Spotify AI DJ review: an underwhelming novelty that falls short on discovery - for now
An AI voice mispronouncing up-and-coming acts won't change discovery habits
and live on Freeview channel 276
If Spotify's new AI DJ feature (recently rolled out to all Premium users in the UK and Ireland) learns more about you the more you use it, mine presumably thinks I'm an indecisive, musically fickle person with the attention span of a fly. That's because I spent most of my time with it hopping around genres, trying to predict what it would throw at me next.
The gimmick here is surely the DJ's synthesised voice - after all, Spotify's on-the-fly curation systems have to this point been fairly robust. The 'algorithm' already churns out five or six Daily Mixes a day, slicing your tastes into vaguely cohesive chunks while throwing in some genuinely impressive new recommendations.
Even spontaneously asking for a playlist based around a certain artist brings up a seemingly tailored list of songs - I'd hazard my 'Strokes Radio' might look a little different from yours.
So hearing it all delivered through a voice that has been described - by Spotify - as "stunningly realistic" (and modelled after Spotify’s Head of Cultural Partnerships, Xavier “X” Jernigan no less) is surely the big selling point?
Which brings me back to my constant genre hopping. The feature gives users the ability to move on to a new genre or category at the touch of the button if they aren't "feeling the vibe" - the AI already shifting the blame onto you, rather than the computer's rubbish picks.
Each new hop brings with it a personalised intro, and so I was more interested in seeing how the feature's digitised voice would handle some of the more esoteric artists in my rotation. Indeed, it quickly tripped on Brighton electronica/noise trio CLT DRP (it's pronounced "Clit Drip") with the rather long-winded, "cee-ell-tee-dee-arr-pee".
Apart from that, the main fun seemed to be in seeing what exactly the DJ listed under its sometimes obvious genre suggestions.
Presumably nobody in the world would be surprised if they were occasionally served up a genre as ubiquitous as indie, but even more so for me, throwing out Pavement as a first suggestion is like blurting out the Generation Game's cuddly toy and then struggling on the more nuanced choices.
Anyone who's spent 10 seconds with my Spotify account knows Malkmus is God. Its proud suggestion of "noise-pop" was perhaps slightly more leftfield. But to follow that up with Dinosaur Jr? Built to Spill and The Walkmen? Debatable.
One area where the AI DJ does seem to have an edge over those Daily Mixes is in its apparent ability to reach for the deep cuts. When “X” promised some tracks from Idlewild, I got ready to hear the obvious single picks that I've heard on 'shuffle' time and again.
But no, what I instead got was a pleasantly surprising smattering of early career B-sides, a time at which the band were - in this writer's humble opinion - at their best. Though whether this was the AI's genuine attempt to surprise me with forgotten gems or a quirk in its early-days programming remains to be seen.
At this stage, I find it hard to imagine using the AI DJ for anything more than the novelty, or perhaps as a more-random-than-random starting point from which I'll take back full control. But if it really does get better with more use, perhaps in the coming months it just might evolve into an invaluable tool.
According to Spotify's data, on "days when users tune in", they spend 25% of their listening time with the DJ, and more than half of first-time listeners come back to listen to DJ the next day. At the time of writing, I have not yet gone back to it on Day Two, though the platform insists it has "especially resonated with Gen Z and Millennials", who make up 87% of DJ users.
Whether I'll be one of those Millennials with which it will one day especially resonate remains to be seen, but for now, Spotify's already strong recommendation tools are the way I'll be enjoying music. An AI voice mispronouncing up-and-coming acts isn't going to change that.
NationalWorld Editor Nick Mitchell seems to agree, saying his first impressions of Spotify's new AI DJ were "slightly underwhelming," with the platform "already crammed full of personalised playlists" and its algorithm being "undeniably impressive at serving up what it thinks listeners want to hear."
"What the new 'AI DJ' seems to do," he says, "is cherry pick from these playlists across genres, and interject every few songs with an admittedly realistic sounding fake human voice. It served me some Daft Punk and Yeah Yeah Yeahs, which was all to the good, but it didn't seem particularly ground-breaking, or great at finding new music for me.
"My suggestion is that they give people a choice of accents or celebrity voices in their next version. Jeff Goldblum introducing the new Kendrick Lamar track would be a novel experience." Now that's a thought.