The Machine in the Music

Where AI came from, what it's changing, and why I still play every note

---

Let me be honest with you before I say anything critical: what these AI tools can do now is genuinely impressive. I build my own guitars in a workshop in Zurich, I play and sing and produce almost everything myself, and even I sat there the first time thinking, "okay, that's remarkable." So this isn't going to be a rant. AI is going to change how music gets made, whether I like it or not. What I want to do instead is understand it properly — where it came from, what it's really doing to musicians and listeners, and why, after looking at all of it, I still do everything by hand.

Let's start by killing one myth: that AI music appeared out of nowhere a couple of years ago.

A short history of the machine in the music learning to sing

https://www.youtube.com/watch?v=v5z3t2qz4qo

The first piece of music composed by a computer is nearly seventy years old. The Illiac Suite, written in 1957 on a room-sized machine at the University of Illinois, used mathematical rules to generate a string quartet. In the 1980s, a composer named David Cope built a system called EMI that could study Bach or Chopin, learn their fingerprints, and write new pieces good enough that audiences couldn't reliably pick the human from the machine.

Then machine learning showed up and everything sped up. Google started its Magenta project in 2016. Sony made a Beatles-style tune called "Daddy's Car" the same year — though a human still had to arrange it and write the words. In 2020, OpenAI's Jukebox produced raw audio with actual singing, rough as it was. And then the tools went consumer: Suno opened to everyone in December 2023, Udio a few months later, spitting out a fully mastered track in about forty seconds from a line of text. Seventy years of slow research turned into an overnight flood. Suno alone is now worth more than five billion dollars.

How it actually works — and why you can still hear it

Here's the plain version, no magic. These systems are fed an enormous pile of existing music, and they learn the patterns of it: what tends to follow what, what a certain kind of voice or groove is shaped like. You type a prompt, and the model assembles new audio that fits those patterns. It isn't feeling anything. It's producing a very clever statistical match of a prompt against everything it was trained on.

And honestly, right now, you can often still tell. This is where I'll give you my take as someone who works with sound every day: the female voices have gotten scary good, but the male voices still sound generic to me — the same character over and over, and little artifacts that give the game away if you're listening for them. That gap is closing fast, though. Be honest with yourself: in a few years, most of us won't be able to hear the difference at all. Which raises the real question: if we can't hear it, does it still matter? Hold that thought.

One more thing worth knowing: Suno and Udio have never said what they trained on. That silence is the whole legal fight, and we'll get to it.

What it's actually changing

The flood is real. More than 100,000 new tracks land on streaming services every day. On Deezer, which actually tags AI music, fully AI-generated tracks hit about 75,000 a day in early 2026 — 44% of everything uploaded. Nearly half of all new "music" arriving there is made by machines.

I'll give you the fair nuance, though: those AI tracks are still only around 1 to 3% of what people actually play. Uploaded in the millions, listened to barely at all. So the panic that AI is already draining money from human artists is smaller than the headlines suggest — for now.

The money side is where it gets ugly. One man in North Carolina pleaded guilty in 2026 to pocketing more than $8 million by generating hundreds of thousands of AI songs and using bots to stream them billions of times. That's money taken straight out of the pot that pays real musicians. And the lawyers are busy: the major labels sued Suno and Udio in 2024 for training on their recordings without asking — then, a year later, several of them quietly settled and signed licensing deals instead. Train first, settle later.

And then there's the voice-stealing. In 2023 a fake "Drake and The Weeknd" track fooled millions before it got pulled. In 2025, an AI "band" called The Velvet Sundown passed a million monthly Spotify listeners before anyone realised nobody had ever been in the room. Here's my problem with that story: Deezer flagged it as AI, but Spotify didn't. The big platforms are not drawing a clear line. They're strangely shy about simply labelling what's AI and what isn't — and I think that's a choice that protects them, not us.

Where I actually land: the soul is the thing that's missing

Now the part I really care about. Strip away the numbers and here's what bothers me about AI music: there's no story behind it. Nobody sat up at two in the morning and wrote that song because something inside them had to come out. It wasn't shaped by grief, or a father who died too young, or a bad year that finally made sense in three chords. It's a calculation — a prompt matched against a library of everything that already exists. It can copy the shape of feeling. It can't have the feeling.

And when the soul goes, the value of the work quietly goes with it. If a song is free and instant and endless, we stop treating any song as something a person bled for.

I think that's why I'll make a prediction I'd bet on: AI music will get heard more and more, mostly without us even noticing. In the background, in the ad, in the playlist we don't choose. But I don't believe that in thirty years we'll be loving an AI song the way we still love Michael Jackson or the Rolling Stones. It doesn't have the depth that actually moves you. It's elevator music for unconscious listening.

The chance — and why I'm not afraid

Here's the twist you might not expect from a guy who builds his own guitars: I think this is a huge opportunity for real artists, and I'm not being naive about it.

But here's the thing: I don't think that chance is about the AI tools themselves. I'm not the guy reaching for AI mastering or a one-click "fix my mix" button. A song doesn't come alive from a random, data-based shortcut. It comes alive from people. Musicians in a room, talking, pushing each other, inspiring each other, chasing the same feeling until it finally clicks. That human exchange is the part I'll never trade away, and it's exactly the part a prompt can't give you. Yes, a few artists use AI carefully, with consent and credit. Grimes even offered her own voice for a 50% royalty split, and fair enough. But that's not my road. The more the world fills up with machine-made solutions, the more I'm convinced the answer runs the other way: humans creating with humans.

But the biggest opportunity is the strangest one. Being imperfect is now a superpower. When the whole internet is drowning in flawless, frictionless, machine-made audio, the human cracks are what cut through. The take that's a hair behind the beat. The voice that breaks on the honest line. My accent — which, let's be honest, no AI is ever going to bother copying :-). Those aren't flaws to fix anymore. They're the fingerprint. They're how you get recognised in the noise.

Where I stand

So here's my line, and I'll keep it simple:

Every note you hear from me is written, played and sung by a human. Composition, performance and vocals are 100% me. I use the normal studio tools every musician has used for decades. What I won't hand to a machine is the soul of the song.

That's the whole reason I build the guitar before I write the song. Not because handmade is automatically better, but because when every stream I earn goes to kids who need it, the human on both ends of that connection is the entire point.

The ground is moving fast and I could be wrong about parts of this. So I'll leave it as a real question, and I'd like your honest answer:

When you can't hear the difference anymore — why does it still matter to you that a human made it?

Tell me in the comments. And if you want to hear what one stubborn human still does by hand, my music lives here.