Why AI could make human jazz popular again

By Jonathan Levin

AS A MUSICIAN and music-lover, the artificial intelligence (AI) revolution terrifies me in many ways. AI apps such as Suno have already shown extraordinary potential to generate catchy and professionally produced music in certain genres. So it isn’t hard to imagine a world in which, for example, session musicians, jingle writers and purveyors of educational music for kids could soon lose their livelihoods to machines.

At the same time, I’m fairly optimistic that jazz — one of the most commercially underappreciated of all the musical styles, and the one closest to my heart — will survive and thrive in the new AI ecosystem. A 2024 by Luminate ranked jazz 10^th out of 11 “selected top genres” in the US, where it was nestled between classical and children’s music and commands less than 1% of total on-demand streams.

AI might be the key to improving on those abysmal numbers by highlighting what I call the “jazz model”: a way of making music that puts live, verifiably human performance at the center. And that model may point to a path of survival for other human artists looking to carve out a niche in our AI future.

To see why, it helps to look at what generative AI actually does well — and what it struggles with. It can mine vast troves of patternistic text, images, audio and video, then turn it into something you might want to consume. That works quite well, for instance, for pop and rock music, in which songs tend to clock in at around 3 to 4 minutes and follow the predictable pattern “verse-chorus-verse-chorus-bridge-chorus-end.”

But great jazz has two things that help set it apart. First, it is often harmonically groundbreaking (Miles Davis’ Kind of Blue, which introduced the world to modal jazz; John Coltrane’s Giant Steps, which cycles through mind-boggling key changes). Second, the genre experiments with novel song forms (Ornette Coleman’s Free Jazz: a Collective Improvisation). More than a century after the birth of jazz, my favorite contemporary players — including the guitarists Julian Lage and Kurt Rosenwinkel — continue to push the boundaries of timbre and harmony.

I’ve tried to test AI’s ability to replicate the depth of the sound and have been disappointed in the output. Suno prompts such as “create an instrumental jazz recording that pushes boundaries of form and harmony; experiment with dissonance and key changes” result in something I might play on my stereo at a polite holiday party with extended family. But there wasn’t anything novel or boundary-pushing about it, nor did it hit me on an emotional level.

Obviously, it’s premature to say that the technology won’t ever be able to create good jazz. Yet even if that happens, it’s likely that we’ll start to distinguish more sharply between craft — polished, repeatable style — and art, which we’ll reserve for creative work that is visibly, even vulnerably, human. A recent on how US adults viewed AI found that 53% of people thought the technology would worsen the ability to think creatively, which suggests many will be looking for ways to believe that creativity still exists.

What genre can do that better than jazz? Imagine piling into clubs like the Village Vanguard, where we can sit so close as to watch the performers sweat and where each performance is improvised, unique and imperfect. In that moment, one can marvel at the way human lungs produce expressive trumpet solos and the human fingers sliding up and down the upright bass. Virtuosos will be celebrated, much as great athletes are, as living celebrations of what we can accomplish with hard work, even without machines.

I suspect similar dynamics will extend to other art forms as well. AI will excel at making pastiche — knocking out competent genre fiction, portraits, and decorative sculpture. But the work we prize the most will be the avant-garde, the risky and idiosyncratic, and there will be greater demand for methods to authenticate that it was produced by human hands. Even creators whose art isn’t traditionally performance-based may have to show their process, perhaps by livestreaming from their studios or sharing unedited drafts, precisely so audiences can experience and reward the distinctively human labor behind the finished piece.

Long before I became a markets and Fed columnist (my day job for this publication), my first dream was to become a jazz guitarist. Seeing how hard it was for working players to earn more than a modest middle-class income eventually pushed me toward another career, but I’ve never stopped cheering for the people who stayed in the music.

It’s been nearly a century since jazz dominated popular music and some six decades since the massive hit albums of Davis, Coltrane, Dave Brubeck and others. Yet, as scary as AI is for musicians on the whole, I’d love to believe that the upheaval will finally bring about a renewed appreciation for the jazz performers that I hold so dear — a group of artists that the world has long taken for granted. — Bloomberg Opinion