FYI.

This story is over 5 years old.

Tech

There Are 800 Musical Genres and Spotify Wants to Find Your Favorite

Spotify is betting that artificial intelligence can do what record store snobs never could.
Image: Brad Cohen/Flickr

The "music intelligence" startup Echo Nest was just bought by Spotify for undisclosed sum, proving that, in the streaming music business, it’s not just about giving people what they want, but also what they’ll want next. Echo Nest's machine learning technology, which already powers Spotify's radio feature, and will now get to work improving its relatively new Discovery section, by predicting what you'll like and suggesting new music based on its analysis of your listening habits.

Echo Nest’s massive data collections—35 million songs, broken down into 1,168,162,257,347 datapoints—have also been powering personal recommendation for Pandora, Rdio, Twitter, VEVO, and its API will remain up and free for developers to keep building third-party music apps.

Advertisement

Even if you’ve never heard of the Echo Nest until now and you don’t use any of those streaming services, you likely still know its work. You know those maps of America and its musical taste that have been bombarding our Facebook feeds for a few weeks now? The man behind Your State’s Most Distinctive Musical Artist—which among other things revealed New England’s love-affair with the guitar solo —as well as Your State’s Favorite Artist—which revealed that states only like Jay-Z or Macklemore—is named Paul Lamere. He’s Echo Nest’s director of developer platform.

Following the trend of geography and musical artist to the municipal level, yesterday Echo Nest revealed an index of which genres cities are most strongly associated with. Perusing this list, it appears Philly’s full of ‘Chicago Soul,’ Chicago’ is full of ‘Louisiana Blues,’ and Los Angeles has got ‘Jazz Christmas,’ ‘Punk Christmas’ and ‘Jerk.’ But then, what’s in a name?

The list is fodder for more arguments, I guess, but it also drives home just how tired and imperfect musical genres have always been, and how, if we’re going to have access to millions of hours of music, how we categorize it is going to change too. Genre reigned as a way of helping people find music—you want some rock music, it's all over there, they'd say. Now big data and artifical intelligence are taking that mantle.

"As helpful as [genres] were for categorizing three rows of vinyl in a record store—they are of little use to today’s music fans, as they stare into an abyss of 30 million songs from all over the world,” Echo Nest wrote on its blog. Asking for ‘Rock’ is only slightly better than 'Play me some songs that are music.’”

Advertisement

"Genre” has never been a very tidy category—some genres divide up by a shared history, some by aesthetic, some by beats per minute. What is it that people who dislike country music like about Johnny Cash? Would they prefer rockabilly? Would they prefer spoken-word James Earl Jones? Free from the mass media of radio and television, the only category that really matters is "do you like it?"

Echo Nest still relies on genres, but unlike a record store that just had Rock & Pop, Jazz, and Classical, it has 800 genres, doesn’t limit artists to only one, and categorizes artists still further within the genres—are they a core member, an up and comer, or an outlier?

Nuance is a good thing, and allowing shades of gray and mutability keeps Television-punk-fans from getting stuck listening to Blink 182, or even having to admit that they exist. Jazz on the other hand covers an entire century of music. It’s conceivable that someone who wants to hear Jelly Roll Morton would also want to hear Sun Ra, but that’s not exactly why people are signing up for customizable services.

It’ll be interesting to see if Echo Nest’s artificial intelligence can balance out giving people what they want while still stretching what they hear, given that “unity” is something just a function of the listener’s intelligence. Early in my music streaming life, the worst Pandora stations just came off as insulting and seemed vastly inferior to human-curated stations and playlists that defied genre—the online-only station Luxuria FM plays “Surf Music, Bossa Nova, Exotica, Space-Age Bachelor Pad, Jazz, Soft-Psych, Sunshine Pop, Wall of Sound, Latin, Go-Go, and Film & TV music,” the unifying principle seems to be a certain evocative sensibility—but couldn’t that also have been a lack of charity towards a computer, and faith in the curator?

Spotify, the biggest and most successful music streaming service globally, is betting that AI can sort what record store snobs never could, not by striving for objectivity, but by allowing for subjectivity, which for something as personal as someone’s musical choices is probably best. If you’re going to be stereotyped by your musical taste—and Echo Nest has a mechanism already built specifically for that—then you want to make sure it’s getting your taste right.