What My Toddler Taught Me About Music and AI

When she was three years old, I watched my daughter operate a drinking fountain for the first time. As she leaned in, she loosened her grip on the button, cutting the flow. She tried this several times, with admirable persistence but zero success. Eventually she had an idea: Lean in first, then press the button. That worked, kind of: She got a noseful of water (footnote 1). On the next attempt, the water squirted into her mouth. The delighted look on her face suggested—correctly—that she'd solved something important.

That makes her at least as smart as a sulphur-crested cockatoo. Last year, news media reported that in Sydney, Australia, a crackle of the birds learned to operate a drinking fountain by perching on its spring-loaded handle, one after the other. There was a creek nearby that they could have been drinking from, so the birds appeared to enjoy the challenge. "One of them got it, and then the others were like, 'Ah, this is fun'," a local ecologist told The Guardian. The flock, he surmised, was "having a bit of a relax and a bit of a laugh."

The birds, and my daughter, know something that Mikey Shulman doesn't. Shulman is the CEO and cofounder of Suno, a billion-dollar AI music-generation company that lets users create complete songs via text prompts. Type "bluesy '70s rock ballad with horns," and Suno does the rest, quite credibly and in just a minute or two.

Shulman sees himself as a liberator—the anticockatoo, if you will. "It's not really enjoyable to make music now," he stated in a recent interview. He meant, with people: musicians, vocalists, conductors, arrangers, engineers. "It takes a lot of time, it takes a lot of practice," he explained. "You need to get really good at an instrument, or really good at a piece of production software. I think the majority of people don't enjoy the majority of time they spend making music."

You almost have to admire the willful lack of understanding, indistinguishable from chutzpah. Music-making, Shulman says, is a burden. The effort is the problem. The cure is his software.

Jazz pianist Brad Mehldau responded on Substack: If making music were unpleasant, musicians would stop. The fact that they persist through years of grinding practice, boredom, humiliation, obsession, small breakthroughs, and occasional despair is not evidence that the process is miserable. It's evidence that the process is the point. "We love to strive," Mehldau wrote. "It's how we grow."

The joy of craft
He's right, but the gain goes deeper. There's profound satisfaction in getting good at something difficult. A paycheck helps prevent dissatisfaction (and destitution), yes ... but mastery is its own private currency. It doesn't matter whether you're a woodworker, a composer, or a comedian; we revel in our command over a piece of balsam, in the quiet beauty of a contrapuntal melody we crafted, or in the precision of a joke that lands just right. We made something that met our own standards before it met anyone else's.

And making music takes this into a higher realm. Players instinctively adjust to one another. The drummer hangs back a fraction of a beat, while the guitarist bends toward a change and the rest of the band picks it up, wordlessly. A groove forms. There's eye contact. There are smiles. Suddenly there's a shared current in the room that feels like magic or telepathy. When I played and sang with others, it was a source of joy to achieve this "mind-melding" mental state. If you were lucky enough to get there, it could feel as if you were mining something deep and sacred. There was affection in it, and a slight, pleasurable loss of self. That wordless reciprocity and keen attunement can be as good as making love. Where's that button when you use Suno?

Speaking of buttons: I don't dislike them as a means of music-making. At all. I have no problem with drum machines, sequencers, samplers, digital effects. They gave us Yello, Kendrick Lamar, Billie Eilish, Daft Punk, Infected Mushroom. Guitarist Jack White once said, "Technology is a big destroyer of emotion and truth. ... That's the disease we have to fight in any creative field: ease of use." White has been known to record on an eight-track tape machine because the constraint forces decisions. You can't tinker without end. When everything is revisable, maybe nothing requires your full presence and attention.

AI-generated music is ease of use taken to its logical end. There are no personal stakes. There's no moment that demands, this is the take. What disappears is effort and the sense that someone risked failure.

If you can't tell, does it matter?
The obvious defense of AI-generated music is also the oldest defense of convincing imitation: If you can't tell, does it matter?

I found out recently that it might. On a whim, I'd asked Suno to generate a funk track with a marching band rhythm. It produced a piece of swampy Americana with gravelly male vocals. It was sophisticated and sounded great. My foot tapped. And then something soured, because I knew what this music wasn't. It wasn't a band's Wednesday night in a rehearsal space, or a vision forged through years of almost no pay and stubborn ambition. It sounded like a band, a good one, but it didn't feel like one. In the end, it warmed me about as much as a screensaver of a fireplace.

And I thought of Mikey Shulman, who would have handed my daughter a bottle of water, thinking he'd solved the problem.


Footnote 1: Caution, unbearable cuteness ahead: tinyurl.com/4me6jdda .

Advertisement
Advertisement
Advertisement