Robot rock: can big tech pick pop’s next megastar? | Music


One lunchtime about three years ago, Hazel Savage Aron Pettersson set a new piece of software running on a laptop then went to a nearby mall for a sandwich. They hoped, on their return, to have the answer to a question that would change the music industry: can a computer pick a hit record?

The pair had just founded their firm, Musiio, in Singapore’s Boat Quay district. Pettersson, who is Swedish, was a specialist in artificial intelligence (AI) with a background in neuroscience; Savage, a British music industry professional with tech pedigree, had worked for Shazam the Pandora streaming service. They let their software loose on the Free Music Archive, one of the world’s largest collections of copyright-free songs. These are written by little-known artists commonly used for soundtracks podcasts. They asked their computer to pick 20 songs from the archive, based on their similarity to a tune Savage liked: I Wanted Everything by the US indie star Kurt Vile. Back in the office, they listened. “Every song was great,” says Savage, “every song was of a similar genre.”

Savage says Musiio can now run through thousands of songs – submitted as demos or uploaded to streaming services – sort them, according to whether they contain a vocal, whether they’re trap, indie or classical, even whether they bear resemblances to an existing hit, say Uptown Funk by Mark Ronson. Musiio is just one of many hi-tech firms changing the way songs are categorised, playlisted promoted, to eventually reach the ears of millions. They’re fast, efficient get attention for unheard-of acts. But are they also a little, well, inhuman?

Test run … Kurt Vile, whose tune I Wanted Everything was used in Musiio’s experiment. Photograph: NBC/NBCU Photo Bank/Getty Images

For decades, talent scouts or record company A&R professionals used to find new singers, musicians MCs by going to concerts, listening to radio, talking to people in record shops, receiving tips from well-connected pros such as gig promoters, listening to unsolicited demo tapes. Over the last 15 years, though, the process has changed. In 2006, Adele came to the attention of XL records in London via the songs she had uploaded to MySpace. In 2008, Justin Bieber was plucked from internet obscurity after US talent scout Scooter Braun accidentally clicked on a YouTube video. Billie Eilish found stardom via a 2015 SoundCloud upload. In 2019, Lil Nas X was signed to Columbia after reaching a huge audience on TikTok.

Today, anyone can put a song on a streaming service with the hope of being discovered, those platforms have effectively become reservoirs of largely untapped talent. In 2019, Spotify co-founder Daniel Ek said that nearly 40,000 songs are uploaded to his platform every day – way more than could be manually assessed. Most don’t trigger a call from a talent scout, but a few do. While Savage co scan sound files, other companies are digitising the tasks of riffling through social media streaming platforms to count followers, plays “likes” as a guide to a possible hit.

Conrad Withey is the CEO of Instrumental, a British company that uses data analysis to identify, track, profile, rank sign overlooked recording artists across the globe – doing digitally the sort of number-crunching an A&R professional might once have done manually. Most of the musicians, composers, singers songwriters Instrumental signs aren’t going to bother the charts. Instead, they appear in streaming playlists that listeners play again again. “If you’re looking at a return on investment,” Withey says, “getting on those playlists can do really well.”

Illustration
Illustration: Joe Magee/The Guardian

In the past, record companies have settled for one in 10 of their acts finding true success, with big hits underwriting the less profitable ones. But Withey says pretty much all of his less prominent signings turn a profit. He highlights one, Eamonn Watt, a composer from Shetland, whose restful keyboard appears in such Spotify playlists as Peaceful Piano, Classical Sleep Instrumental Study. “He popped up on our tech,” Withey says. “He’s now doing millions of streams but I would never have found him otherwise.”

Instrumental typically picks up on artists who are doing well but registering below 1m streams. “We help artists manage their socials better,” says Withey. “We develop their brand, invest in their new music grow their reach. We also have our own playlists that reach more than 1.5m listeners those help recordings reach new audiences – trigger other playlist editors algorithms.”

Major record labels are using similar techniques. In 2018, Warner Music Group – home to such stars as Madonna, the Red Hot Chili Peppers, Dua Lipa, Coldplay Beyoncé – bought Sodatone. This Canadian startup combines streaming, social media other data (even touring stats) with AI to identify talent. Founders Arjun Bali Jerry Zhang see their organisation as augmenting, rather than replacing, the role of talent scouts. “An A&R used to go to one or two shows every night,” says Zhang. “But the amount of music that’s being released has increased by orders of magnitude. An A&R can’t go to 100 shows a night. We’re just trying to help them to explore the music landscape.”

Could AI have spotted the sea shanty craze? … Nathan Evans, singer of Wellerman.
Could AI have spotted the sea shanty craze? … Nathan Evans, singer of Wellerman. Photograph: tiktok.com/@nathanevanss

Sodatone can process huge amounts of data, picking out tiny events that might indicate something larger. “One artist, one influencer, with one social post can cause a significant shift in what’s cool,” says Zhang. “Or it could be a million teenagers gathered in a specific pocket of the internet.”

