BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

Spotify Wants To Hook Users On AI Music Creation Tools

Following

One tool will let you tweak a song’s rhythm or melody. Another can take a pop song from, say, Justin Bieber or Drake and combine it with a fugue by Schubert or Bach, if that’s your thing.

“You’ll be able to try all sorts of combinations,” says Francois Pachet, director of AI research and development at Spotify. “And it’s really fun.”

It’ll be really fun for sure ... assuming copyright clearances from publishers, labels, artists and songwriters get sorted out regarding any pop music that gets modified or mashed-up with AI modifications. But the algorithms don’t need pre-existing recordings, because they can make original music too.

And will that computer-generated music have a copyright? That very question is being studied by the UK’s Intellectual Property Office, it announced this week.

Meanwhile Spotify is forging ahead with its AI Music agenda, although no date for the launch of its suite of AI tools has yet been publicly disclosed.

To highlight what they’re up to, Spotify’s AI team, based in Paris, has developed Skygge, a duo with songwriter and producer Benoit Carre paired with an AI program developed by Pachet called the Flow Machine.

Competing AI Music platforms are popping up like mushrooms, some of them psychedelic, like the one from Sony’s Paris-based computer science labs, formerly headed up by Pachet before he moved to Spotify in 2017. They used AI to create a Beatle-esque track called “Daddy’s Car.” Call it “Lucy In The Sky With Cubic Zirconia,” a trippy ride through a winding wonderland where instead of “newspaper taxis appearing on the shore” you get “take me to the diamond sky … take me on a distant sky,” suggesting the data-driven taxi driver may not speak English as a first language.

But Spotify’s project smartly moves AI music from a lab experiment to a recreational toy, a better bet than trying to best the Beatles with bytes. The idea is that Spotify users will engage more when they have a hand in creating the music with the help of AI.

“It’s like the instant cake mix story,” Pachet says. As the story goes, a food conglomerate in the 1950s boosted sales when they figured out that when you get the customer involved in the cooking, even a little, sales rise. They removed dried egg from an instant cake mix, so the customers – usually housewives back then – would have to add a fresh egg. “It was that bit of effort,” Pachet says, “so she could say she made the cake... that made all the difference.”

Just as someone who doesn’t know how to cook could make a perfect cake with the right mix, so too a Spotify user who doesn’t know where middle C is on a keyboard could make perfect music, with AI assistance.

So will AI Music take gigs away from human musicians? It’s the same fear facing most professions these days, the same dread echoed in Elon Musk’s warning that we’re headed toward the “Singularity,” a Matrix-like existence where AI takes control away from humans. But AI Music shows that computers will not replace humans any time soon, at least when it comes to music, or at least meaningful music.

Could a computer create something as life-changing as Leonard Cohen’s “Hallelujah?” That masterpiece was the result of around 80 drafts, some reportedly spawned in New York’s Royalton Hotel where the songwriter in his underwear said he banged his head on the floor to coax some of the song out of him. Cohen knew that human flaws were a key to great art. “It is by intimacy with the flaw that we discern our real humanity and our real connection with divine inspiration,” he said.

Human flaws may make music more lovable.

Take John Paul Jones’ bungled bass stumble at 1:50 of Led Zepellin’s “Good Times, Bad Times,” where he misses the change from verse to chorus but recovers at lightning speed. Or consider Pharrell Williams’ pitchy vocals on N.E.R.D’s “Run To The Sun,” where autotune would have fixed it, but also would have ruined the magic human error sometimes produces. Human slip ups are strangely captivating and lend authenticity.

So, when the Singularity comes, what singles will we want to hear, if we’re still here?

We’ll want songs to soothe, excite and change our lives, and that can’t come from a computer. Or can it? Some of today’s pop music already sounds like it comes from a data processing device. Take Katy Perry’s “Dark Horse,” the target of an ill-fated copyright infringement lawsuit. The disputed passage in the song sounds like it was composed by a metronome. Did it change any lives? Maybe.

And when the Singularity comes, will there be any more music copyright infringement lawsuits? Maybe not, but while we’re waiting for the Singularity, if it ever comes, AI music could be the subject of a copyright, trademark or name and likeness lawsuit if the music simulates the style of a human musician or copies a song. But who would be the defendant? Is it the computer, the composer of the software, or the company releasing the music? Likely, the defendant will have a central nervous system rather than a central processing unit.

But AI Music can provide many useful functions to us humans, including as a tool for complex film and game music or as a service to provide soothing music near a crying baby at a shopping mall, for example.

“We often use music to regulate our emotions in some way,” says Pamela Pavliscak, a professor of design at the Pratt Institute, who advises tech companies on emotional issues around computing. “You've had a bad breakup and at first you want to wallow in it. So you listen to sad songs and you get yourself in a mood. And after a while it’s like, OK, I’ve gotta pick myself up. So you switch from ballads to upbeat, happy music. AI can assist with this, but for it to truly manipulate people’s feelings with music? No, it’s not good enough to do that. Only a human can to that, and only a few at that.”

Another reason computer music won’t dominate the Billboard Charts and Spotify playlists anytime soon is that music’s popularity depends on more than the music itself. There’s the personality and back story of the creator; the early years, the screw ups, the surprise successes, the drugs, the rehab, and the redemption that makes the masses fall in love with the artist who created something from the heart.

Computers don’t have back stories.

That is, except for HAL 9000 from Stanley Kubrick’s 2001: A Space Odyssey, which recited its history, sang the song “Daisy Bell (Bicycle Built For Two),” and then pleaded for its “life” before astronaut Dave Bowman pulled its plug, as punishment for HAL’s murderous insubordination.

Whether the Singularity will ever arrive and replace musicians, or music itself, is debatable, but we can have fun with AI music in the meantime.

Follow me on Twitter or LinkedInCheck out my website