful romances, close to the heart, hard to dislike.
Unknowingly to them all, this song,”Jag vet, du är inte min” by Jacub, made headlines in the Swedish music community. It was kicked off the Swedish Billboard chart because it was created with AI.
Despite the fact that the composers and songwriters behind it, a Danish company working with a Swedish singer under a pseudonym, had to write a long letter explaining that they had already done most of the work themselves: the lyrics, the melodies, the compositions. They had only used AI for polish. The damage, however, was already done. To many, it was simply cheating.
However, the song lives on in consumers’ minds as a beautiful contribution to Swedish music.
I don’t know what struck me the most: the fact that AI was seen as having claimed the credit, although the piece was clearly written by humans in all the important ways; or the fact that those who were loudest in their criticism were often musicians who had not really engaged with these tools themselves. If they had, they would know that what was produced was not some simple text-to-music output, but something much closer to a hybrid process, music shaped by human intent, with machine assistance in refinement rather than origin.
What stayed with me most, however, was the possibility that this is, in fact, one of the most “human-like” productions we have seen in a long time.
Autotune has made it possible to correct pitch to near-perfection, producing a sound that often moves away from natural performance.
Quantisation has likewise allowed musical phrases and percussive elements to be placed perfectly on the grid, locked to the metronome with mechanical precision.
The result of modern music production, even before AI, is in many cases closer to machine logic than to human imperfection. Not always, of course, but often enough that the aesthetic baseline has already shifted. The paradox is that AI, depending on how it is used, can reintroduce variation, phrasing irregularity, and non-standard timing in ways that feel closer to human performance than overly corrected digital production.
Since the 1980s, the art of music engineering has steadily become more valuable than formative artistic knowledge.
The ability to draw, write lyrics, and compose, the fundamentals of artistic creation, has, in many corners of the industry, become less important than the ability to produce, engineer, and package output at scale. Right now, we may be returning to a situation where production itself is becoming less central, while human creativity and direction become more valuable again. The advantage shifts toward those who can generate ideas and aesthetic direction, not only those who can technically execute them. In other words, those who can originate musical ideas from scratch regain importance in the creative process.
Most people say that AI will eventually learn human traits and surpass humans at being human. I don’t think so. The progress of music is non-statistical. It is saying something, playing something, or drawing something in ways that have not been done before. AI is still based on what is probable, not what is improbable. That is where the unique experience of a human life becomes unnecessary to an AI model, but highly essential for producing genuinely unfamiliar perspectives.
I’d add one more thought, one that is perhaps more philosophical than musical. Both consciousness and AI can be thought of as self-referential systems. In that sense, they share a structural similarity. But they are self-referential in fundamentally different ways.
An AI model is ultimately confined to information that has been translated into data. Everything it knows (actually merely statistically predicts) has first been described, recorded, labelled, or otherwise made legible by human beings. Its world consists of our representations of reality.
Human beings begin somewhere else entirely. Our first source of knowledge is not language or data, but reality itself. We encounter the world before we describe it. We live through experiences that often resist explanation until much later, if they can be explained at all.
That difference matters for art. The purpose of art is not simply to rearrange what is already known, but to notice something that has not yet been expressed in quite that way. It is a conversation with reality before it becomes a conversation with culture.
This is why I find AI remarkable as a technological achievement, while still seeing it as exactly that, a remarkable tool. It cannot step outside the world we have already translated for it. It cannot encounter reality except through us.
Perhaps what concerns me more than AI itself is that we increasingly experience the world through digital representations. We risk mistaking the map for the territory. Technology has become so immersive that we sometimes forget we remain in direct contact with something far stranger than any model can capture, a universe that is still largely mysterious, and whose truths may not ultimately be technological at all.
I don’t believe we are at the cusp of erasing the uniqueness of being alive. I do, however, believe that those who predict AI as the doom of the world are also contributing to the AI economy by amplifying cases like this nostalgic Swedish song as examples of AI-generated music, even when much of the creative credit still belongs to human work and intention. Perhaps the greatest irony is that, in trying to warn us about AI replacing artists, they end up erasing the artists who were there all along.
Fred from Fred & Yang
After hours
This week I had an interesting encounter where I met a booker who had the exact same favourite song. Big Iron by Marty Robbins, outlaw country at it’s finest and we started singing it acapella in public. It gained me access to a group called Stockholm Cowboys, a little community with people who like country music. Now looking for a hat that works in that group 😉.