A white, chart-topping rapper criticized as “an avatar of algorithm culture.” A young, major-label rock debut dismissed as “an algorithmic fever dream.” A 20-year-old band’s first release after a five-year hiatus bashed as “more like a streaming algorithm than a coherent album.” A mainstream, established pop-rock group denounced as “the machine learning output of the Lumineers, the Chainsmokers, and a SoulCycle playlist.”
Music critics are lamenting the possibility of a machine-driven world that rewards artists not for their originality, creativity, or emotional authenticity, but for their ability to replicate proven, predetermined formulas. Studies show that pop music and lyrics have grown increasingly repetitive and homogenous over the past few decades, and there is a whole graveyard of startups mining streaming and social data to predict the next big hit. Research initiatives like Google Magenta and Sony’s Flow Machines are even training machine-learning algorithms to compose songs on the spot, aiming to be indistinguishable from human songwriting.
ICYMI: We compared a WashPo story about Amazon to a WSJ story about the same topic. We noticed something.
This warped reward system, by which musicians climb the streaming charts, can influence every aspect of a song, from the music and lyrics to the artists’ wider visual and social personas. In the aforementioned “algorithmic fever dream” review, for instance, Pitchfork’s Jeremy Larson described musician Greta Van Fleet’s attire at live shows as “hippie costumes they 3D-printed off the internet.” And Corban Goble wrote in his review of Imagine Dragons that the emotional depth of the group’s lyrics was the equivalent of “Instagram-quote culture—the idea that any snippet of thought, removed from context, can build a base of inspiration.”
“After my review got published, I received a lot of emails being like, ‘Why are you talking about all this other stuff, not just about whether their music is good?’” Larson says. “But criticism isn’t just talking about the music: it’s also talking about the context and environment in which the music is created and consumed.”
One could say the same about journalism, where falling budgets and massive layoffs shape, and often burden, its output. Musicians and the writers who cover them are both working in technologically unique and financially fragile economies. Although the two industries can now distribute their core product online at little to no cost, lower barriers to participation have not translated to better financial health. A recent survey by the Music Industry Research Association (MIRA) found that the median income for US musicians in 2017 was just $35,000; the US Bureau of Labor Statistics claims that median annual wages for reporters and correspondents are not much higher, at $39,370. Meanwhile, virtually every major music magazine has been hit with layoffs in the past two years, in part because of the failure of ad-supported business models and the unrealistic expectations of return from venture financing.
On the consumer side, streaming and social-media platforms have transformed the nature of music discovery, which was previously more proactive by necessity—requiring manual effort to open up a newspaper, dig through crates at a record store, or attend a live show. Nowadays, “discovery” can be as easy and passive as scrolling mindlessly through a personalized feed or shuffling an algorithmically -curated playlist in the background of a holiday party, without help from a critic or other human guide.
Because of its inherently passive nature, algorithmic curation has also made one core function of criticism defunct. Traditionally, critics acted as trusted tastemakers and, in the words of Larson, “consumer guides”—drawing upon their decades of subject-matter expertise to convince music fans about which CDs and vinyl records to buy at their local store. Now, streaming algorithms arguably have more influence over consumers’ listening habits, but in a rather different way: they don’t serve as tastemakers so much as “taste-reflectors,” serving up music with the highest quantifiable chance of reflecting a user’s already-existing preferences.
ICYMI: A master audio storyteller on how to create a powerful podcast
The resulting experience is akin to each individual streaming listener having their own personal “critic” recommending music they’ll like, with a higher hit rate. “There’s a certain sense of music writers thinking, ‘I don’t want a robot to take the job that I and other people in this industry have worked so hard for,’” says Larson.
Perhaps with an undertone of personal resentment, phrases like “algorithmic culture” and the “algorithm economy” have cropped up among critics to illustrate the way aesthetic and commercial motivations shift in this world of passive, automated discovery.
For example, writing for Racked in April 2018, Kyle Chayka suggested that a platform abides by algorithmic culture if it encourages products that reproduce a “Generic Style,” rather than those that reflect more individualistic expression. In such instances, “we find ourselves in a cultural uncanny valley, unable to differentiate between things created by humans and those generated by a human-trained equation run amok,” Chayka writes.
