My short answer to this question is: it’s a real risk, but I have confidence that we will keep re-discovering the essential social benefits of creating, performing, and listening to music with other humans. These nourishing experiences help us get through the challenges of life in a way that AI can never replace.
I am of course deeply troubled by nascent efforts to replace artists with algorithms that can write music faster and cheaper. One of the main reasons I’ve created qBrio is to help demonstrate that the creative ceiling is highest when we empower musicians with AI and new technology. It’s not that I think an AI will never be able to craft a masterpiece (its doubtful but it could happen). Rather, I think outsourcing music making to AI fundamentally misunderstands what music is and the role that writing, performing, and listening to music plays in all of our lives.
Music is not a product like a car. We get the same functional benefit from a car regardless of whether there are humans or robots working on the assembly line. But some of music’s most profound benefits lie in the fact that it is social; these benefits emerge from making music with other humans or being moved by music that you know was written and performed by another human. Some of my most valuable experiences with music include the memory of my mom singing “You Are My Sunshine” to help me fall asleep when I was a child, connecting with the rebellious lyrics of The Offspring when I was in high school, the sense of achievement I felt with members of my college orchestra after we performed Mahler 2, and the camaraderie I felt with the rest of the audience at a Chance The Rapper concert in Chicago. The common element here is music as a means of emotional expression and social connection .
Just think how hollow these experiences would be if it were an algorithm singing me a lullaby instead of my mom, or if I were performing Mahler with an orchestra full of robot musicians. Regardless of how well crafted the lullaby or talented the robot musicians in these scenarios, there would be something lost. This is because music is not a product, its an enriching experience where the enrichment comes from interaction with other humans.
This is not to say that music has no functional benefits like other consumer products. There is overwhelming evidence that music can produce specific physiological and emotional responses that are good for us. Certain types of music, whether written by a human or an AI, can systematically help us sleep, exercise, focus, relax, and even recover from disease. These are remarkable breakthroughs. And if AI-written music can help me exercise more effectively or relax more quickly than human-written music, I say bring it on.
AI might also help more people experience the joys of playing music. To take my example from above, while there would be something important lost if I were to perform a Mahler symphony with robot musicians instead of humans, it would still probably be fun and even exhilarating. What’s more, not everyone has access to a 90-piece orchestra. If AI musicians are more accessible and make it easier for more of us to play music when we don’t have a local orchestra or can’t find people to join our band, that is a good thing.
We need to be vigilant, however, of the key, delicate ingredient behind great music: artists. Masterpieces emerge when artists dedicate themselves to their craft and apply their full passion and creativity to music making. In order to support this time, effort, and focus, artists need paying gigs. From Bach working as the music director at a cathedral in Leipzig to Drake working a songwriter for Dr. Dre early in his career, artists have always needed gigs to get started.
An uncomfortable truth is that there are soon going to be some gigs an AI can do better than a composer. Jingles for commercials, for example, don’t need to be masterpieces. They might even be more effective if they are composed by an AI that can tailor the jingle to each of our listening histories and adapt in real time to our biofeedback. I don’t dispute that an AI might eventually do a better job of this, on average, than a composer. But at the same time, we can’t lose sight of the fact that these are the types of gigs budding artists use to support themselves. You never know how early gigs will contribute to an artist finding their unique sound or stumbling across the seeds of a masterpiece. Childish Gambino and Ludwig Göransson met while working on the TV show Community. Now they are crafting number 1 hits and composing groundbreaking soundtracks for Hollywood blockbusters.
We need to maintain an ecosystem where there are enough opportunities for artists that want to dedicate their lives to music, while letting AI take over gigs that don’t require or benefit from an artist’s full creativity.
On a concluding note, one part of music that I especially enjoy are covers. I am constantly surprised and moved by the creativity of artists covering famous pieces. And I think the very act of creating covers, using the same melodies and harmonies to convey new, different emotions, is fundamentally human. Here are three wonderful covers that were popular submissions to our frisson dataset:
There is already an AI that, after being fed every Bach chorale ever written, can produce new (marginally different) chorales such that human listeners can’t distinguish whether they were written by the AI or by Bach himself. But I think the creativity and audacity to create covers like the above, which give listeners new, authentic meaning from a piece they already know, that is a part of music that AI is unlikely to ever replace.