Part machine, all pop
YouTube sensation’s album composed entirely on artificial-intelligence platforms sparks a conversation about the future of music
If you’ve ever lamented that everything you hear on the radio today sounds like it was written by a machine, brace yourself for the future: The robots are coming — and they’re bringin’ the jams. Well, not nuts-and-bolts robots. Not yet, anyway. But artificial intelligence (AI) is being used to write music, and Los Angeles-based musician and YouTube video star Taryn Southern is getting in on the ground floor by recording the first album composed entirely using AI, which should leave her well placed for sympathy when the robots eventually take over — assuming, of course, that AI is by then sufficiently advanced that robots can actually feel sympathy.
Southern hears jokes like that a lot, as you might imagine. But no, she’s not worried that AI will eventually render human musicians and composers obsolete. And if the day comes when AI does take over music, she says, we’ll all have far bigger worries on our hands.
“I always say if that happens, then we’re all done because it’s not gonna just happen in music. Music would be, like, the most benign area that it could take over,” laughs Southern, 31, who was in Toronto last week to speak (and sing) at the Goo- gle-sponsored, AI-focused tech summit Go North at the Evergreen Brick Works.
Southern, a Kansas native, sometime actress and TV host who’s logged nearly half-a-million YouTube subscribers doing amusingly puerile comedy songs such as “I Think I Farted,” “Coffee Makes Me Poo” and “Wrong Hole (The Untold Story),” also runs a digital consulting firm and production house for brands and media outlets looking for online video exposure called Happy Cat Media.
She does believe a revolution is coming; it’s just a musical one and not a robot one.
While the prospect of using artificially intelligent computer algorithms to create music might inspire griping that such music isn’t “real” or that the person using AI to compose is somehow “cheating,” consider that the electric guitar, the synthesizer, drum machines, samplers and digital production software were all once viewed suspiciously in much the same way. AI, Southern reckons, is simply the next great leap forward in music-making.
“I think it will change music,” she affirms. “I think it will change the music industry completely. Completely, in the same way that beats and samples did whenever they started and everyone was, like, ‘Oh, my god, this is the end of the studio musician. Who’s going to want to bring a studio musician in and pay them $700 when you can use a beat or a loop?’
“But because of this, because of the barriers to entry being so much lower and the cost of producing music being lower, you had the advent of hip hop and EDM. Those two genres of music would not have been born had it not been for samples and loop as a new sort of technological input.
“I think the vast majority of musicians will figure out how to use it to their advantage and work with it in really interesting ways. Of course, there are always going to be people who’ll feel threatened by it, but that’s the case, as you know, with every single technology. And of the people who embrace it and adapt it, they’re going to create all kinds of things which we can’t yet even conceive of.”
Attentive internet watchers have already had a preview of Southern’s forthcoming I AM AI album in the form of “Break Free,” a glacial synthpop track about transcending the limitations of the human body that, at 1.4 million YouTube views, threatens to eclipse her minor 2013 radio single, “Crush,” as her biggest “real” pop hit.
The record is only about half-done, in part because Southern recently received a grant from Google and You- Tube to make immersive virtual-reality videos for three of the songs, but also because she’s using several different AI platforms to compose the music and the process is “totally different depending upon the tool.”
At the moment, she’s using the soundtrack-geared Amper and Aiva programs, Google Brain’s Magenta and the more advanced IBM Watson, which — like France’s Flow Machines, the source of the world’s first AI-composed pop song (“in the style of the Beatles”), “Daddy’s Car,” last year — requires “a fair amount of coding knowledge to work” but allows you to “train” your compositional computer by inputting actual pieces of music.
The less-advanced platforms ask the user to designate various inputs, such as genre, instrumentation, tempo, style and mood, and then deliver a piece of music that fits those specifications according to a complex system of internal algorithmic rules.
The music to “Break Free” was created using Amper.
“It’s a very vague idea and then the algorithm spits something out and I’m, like, ‘Whoa, that is not what I like’ or ‘Oh, wow, that’s better than what I was thinking,’ or ‘That’s different than what I was thinking’ and it’ll inspire a new idea, Southern says.
“So now I write the lyrics to the music because I wait until I get a piece of music that I’m really inspired by and it conjures up all sorts of images in my head, which is the opposite of how I used to write. But I find that part really fun and interesting. Once you find something you like, you can iterate as many times as you want. So there’s a lot of arrangement required to make this work. It’s not like you just press a button and the song is there.
“I can rearrange the tempo, the BPM, the key. I can take segments of the song and take out certain instruments. I can shift the melody and do certain things so that I get to a place where I like it. So it’s a lot of back and forth, just like human collaboration.”
Southern does have a point. With the pop charts mostly dominated in recent years by a small group of songwriting giants such as Max Martin, Dr. Luke and Greg Kurstin, it’s hard to muster a complaint that machinecreated music will make Top 40 radio sound any more robotic. And what, anyway, are all these highly paid songwriter/producer types but “advanced AIs” themselves?
As Southern put it in her talk, Martin is “basically just like a walking AI because he has such incredible pattern recognition for pop music.
“Working with a producer and working with AI are very similar. You’re working with something that is incredibly astute at pattern recognition and understanding what people like to hear within certain genres.”
And what do human songwriters think of all this?
You might expect him to feel threatened, but Toronto singer/songwriter Donovan Woods, who makes a living co-writing with musicians in Nashville, isn’t all that concerned. Indeed, he views the advent of AI composers in much the same way as Southern; if you’re working with a producer, he or she is doing much the same work.
“It’s not much different than that. You’re just feeding a musical idea into another person’s brain and then filtering through their frame of reference and spitting it back out,” he says.
“The thing is, (AI) will just become another tool that kids use to make new, dope music. Just the other day, a producer friend told me about writing with a young producer and realizing that, while the kid is using digital tools, he’s not thinking of recreating versions of actual instruments. “New producer kids are not hemmed in by any constraints of even knowing, say, how a pipe organ operates or what it’s supposed to sound like. They’re just operating on instinct.
“So even if it becomes robots, my guess is that kids will somehow make it dope and awesome, and it’ll still be human.”