Is Artificial Intelligence Saving or Destroying Music?
AI can now write and record songs incredibly well. But is there something about music that is distinctly human, something that an AI cannot replicate?
What makes great music great? Is it technical mastery, emotional payoff, or some other inexplicable x factor, something that could never be replicated by an algorithm?
The answer might be the key to placing judgment on music created by artificial intelligence. If AI can make music that achieves or supersedes human-level mastery, then we humans could be left in the dust as supercomputers build on and refine their own work more expertly than we ever could.
So we're left with the question: is there something about music that is distinctly, inextricably human?
Nick Cave thinks so. In an open letter to a fan, he argued that when we listen to a song, we're actually listening to many songs at once—songs of a person's past, of their identity, struggles, and shortcomings—and AI could never replicate the imperfections and histories which form the bases of music's power.
When we listen to music, "what we are actually listening to is human limitation and the audacity to transcend it," wrote Cave. "Artificial Intelligence, for all its unlimited potential, simply doesn't have this capacity. How could it? And this is the essence of transcendence. If we have limitless potential then what is there to transcend?"
A fair argument. But it's easy to imagine that AI could learn to replicate flaws, that it could texture its compositions with bits of hoarseness and the little trips and stumbles that define some of our most beloved recordings, making us feel connected to the musician on the other side.
Though AI is a relatively new phenomenon, its arrival in music is far from unprecedented. Music has been connected to technology since the phonograph was invented in 1877, making recordings replicable on a massive scale. Before that, all music was played live and listening was an exclusively communal experience. Now streaming services let you listen to any song, anywhere.
AI began to creep onto the scene about a century after the phonograph revolutionized the recording industry. In 1995, David Bowie was one of the first mainstream musicians to use AI by creating the Verbalizer, a computer application that took quotes from various literary sources, cut them up, and reordered them into verses—sort of like your average TwitterBot. Bowie used the lyrics for his album Outside. (Flash forward to 2019, and Bowie is the star of anaugmented reality app that allows you to project a 3D exhibition experience into your bedroom).
Flash forward to today, and intelligent recording softwares are allowing people of all musical skill levels to create far beyond their abilities, and digital workstations allow for endless refinement. Most softwares let users endlessly re-record any section, adjust any note via pitch modulators, restructure rhythms, and program whole orchestras—all at a computer's keyboard. Logic's trigger function allows you to link letter keys to complex chords, which build off each other on their own to craft perpetually mistake-free compositions.
Today there are dozens of AI music composition programs on the market, and some have partnered successfully with humans. In 2016, Sony Music commissioned the songwriter Benoit Carre to write a song with the AI software Flow Machines. Carre penned the lyrics and arranged the song, but the software wrote the music and melodies, and the result was "Daddy's Car," a disconcertingly catchy, Beatles-esque tune that could easily pass as manmade.
Carre is far from the only artist to collaborate with artificial intelligence. In 2018, Taryn Southern, former American Idol contestant, released her debut album I Am AI, a collaboration with the software Amper. "I imagine in 20 years, 'coding' songs will be commonplace," Southern told VideoInk. "It's still incredibly early for AI, but I could see artists using machine learning for all kinds of applications: to mix and master their songs, to help them identify unique chord progressions, alter instrumentation to change style, determine more interesting melody structures based on a musician's given sound and style preferences, even gage their audience's emotional response to a song. The sky is the limit."
If you agree with Southern, then AI is simply leveling the playing field, making music more accessible to people with different skills and opportunities. AI mastering services are saving artists thousands of dollars on an expensive and time-consuming aspect of post-production. And of course, now people who can't play any instruments can suddenly compose and record.
On the other hand, these softwares are sparing people the work of actually learning an instrument, letting them attain the satisfaction of creation without the dedication it used to take to get there—potentially perpetuating the social media era's trend of instant gratification that some argue is putting us all constantly on edge.
But progress will not be stopped, and scientists are making AI more virtuosic every minute. Douglas Eck, the creator of NSynth—a project that utilizes deep learning to dissect and combine sounds to create completely new ones—views AI's advancing musical aptitude as progress, just like any other invention. "We're making the next film camera," Eck told theNew York Times. "We're making the next electric guitar."
Investors seem to agree. AI compositional softwares have amassed millions in funding, and some AI-human collaborations are even edging towards mainstream success. In 2017, the producer Bauer collaborated with musician Lil Miquela—who also happens to be an artificial Instagram influencer, and whose synthetic vocals and CGI-generated appearance garnered her a Times Square billboard ad. And now Amazon's Alexa has a new feature called Deep Music, which automatically weaves audio samples together to match whatever 'vibe' it detects in the surrounding atmosphere— because Alexa wasn't already listening too closely.
But none of this answers the central question: can an AI ever make music as well as a human? Or does it lack something fundamental, some mysterious quality?
According to posthumanists like Claire L. Evans, there shouldn't have to be an either/or, a man versus machine binary. "It's almost exquisitely myopic to judge posthuman music on its ability to "pass" as human," she wrote on Motherboard. Instead, she argues, AI might help us access unprecedented levels of mastery. Combine the technical skills of an AI with the emotional experiences of a human and you might just have a new kind of hit, one that could achieve a cosmic level of brilliance.
After all, music, science, and the cosmos have always been interconnected. Recent revelations in quantum physics have revealed that the universe is inextricably connected to sound; superstring theory proposes that the universe is a form of oscillations that resemble the compressions and rarefactions that form sound waves. Basically, each individual sound wave contains harmonics, which align to form spiral formations when transcribed onto staff lines. In what could be an uncanny coincidence (or proof of divine providence), these formations resemble the architecture of shells or the spirals of galaxies.
Essentially, musical harmonics are mathematical systems reflected by (or reflecting) the structure of the universe. It's not hard to imagine a scenario where AI could develop an understanding of harmonics that's far more advanced than anything we could come up with on our own, one that accesses secrets of the universe we couldn't imagine.
And yet there's still something unknowable involved in artistic creation, something that makes certain pieces resonate while others falter. Music has always been a combination of technical skill, inspiration, science and faith; it seems that ultimately, one can't exist without the other.
Eden Arielle Gordon is a writer and musician from New York. Follow her on Twitter at @edenarielmusic.
POP⚡DUST | Read More...
Fandom for the Faithless: How Pop Culture is Replacing Religion