I’m listening to Radiohead’s Creep on the radio. “You may not know this,” the DJ coos patronizingly, “but this song turns 30 this year.” So far, so titbit of trivia on FM drivetime. The only difference is this DJ is not a real person.
AI DJ is the next move in Spotify’s never-ending goal to “personalize” our listening experiences. Like its Discover Weekly new music playlist, or its end-of-year Wrapped recap, the AI DJ curates a stream of songs it thinks I’ll like based on my listening history. Along with the tunes, I get segues of “commentary” from a male AI voice, which bursts with the forced friendliness of an over-invested high school guidance counselor.
The feature is currently being rolled out, so you may have already noticed it on your Spotify app. Like just about everything this year, it’s made with OpenAI, a chatbot known for its distinctively human – and sometimes quite creepy – responses. Spotify music editors also give the feature a boost, as they’re the ones who write up the fun facts the voice gives you every few songs. “With this generative AI tooling, our editors are able to scale their innate knowledge in ways never before possible,” said a press release.
“Let’s keep this vibe going,” the voice says as I skip through the next songs: Television’s Marquee Moon, The Slits’ Typical Girls, and Jimmy Eat World’s The Middle. I learn that robots like dad rock.
While the honey-voiced guide to my listening session is not a real person, it’s based on one: Xavier “X” Jernigan. He’s the head of Spotify’s cultural partnerships and host of The Get Up podcast, a daily morning show about pop culture that also gives listeners a “personalized playlist”.
According to a Spotify press release, “[Jernigan’s] personality and voice resonated with our listeners” well enough for him to land the gig as the “first model for the DJ”. But more options could come in the future, the brand noted.
I liked the voice, even if it came off as stalker-ish. “I know what you listen to, so I’m going to be here every day,” it told me, kind of threateningly. And it is a little scary just how much the AI DJ knows about me.
With the tap of a button, it launches into a “look back at 2018”, but not the songs that came out that year – the songs I listened to. It starts blasting three Celine Dion anthems I played nonstop during a particularly self-indulgent time in my life. As I get a weather alert about an incoming snowstorm, it switches to my “favorite summer jams” from past years.
Was this “personalization”? I felt like the AI DJ was less picking out songs from its digital crystal ball of predictions and mostly just playing songs I’d listened to before. It was also happy to serve up songs supposedly “picked just for me” by their editors: indie and R&B playlists that seemed to hype up new releases.
It’s tough to imagine who, exactly, this feature is for. The app’s promotional videos brand it as a kind of personalized crate-digger serving up deep cuts and commentary. But I assume that those looking for human curation will keep listening to, well, actual radio (or maybe rival Apple Music’s take on “curated” services, which taps experts like Ebro Darden, Zane Lowe, and Elton John as alternatives to algorithms.)
For the most part, the AI DJ wants to please me. When I tap a button to change the genre, I get an amiable: “Not into it? I got you.” There’s no waiting around for a song I like to enter the rotation, as I do with the FM station in my car. It’s effortless. But after some time, the AI DJ starts to speak to me less and less. Soon, it only lets me know what artist is on deck. I wonder why I need it at all.
After 30 minutes, I realize I’m just listening to my top-played songs. I’ve heard them all thousands of time. I turn off the AI DJ, and go look for a new album I haven’t heard yet.