The auto-generated "favorites" that YouTube Music pitches me can get a little stale, as one overfamiliar song after another is served up. Sure, they're songs I like, but algorithmic recommendations start feeling a little bit like the musical equivalent of Bread and Jam for Frances. "Sweet Lord Jesus, this again?" I'll mutter, skipping to the next one.
Usually, new additions are welcome. Only now, I find myself having to be warier and warier around new music, particularly if it feels a certain way. More and more new music isn't produced by human beings at all, but is instead the result of a short prompt dumped into Suno or some other AI. I created a few songs myself, back when it was unusual and felt like a fun thing to try. But I'd rather not listen to it.
As every ad for Suno reinforces, that creation process requires exactly zero talent, as people who haven't the skill or won't put in the work are encouraged to pretend they've made something. So when I hear a song that exactly matches a particular vibe, I'm on alert. Machine music takes advantage of our distraction, engaging parasitically with the musical styles we enjoy and crowding out human voices and talents.
Like, say, my listening preferences when driving on a cool summer evening. I drive with the windows down and the sunroof open at such times, letting the wind play through my hair. I want stuff that's utterly relaxing, both mellow and sophisticated. Give me a tree-lined country road paired with a sultry chanteuse purring over a tightly constructed jazz standard, and I'm as happy as a clam.
If clams liked driving briskly whilst listening to jazz, that is. It may or may not be a mollusk preference.
So the other day while driving home from church, I heard a new tune that landed perfectly. Bubbly and playful, with lyrics that danced between English and Japanese, it was just one hundred and ten percent chill vibe. "Little Sunny Moments," it was called, and I enjoyed it as it hummed away in the background. But as it progressed seamlessly from verse to chorus and came to a quick and easily digestible conclusion, I began to wonder.
Is this a real song?
It felt too tight, the vocalizations too sharply defined and breathless around the edges. Each "Doo Doo Doo" didn't quite have the shape and form of a human utterance. I'd not really paid attention to it, so I listened to it a second time, tending to the lyrics, trying to come to a determination as to its provenance.
[garbled] is shining in the light.
Everything just feels so right.
Tiny waves go up and down.
Nothing special, nothing big.
Little sunny day.
Doo doo doo. Drifting far away.
Doo doo doo. Come and stay.
Seagulls dancing in the blue.
I just sit and think of you.
Clouds are floating like a dream.
If I laugh a little more,
will this day just last forever?
Doo doo doo. Little sunny day.
Doo doo doo.
Doo doo doo. Drifting far away.
Doo doo doo. Come and stay.
The lyrics sort of made sense, but were more a feeling than anything coherent, and the garbling of some words and hyperprecision of others felt less organic and more generated. Odds of it being AI? Now at 90%.
And then came the final verse, sung to precisely the same tune as the rest of the song:
Smooth relaxing jazz Bossa Nova.
Similar mood to Quiet Glow.Soft breathy female.
Soft Doo Doo Doo
Soft Breath Doo Doo Doo
Soft breathy female.
Smooth.
Soft breathy female.
Smooth.
It. Was. The. Prompt. O Lord have mercy, how perfect. From behind the wheel, I cackled with delight.
Somehow, the prompt used by a human to tell an AI what kind of song to make had been accidentally folded into the lyrics, and...this being a low effort sort of thing...they'd just burped the thing out without even noticing.
"This," I thought to myself, "should be mandatory." Meaning, at the end of every AI song, the final lyric must be the prompt.
Just so we know. So it's clear how much effort was involved, and how little humanity.
Posts