A “Generic Style” in music takes the form of artists and songwriters deliberately tailoring their sound to maximize placement on specific Spotify playlists—a practice that already pervades more mainstream circles. This is what Jeff Weiss had in mind when situating Post Malone as the face of a music culture dictated by machines for The Washington Post. “By ‘algorithm culture,’ I meant the notion of art as something reduced to an integer and formula—a constant infinity loop of ‘recommended if you like…’ playlists,” Weiss says.
Beneath such callouts of “algorithmic culture” lies a scramble to preserve and promote independent artistic thought—and criticism as a creative form. In a world where tech platforms capture a disproportionate amount of financial and cultural value, one can arguably boil down an artist’s chances of commercial success to a list of data points to be mined, including but not limited to streams, likes, and followers. In addition, crowdsourced models of music evaluation—automated blog charts like Hype Machine, DIY lyric annotations on Genius, aggregated rankings on sites like Rotten Tomatoes and Metacritic—paint an illusion of certainty onto public opinion about art.
This climate presents a dilemma to cash-strapped music publications: Is a trend worth covering if it disappears before one is really able to understand its critical and historical significance? If yes, doesn’t our critical language for emerging music become weaker if our writing simply reflects the increasingly transient nature of its subject matter?
Liz Pelly, contributing editor at The Baffler, recently tweeted about this shift in critical language, pointing out that music press outlets have been leaning more and more “away from investing energy into reviews, interviews, editorial, writing in general . . . and into ‘sessions,’ ‘experiences’ [and] playlists.”
Press coverage of music, in other words, is suffering from the same algorithmically borne disease. The most historically renowned music critics, such as NPR’s Ann Powers and the late Lester Bangs, have been unafraid to “go against the grain” in their recommendations—vouching for artists that would have otherwise been doubted, ridiculed, or marginalized. In contrast, by rewarding familiarity, “algorithmic culture” potentially penalizes the very adventurousness in taste that gave these critics their reputation. “It says a lot about the devaluing of creative work into ‘content’ optimized for the most clicks, plays and views—and into fodder that can be more easily branded,” Pelly says.
This is made all the more ironic by the fact that music blogs and publications are part of what make streaming algorithms do their job better. Spotify employs natural-language processing (NLP) models in its recommendation algorithms, analyzing text from blogs, news articles, forums, and other sources to draw connections among different artists and songs, and to figure out what adjectives and moods people associate with these artists online.
Of course, not all writers and editors see streaming algorithms as a threat to their work. “It’s the responsibility of writers and editorial publications . . . to look at how we can incorporate those platforms into the work we do, and use them to broaden the reach and profile of what we do,” Nick Sabine, co-founder of electronic music magazine Resident Advisor, says. In fact, music publications remain some of the most popular independent curators on Spotify and other streaming services. Resident Advisor’s Spotify account currently has over 22,000 followers and has published around 50 public playlists, many of which focus on niche and local electronic styles like proto-house, Chicago footwork and UK garage. Mainstream and generalist music publications have an even bigger presence: Pitchfork’s account has nearly 370,000 followers, Rolling Stone has over 720,000 followers and Billboard boasts over 1.9 million followers.
“Resident Advisor attracts 40 million readers a year; Spotify probably attracts more users than that every single day,” Sabine says. “We could play an absolutely pivotal role in someone’s relationship with and discovery of new music, but they might never come to our website. I see this coexistence as an opportunity for us, rather than as a threat.”
But most critics live on a spectrum from skeptical to resentful. “Any critic who is evaluating art solely on the basis of [algorithmic] influence,” Weiss says, “deserves to be locked into club Liv with YesJulz for 48 hours while a playlist of DJ Khaled screams ‘They don’t want you to win!’ at levels so screeching that it’s only previously been heard in the overthrow of obscure Panamanian dictators.”
ICYMI: Equipment you’ll need to start your own podcast
Cherie Hu is a freelance music and tech journalist based in New York. She writes regular columns for publications including Billboard, Forbes, and Music Business Worldwide.