Since ChatGPT wrote itself into the tech history books, much has been made of what artificial intelligence (AI) means for the written word. But intelligent machines are also making some serious soundwaves across the music industry – and as an ex-musician, this journalist was keen to find out more.
Playing in a band was a mixed experience – some highs, some lows. But one thing that always bamboozled me as a guitar-slinger with not a lot of facility for sound engineering was the technical side of recording music, which has only become more complex since I chose to hang up my Telecaster.
Back in the 1990s, the debate raged: would electronic dance music end up replacing rock’n’roll, as traditional instrumental bands found themselves consigned to the scrapheap by an ever-evolving musical movement of sampled, looped, and remixed beats, grooves, and riffs? As it turned out, that never happened, and the two genres learned to live side by side, sometimes even collaborating.
Now the advent of AI-driven music generation has revived that debate in a format rebooted for the 21st century, moving it one step forward as fears grow that self-improving machines will ultimately replace human composers and producers in the electronic music industry.
Meet Aimi: Alexa’s musical little sister
On the contrary, insists Edward Balassanian, CEO of Aimi, the company behind the interactive AI music-generating app of the same name: “generative music,” will instead lower the bar of entry into the industry. It allows humans and machines to work side by side to create tracks that can continuously tailor and adapt themselves to a listener’s individual mood or need.
“The way we look at making music is very creatively focused,” he tells me. “So we've trained our models to make music the way producers make music. Aimi assembles music and applies effects in real time. It makes decisions about which loops to combine and has generative instruments as well that can play along with other loops.”
I played around with the upgraded Beta version of Aimi, and it is quite impressive – the soundtracks, heavily rooted in dance, electronica, and house, sound much like anything a live DJ might spin. The user can select preferences, programming the app to play more or less of certain beats, melodies, and harmonies across a range of synthetic styles.
Now, as a lifelong fan of old-school rock, country, folk, and blues, I may not be as well qualified as a dance music pro to say whether Aimi can produce longer-lasting and more nuanced electronic tracks.
But what I will say is that, as I write this article, she’s playing a soundtrack that melds synthesized dance beats overlayered with African vocal stylings – and I haven’t gotten bored yet. The music certainly seems to have an uncanny ability to shift and change subtly, keeping to the broad script I opted for at the outset without sounding too repetitious.
Balassanian hopes his AI-driven music generator will be a tool to augment human digital composers rather than put them out of business – for every hour of music recording a producer releases, a hundred are left on the cutting room floor, he claims.
“That's gold to Aimi – we extract close to 500 features from every little bit of music, and our algorithms know how to use those to create exceptional music, much the way a producer does,” he explains. “A lot of that intuition is now distilled into algorithms that can be reused by people who don't have that intuition. But for producers who are sophisticated, this is a huge head start for them. They can drag their studies in and export them in seconds into a multitrack audio that they can do whatever they want with.”
I point out that this doesn’t sound like something that would be employed very far beyond the sphere of electronic music, and Balassanian doesn’t disagree – although he insists that keeping Aimi away from more conventional styles such as country, blues, and rock’n’roll was more a business decision than a technical one.
“We've developed a programming language for generative music. You can write a symphony using any script. Electronic music, from its genesis, has been about sampling and reusing musical ideas,” he says. “And that's something very central to our thesis. We want to encourage not just sharing of the musical ideas, but the techniques that go along with them. And electronic music really lends itself well to that.”
The healing power of AI?
AI in music therapy could constitute another beneficial development. By being tailored to the individual patient, a soundtrack can be created that continually shifts and alters according to their emotional responses – albeit within algorithmically defined parameters – and can, in theory, play infinitely.
“If you use calm or other meditation apps, one of the challenges is they have to record the music,” Balassanian says. “And even if they do a two-hour recording, you'll get tired of it after a while. With generative music, it can just play.”
Because the science behind music therapy – whether AI-driven or not – is also in its infancy, he believes that giving patients more control over what kind of soundtracks they listen to should help them choose the most beneficial therapeutic solution.
“A lot of this science behind music therapy is inconclusive right now,” he says. “There's a lot of evidence to suggest it's good and it works, but we don't quite understand how it's functioning and affecting the brain the way it is. We've talked to some scientists that are using it to affect the way your organs work, using sound frequency to focus on specific parts of your body. And it's fascinating, but difficult to understand how it impacts individuals. So giving individual control is a powerful way to let the patient, in effect, have some input into how the music is affecting their mood. They're the best person to judge whether their mood is good or not.”
Balassanian also believes this generative music will benefit employees in, say, meditation or yoga centers who likely do not appreciate having to listen to the same set of identical recordings on a loop all day at work.
“In the health and wellness space, one of the things that we've learned is that, especially talking to hospitality and other premises, staff well-being is critical,” he says. “And one of the things that they are very concerned about is a playlist that repeats for 12 hours of the day. Customers don't hear that repetition, but the staff does. And it's a problem, right? If you're in a spa, it doesn't matter how calming that music is – if you heard it 40 times in the last two hours, you're going to start getting Looney Tunes.”
They used to be afraid of the Yamaha…
Balassanian and his team consulted more than 200 producers around the world, so they could distill their shared knowledge into Aimi, an app that – he hopes – will allow novice users to tap that expertise.
“You don't need to be consumed with music theory, music composition, or the nuances of your particular style,” he says. “We've been able to distill expert techniques from top-tier producers into what we call Aimi Studio, which is our creator platform. And that means you don't have to know those techniques. We spent the last year working with a lot of these producers to essentially translate their knowledge into reproducible and reusable expert algorithms.”
Didn’t he fear a backlash from producers and DJs who would reject AI-driven music tech as an existentialist threat to their careers?
“I got a Yamaha DX7 as a birthday present in the late Eighties, " he laughs. “And I remember all the controversy: ‘This is going to put people out of business.’ It created a whole new industry of music. I think whenever you're talking about creative disciplines like this, the only threat is really to the lack of creativity. If all your value is work, then yeah, tools will eventually take over that part of it. But if your creative elements are being held back by tedious work, you welcome tools like this.”
Audience participation, digital-style
This revolution in composition and performance need not be limited to the aural sphere either, for the simple reason that music touches on and intertwines with so many other creative media.
“Music is at the heart of so many different mediums of consumption, whether it's video, movies, social media,” says Balassanian. “Music is the one common thing that's across all of them. And generative music has a really unique part to play that we're seeing now, creating spoken audio tracks to remaster the movies for different languages. What they did with Avatar, where they basically used AI to have lip-synching perfectly synchronized to the other languages, that's awesome. You don't get a crappy dub of an actor's mouth moving without synchronizing to the audio.”
That’s why one day he hopes to see a similar principle applied to music for movie soundtracks, where somebody watching a film could select their own generative score to suit their tastes – and perhaps even use that to interact with the narrative and change it.
“There's no reason why the music can't be generative and trigger specific things happening in the movie,” he claims. “We're working on some really cool technology right now that will allow Aimi to not just generate new music, but also visuals that are choreographed with the music. And that's kind of transformative, when you think about things like live performances.”
He adds: “There's a whole conversation right now about the live performance of music, especially in the electronic music space. If you look at the tools that DJs have been using for decades, every new iteration of those is just more buttons and more dials. They haven't really changed, you're still kind of turning a bunch of knobs on a deck. That whole space is just ripe for disruption.”
More from Cybernews:
Subscribe to our newsletter