Their software tracks bookings at major venues, mentions on music blogs inclusions on playlists charts, as well as support from tastemakers, influencers playlisters. Warner has been using the data to pick up on new talent but also to get the most out of existing artists, by working out where their fans are what they’re into. Towards the end of 2020, Warners indicated that, year-on-year, it had doubled the number of signings identified via Sodatone.

Some believe a greater reliance on hard numbers might lead to less nepotism within the music business fewer opportunities to gerrymander signings charts. But not everyone in the industry welcomes this focus on data; a few wonder how a new artist might fare in this system, plucked fresh from the internet dropped under the spotlight. “When you leave med school does someone say, ‘Here’s a fucking scalpel, start doing surgery?’” asks Lyor Cohen, global head of music at YouTube. “No! You go to residency. So why do people assume a 17-year-old who got some traction knows all the pitfalls that lie before them becoming a successful music artist?”

Cohen is familiar with those pitfalls, having held commanding positions heading up labels such as Def Jam, Warner Polygram helped shape the careers of Run-DMC, the Killers, Jay-Z, Bruno Mars, Public Enemy, Elvis Costello Ed Sheeran. He says that today, an obscure singer, rapper, music producer or bshowing good data points might get multiple offers from labels, little in the way of guidance. “They’re saying, ‘You’re already getting traction so you already know what you’re doing. We’ll only provide you with capital, we’ll get out of your way.’ No one is saying, ‘There are a lot of landmines out here; let’s work together to avoid those landmines.’”

Of course, old methods of star-spotting weren’t always gentle kind. Over the past two decades, pop music’s most prominent system has been talent shows such as X Factor Pop Idol. Withey points out that Simon Cowell effectively shuttered his record label Syco – once home to Little Mix One Direction – last summer argues that talent-show viewers prefer watching TikTok or Instagram to broadcast TV, follow the music from there. “The audience has gone straight to socials. We’ve followed the audience to where they are now.”

Is the music business being “Moneyballed”? Withey co understthe reference to the 2003 book detailing the transformation of the OaklA’s baseball team after the management dropped old ways of scouting players in favour of a data-driven analytical approach.

“What you think ‘good’ looks like is tainted by what has gone before what experience has taught you,” says Withey. “That’s Moneyball; [the old-school scouts] had a fixed view of what a good baseball player looked like.”

Jerry Zhang at Sodatone is more wary. “The rules objectives in sports don’t really change,” he says. “With anything cultural, not only are the participants in culture changing, the rules of the games are changing the objectives of the game are changing. That’s what makes things really tough from our perspective, really fun.”

‘It couldn’t categorise a hybrid of throat-singing electric violin’ … Mongolian throat-singers Huun Huur Tu at Womad in 2013.
‘It couldn’t categorise a hybrid of throat-singing electric violin’ … Mongolian throat-singers Huun Huur Tu at Womad in 2013. Photograph: Andy Hall/The Observer

Savage is aware of her software’s shortcomings. “AI can’t learn something that it hasn’t been taught,” she says. “It can do pop, indie, trap, UK grime, but if you show it a hybrid of Mongolian throat-singing electric violin, then it can’t categorise that.” It probably wouldn’t have picked up on TikTok’s recent sea shanty craze. (Though Sodatone Instrumental’s tech might well have spotted the viewing spikes.) She admits that to truly succeed, you’ll probably need tech services to be complemented by the human touch.

Which means that machines won’t entirely rule the charts just yet. “I believe art music are still in the humanities, in the colleges,” says Cohen with a smile. “I’ll get out of the business when music is in the engineering or math part.”

Yet graduates from those faculties – such as Bail Zhang (both of whom studied engineering at the University of Waterloo, Ontario) – will probably be asserting their influence on the arts for some time. The Sodatone founders believe data analysis informs many aspects of today’s entertainment; firms such as LA’s Parrot Analytics, for instance, predict TV shows’ future success based on viewer enthusiasm.

Hazel Savage is optimistic about her firm’s chances of picking out winning material in other spheres, such as podcasts. She’s also dismissive of big labels that are unwilling to strain a flow of undiscovered tunes through the kind of digital sorting service she is offering.

“If you go on Sony Music’s global website,” she says, “it says, ‘We do not accept unsolicited demos.’ How are you going to be one of the biggest record companies in the world if you don’t want music sent to you? When technology like ours exists, you could find the next great piece of music, why wouldn’t you?”

And if you really love music, why wouldn’t you be thankful for machines that can pick out the first glimmer of the next rising star?



Source link