This yr’s Spotify Wrapped got here with an odd shock. The app’s annual roundup featured extra than simply the standard candy-colored knowledge about which artists, songs, and podcasts customers streamed probably the most. It included a “Wrapped AI podcast” characteristic, utilizing Google’s NotebookLM know-how to verbally recap customers’ listening habits.
This weird addition is extra than simply an audio companion for the data offered in Wrapped. Two NPR-inflected, artificial-intelligence voices undergo your listening highlights in a conversational (albeit nonetheless robotic) method whereas elaborating in your “moods,” “vibes,” and pursuits all year long. It feels each like listening to a health care provider undergo your bloodwork outcomes and a psychic vaguely supposing info about your life.
For instance, my “hosts” instructed that I used to be embracing a “bolder” vibe in the course of the month of September based mostly on the film soundtracks I used to be consuming. “The Challengers soundtrack and Ludwig Göransson’s work on Oppenheimer makes me wonder if you were watching some intense movies during this time,” one among my “hosts” speculated. This wasn’t completely unfaithful. But I additionally wasn’t certain why I wanted to be instructed this details about myself.
The function of this new characteristic isn’t precisely clear. The podcast component feels even much less impressed than a number of the different unusual personalization options the music streamer has launched up to now, from an AI DJ who greets customers by their identify, to “sound towns” the listener supposedly lives in, to this yr’s incoherent labels representing completely different moods (like no matter “Pink Pilates Princess Vogue Pop” is). These options are offered as makes an attempt to attach with customers on an intimate stage, however they’ve largely confirmed to be nice advertising and marketing methods in all their absurdity.
“The goal of Wrapped, in particular, is brand virality,” says Glenn McDonald, former knowledge alchemist at Spotify. “There’s not much in the way this year of data storytelling.”
Overall, the Spotify listening expertise has by no means felt extra alien.
Spotify’s personalization options are getting bizarre
The app, which launched in 2008, has lengthy been appreciated — and bemoaned — for its user-specific, algorithmically generated playlists and myriad compilations designed for any temper, activity, and time of the day. However, its newer personalization options, with an emphasis on AI know-how, have gotten a little bit out of hand. While its know-how is designed to grasp customers’ tastes, it’s precisely that understanding that retains customers in a repetitive consolation zone of songs, artists, and genres ceaselessly.
In 2023, the app rolled out an AI-powered DJ service referred to as “DJ X,” at the moment obtainable as “DJ: Wrapped.” Using an assortment of customers’ most performed songs and forgotten favorites, the characteristic imitates the FM radio expertise — solely with out the essential component of discovery and a human information. The AI DJ — who makes use of the replicated voice of Spotify’s Head of Cultural Partnerships, Xavier “X” Jernigan — gives a imprecise preamble in between songs: both primary info in regards to the artists or what temper they suppose the following music will evoke. It feels extra like an try at a guided leisure ritual than an adventurous listening expertise. (Of course, customers even have the choice to skip songs.)
Another AI-powered software Spotify launched earlier this yr is known as “AI Playlist.” Users can both enter an outline of the kind of playlist they wish to hearken to or choose a immediate, and generative AI will whip it proper up. While you may request any type of combine you need, the prompts appears to encourage listening to music the person is already acquainted with. For instance, the web page at the moment suggests customers strive asking the software to “put [their] top Wrapped songs in order of release date” or “give [them] a playlist of the genre [they] listened to the most.”
Some of those options have gotten much more gimmicky. As a consequence, they’ve grow to be efficient advertising and marketing instruments. Back in February, the app launched the trippy, cosmic-themed software “Song Psychic,” which mimics a Magic 8 ball. Users choose pre-written questions on their future and are given a (acquainted) music in return. More broadly recognized is the characteristic “Daylist,” which rapidly unfold throughout social media when it arrived earlier this yr. The “hyper-personalized” playlist modifications all through the day to supposedly mirror the person’s temper, based mostly on their listening habits throughout sure instances of the day. It’s largely garnered consideration for its AI-generated playlist titles that seem like phrase salad (“rage people pleaser wednesday evening,” for instance).
It’s unclear whether or not the “AI podcast” can be a mainstay of Spotify Wrapped or if Google NotebookLM can be built-in additional into the app. The “research assistant” permits customers to obtain summaries and extra data of paperwork utilizing Google’s Gemini chatbot. However, its perform, so far as the podcast goes, appears a bit redundant for the person. It’s primarily proof of how nicely Google NotebookLM can carry out quite than one thing that enhances person expertise in a considerable manner.
Are all these AI instruments simply making us self-absorbed?
These options don’t simply show how nicely Spotify is aware of you, however how predictable your conduct on the app is. These innovations could be enjoyable to mess around with. Sometimes, they’re precisely what customers want for the time being. But to what finish do streaming providers must hold reinforcing and catering to our personal moods and preferences?
McDonald, who now runs the music listing Every Noise At Once, says it’s a traditional inclination for customers to wish to hearken to music they already like. “There certainly are retention metrics that say that, unsurprisingly, people like listening to the things that they like,” he says. “But unless you really try actual adventurous experiments on what people might react to, you really don’t know.”
During his decade of working at Spotify, McDonald says the streaming service has by no means centered on what “it takes to get people curious.”
While Spotify has a well-liked Discover Weekly playlist that options new songs and artists based mostly round customers’ tastes, there are much less simply discoverable options designed for exploration on the app’s interface. Compare that to playlists the app gives prominently, like Your Time Capsule, On Repeat, and Repeat Rewind, all bringing customers again to their listening consolation zone.
All of those efforts to know listeners ignore music as an interactive artwork type. These Spotify options recommend that music is totally about your personal pleasure and ease and never one thing that’s meant to be engaged with, sought out, and even criticized. The user-focused promotion of AI doubles down on these habits. Listeners are alleged to be delighted {that a} robotic is aware of each little factor about them, whereas by no means difficult them to develop or change. All in all, these AI-powered options promote a narcissistic and, frankly, uninteresting strategy to artwork.
“The argument for streaming services is that all the world’s music is now accessible to you,” says McDonald. “But if all you do is listen to what you already know, it’s like giving someone a teleporter and they only use it to teleport back home.”