One of the bloggers I recently discovered is Alex Hsu, who shared how he’s been making music with Suno. The generated music from Suno V5 was so good he could listen on repeat for hours. He called it like having a private chef: you tell it what flavour you’re in the mood for, and it makes it on the spot.
That intrigued me. The last time I played with AI-generated music was already two years ago, and I remembered not liking it at all.
So I was genuinely surprised, pleasantly so, when I discovered how good the music coming out of Suno sounds nowadays. Here’s one example. The chords are tasteful. The production sounds professional. I was so excited that I almost feverishly tried to explain to Mitch how much AI-generated music has improved. He scoffed. And that scoff got me thinking.
The Resistance for AI Music #
It feels like people resist AI-generated music more fiercely than AI-generated text or images. Yes, there’s plenty of backlash for both too, but music especially so.
Perhaps it’s because when we listen to music, we’re engaging with signals that bridge human emotions, transcending language and borders. When you discover a song you like, you’re naturally curious about the artist behind it. And if you end up liking a variety of their music, you start to feel deeply connected to the person or the band. You collect the stories and anecdotes about them, where they’re touring, how the live version is different from the studio recording but equally good.
That’s how we end up discovering our favourite musicians and can’t stop talking about them and sharing them with friends. I don’t think AI has any chance of replacing this type of music listening. When you hear a piece you love and then find out it’s AI-generated — you’d feel surprised and pleased at best, or disappointed, even deceived. But either way, you’re unlikely to chase the backstory.
However, there’s also a type of music — background music. When you’re studying, coding, or writing, you like to have a certain kind of music in the background to help induce a certain mood: calmness, focus, etc. For this type of listening experience, maybe there’s a place for AI-generated music. At the most basic level, we’re just a bunch of atoms reacting to certain sound wave frequencies. That may sound cold and reductionist, but over-glorifying a human-centric view has caused trouble in the past, and it’ll likely do so in the future.
Of course there’s also the legitimate fairness concern. Many musicians struggle for years with little money despite much devotion.
Mitch believes art is inherently, inseparably tied to being human. If we start to accept AI-generated art — first as a convenience, then as a norm — we eventually arrive at a world where AI art has fully replaced human art. And at that point, art itself loses its meaning.
And even before we reach that endpoint, the process of getting there causes real harm. Artists lose livelihoods. The culture that sustains creative communities erodes. The other day when Mitch was playing Zelda, I noticed that the NPCs just say “hm, ah~” instead of reading their dialogue out loud. I said, “Oh, this seems like the perfect place for voice AI to improve the experience!” He said: that’s exactly the slippery slope. Once you start using AI for NPC voices, soon it’ll be the main characters, soon the voice actors lose their jobs.
Perhaps the chain reaction is what people are worried about. Each step feels small and justifiable. But they add up to something irreversible.
AI Music: Beautiful but Forgettable #
I think I did discover one big weakness of AI music.
Two days after that conversation with Mitch, I was trying to recall the music generated by Suno V5 and realised something interesting: I couldn’t recall what any of it sounded like. I couldn’t hum a single melody — even though I’d listened 5+ times to each piece.
That’s so different from human-made music. When you hear something you really love — even just once or twice — the melody gets stuck in your head. Even if you want to get it out, you can’t.
So I went back and listened again. And it still sounded beautiful. The melodies flowed one to the next seamlessly. But after hearing a track a number of times, I still couldn’t hum it back to myself if I stopped for a while and tried to recall.
Perhaps the biggest issue is this.
Within a single minute of a Suno track, there’s a lot of change. Every phrase flows into the next. It all sounds connected, all sounds beautiful — but it feels like having a conversation with someone speaking eloquently, even floridly. The vocabulary is artistically chosen, pleasing to the ear. But after the conversation, I can’t remember what we talked about. I don’t remember a single memorable story from it.
The Double-Blind Question #
Another angle: if we ran a double-blind experiment, and a listener had no idea whether they were hearing a human or a machine, and they felt moved — why should it matter who made it? The tears still happened. The meaning still landed. If you were moved, you were moved.
We both agree on that part. If you don’t know, there’s nothing to protest about. But once you find out, the whole experience might change. Mitch gave an example: imagine you’re eating a delicious burger, and then the chef tells you it’s actually made of a baby’s head. Now how do you feel about the burger?
People often love a song because they feel bonded to the person or band behind it. The knowing of the artist changes how you experience the art:
- If the artist is a wonderful human being — the song feels even more brilliant.
- If the singer or composer turns out to be awful — the melody stains.
- And if you discover the other end is not a human at all, but a computation optimised to please — I think a lot of people feel something between rage and emptiness. A sense of being tricked.
Because music experience is not a double-blind experiment. We’ve always had the right — and even been encouraged — to inquire about the artists behind the work. That’s how legendary music gets passed along generations, travels vast distances, and becomes timeless.
The Threshold of Art #
There’s also a question about the threshold of “art” that I find interesting. Automation in artistic choices has always evolved over generations. Think of a film editor manually adjusting colour curves frame by frame versus describing exactly what you want and clicking one button. Does the one-button version feel like less of an art? On one end of the spectrum, the artist clicks a million buttons to do everything themselves. On the other end, the artist clicks one button and everything is done. At what point do we stop calling it art because not enough buttons were pushed?
Though this post may read like I’m a big proponent of AI-generated music, that’s not true. So far, human-made music has a magic power to transport me across time and space — an afternoon with my best friend, the memory and feelings of words that we couldn’t say. Music feels like the soundtrack highlighting the journey and landmarks of my life.
But I do wonder if some of the resistance is less about what AI music is and more about the speed at which it arrived. We haven’t had time to prepare. And when we’re not prepared, we rationalise our irrational feelings. Well, I guess that’s also inherent to being human.
If you’re curious, here’s another Suno-generated track — see if you can hum it back after a few listens.