Why AI Is the Future of Music
The music industry finally has some reason to celebrate, thanks to artificial intelligence.
The Recording Industry Association of America (RIAA) recently announced that music revenues in 2016 grew 11.4 percent to $7.7 billion — the highest year-over-year growth rate since 1998. Although the industry is only half the size it was in 1999, double-digit growth is encouraging after years of either declines or flat results. Why the growth? According to the RIAA, the answer is simple: streaming is taking hold. And streaming services — especially Spotify — are lapping the field with AI.
As the RIAA noted, the biggest contributor to growth was a doubling of revenues from paid streaming services such as Apple Music, Spotify, and Pandora. In fact, for the first time ever, streaming music platforms generated the majority of the U.S. music industry’s revenues.
Younger streaming platforms such as Tidal are still too new to contribute significantly to the $3.9 billion that streaming services generated in 2016. Rather, the established streaming leaders, especially Spotify, are hitting their strides by offering better products fueled by AI.
Pandora and the Power of Personalization
Streaming services such as Pandora and Spotify have always created customers by personalizing their vast inventories of music. If you stream music, you already know how well Pandora and Spotify create engagement by offering you customized listening choices based on your personal tastes. I still remember how exciting it was when I first started using Pandora years ago and created my own Pandora radio stations based on names of artists or songs that appealed to me. If I wanted to create a station based on my love of Massive Attack, I could do so. If I wanted to create a station of music inspired by the Cure song “All Cats Are Grey,” I could do so. And Pandora refined my stations even further when I gave a thumbs up or thumbs down to songs that Pandora suggested to me based on my listening tastes.
These customizable recommendations didn’t happen by magic. Behind the scenes, Pandora was using, and still does use, AI to analyze hundreds of attributes of songs in order to find the most interesting matches based on my tastes. AI makes it possible for Pandora to go beyond obvious song suggestions and uncover choices that might not be on your radar screen. That way, Pandora won’t simply recommend songs you already know from an artist or genre. For instance, if you create a Johnny Cash radio station, Pandora will analyze the attributes of Johnny Cash songs and other artists to recommend, say, Lynyrd Skynyrd’s “Sweet Home Alabama” because the song contains similar rock and country influences and mild rhythmic syncopation, even though technically Johnny Cash and Lynyrd Skynyrd belong to different music genres.
The AI engine, known as the Music Genome Project, was co-launched in 1999 by Pandora CEO Tim Westergren and remains the core of Pandora’s recommendation engine. And the more feedback you give to Pandora — such as skipping songs you don’t want to hear or giving a thumbs-up to song you like — the smarter Pandora becomes thanks to machine learning, a form of AI in which technology teaches itself to get smarter.
The Music Genome Project gave Pandora a head start on all the other streaming services. Consequently Pandora is the most popular audio streaming service in America as measured by how many Americans listen to the service each month. According to Edison Research, 32 percent of Americans listened to Pandora in the past month, compared to 18 percent for the next-most popular choice, Spotify.
Spotify Flexes Its AI Muscle
But many of those Pandora listeners consume Pandora’s content for free. In the category where it really counts — paying subscribers — Spotify has an advantage over Pandora. Spotify boasts 50 million monthly subscribers, more than twice the number of its closest rival, Apple Music, which has 20 million subscribers. And AI has been key to Spotify’s ascendance. For instance, in 2015, Spotify launched a feature called Discover Weekly that personalizes playlists based on your tastes – and also the listening habits of other Spotify users.
Discover Weekly uses AI to cull through the approximately 2 billion user playlists to find people who are listening to music similar to your playlist. Then Spotify finds songs to recommend to you based on your own listening tastes and on the music that other users with similar tastes listen to.
For instance, Spotify might notice that you’ve been giving Goldfrapp’s “Anymore” and Bjork’s “Lionsong” heavy rotation on your playlist. Spotify will find other users who happen to have both songs on their playlists, look for more songs that their playlists have in common, and then come up with a recommendation that you hadn’t heard before, such as, say, Thievery Corporation’s “Letter to the Editor,” or Röyksopp’s “Bounty Hunters.” And the algorithm teaches itself to make smarter song recommendations the more you use Spotify and the more the algorithm learns about your listening activity.
As writer Adam Pasick explained in Quartz, Spotify’s algorithm also takes into account a complicated breakdown of your own listening profile into refined categories such as neo soul and new Americana to figure out matches from emerging artists who might not be on your radar screen. By creating a highly detailed and complex profile based on what you listen to, what you skip, and music you browse, overlaid with music people like you are listening to, Spotify can constantly surprise its users with interesting song recommendations.
According to Spotify’s Matthew Ogle, quoted in Pasick’s article, ”We now have more technology than ever before to ensure that if you’re the smallest, strangest musician in the world, doing something that only 20 people in the world will dig, we can now find those 20 people and connect the dots between the artist and listeners.”
Discover Weekly has proved to be so popular that Spotify has rolled out additional algorithm-based features such as Release Radar, which focuses on new releases each week customized for each user based on the artists they listen to the most.
Enter Amazon
But Spotify can ill afford to rest on its laurels now that Amazon has launched its own subscription service, Amazon Music Unlimited. Amazon offers the service through a tiered pricing strategy, with the lowest price point ($4.99) reserved for owners of Amazon Echo voice-activated devices and the next lowest price being reserved for Amazon Prime members.
The Amazon pricing strategy clearly rewards customers who are willing to buy into Amazon’s premium services. I find the Echo-friendly pricing especially significant. Amazon claims that Echo sales for the 2016 holiday season were 9 times higher than the year before, with Amazon selling 5.1 million Echo devices all told. Amazon is credited with pioneering the use of AI to create personalized product recommendations based on a customer’s purchase and browsing history, and the company has quickly seized leadership of voice assistant technology. It’s easy to see what Amazon is capable of developing: smart playlists that you call up with your voice (“Please play for me women singers who write songs like the ones I’m hearing on Father John Misty’s new album”). I predict that in 2017, Amazon will pressure Spotify and Pandora to up the ante in AI-based music. Already, Pandora has rolled out a premium service that offers new features such as the ability to create playlists based exclusively on songs you like, which is crucial if Pandora is going to ever challenge Spotify’s mighty paid subscriber base.
Instead of seeing streaming wars erupt, we’re going to see AI wars intensify, as the major streaming services attempt to make their catalogs even more personal.
Music Streaming Continues to Grow
Music streaming continues to buoy the industry in 2017. According to BuzzAngle Music, audio streams for the first quarter have increased 61 percent compared to the first quarter of 2016. Streaming services have many other cards to play in order to attract more users, such as offering exclusive album releases. But exclusive releases are a limited option. What’s really going to make the difference is providing a better product. Consumers are not going to subscribe to multiple services just to score every exclusive album deal. They’ll become loyal to the streaming service that provides the best product, which is where AI has already been making the crucial difference by making music more personal. With streaming services setting the pace for the growth of the music industry, AI is the future of the music industry.
Media Professional/Music Executive
7yThis is a great article. Thank you for posting this, very informative.
Consultant, Retired Director (CTO, CIO)
7yHow interesting that this prediction may probably be wrong. Analog vinyl sales exceeded the sales of digital music purchases by 2016-2017, proving that record shopping is reinvented and has got mainstream. Music is arising in the grooves.
Problem solving is my thing, let's connect
7yIt sure is. The faster the better
Senior Register Jobcoach bij Van der Horst Jobcoaching & Training
7ymusic in general eas the brain and makes space for other experiences! And the learning of new information is much easier if you listen to music or if you are dance if you like to. Music en moving are very good for the brain and health in general.