![Cover Story: Station to Station: The Past, Present, and Future of Streaming Music]()
I. Merging into the Stream
On-demand streaming music has been part of the collective imagination for more than a century. It can be traced back to the 1888 publication of Edward Bellamy’s million-selling science fiction novelLooking Backward, in which a man falls asleep in 1887 and wakes up in 2000. Amidst the mind-blowing technological developments he encounters on his journey is a “music room,” in which 24-hour playlists are piped in to subscribers via phone lines. With no shortage of astonishment, the man proclaims that “an arrangement for providing everybody with music in their homes, perfect in quality, unlimited in quantity, suited to every mood, and beginning and ceasing at will” is perhaps the pinnacle of human achievement.
The splashy, celebrity-laden debut of Beats Music earlier this year may not have been accompanied by such gobsmacked wonder, but at the same time, the smartphone-based music subscription service sponsored by AT&T is the latest iteration of Bellamy’s fantastic 19th century notion. Beginning with Pandora’s 2005 launch and dramatically ramping up with Spotify’s controversial 2011 debut, streaming has become the preeminent technological force driving digital music into the 21st century. Though the idea of streaming music pre-dates recordings, the industry’s investments in today’s technology is designed in large part to wrench back control via unlimited access after a decade of ceding power to mp3-downloading fans.
So far, it’s working. According to Nielsen SoundScan’s 2013 report, sales of single mp3 downloads declined 6 percent from 2012, while streaming activity increased by 32 percent. The Recording Industry Association of America’s own data reveals that sales of physical media declined 12.3 percent between 2012 and 2013 while paid subscriptions to streaming platforms increased 57 percent. CDs and mp3s won’t simply disappear—they’re still vital parts of digital music's ecology—but faced with streaming, they feel destined to become the digital equivalents of once-dominant analog predecessors like vinyl records and cassettes.
Though streaming platforms are very much a product of the digital-era presumption that all the world’s information should be accessible with a single click, their form and function derives from another early music medium. A few decades after Bellamy’s book captured the imagination of millions, and at the same time that the business of selling records was taking off, “music rooms” were manifested by broadcast radio. Nationwide, parlors were filled with sound by national radio networks like NBC and CBS, which interspersed music with periodic bursts of news, narrative programs, and advertising. From the 1920s forward, the business of selling and consuming music has been structured by a technological dialogue between programmed music streams and individual recordings.
If the recording industry has its way, music ownership will give way to a model completely based on access, but with an important shift. While radio broadcasts are based on a one-to-many model of transmission, streaming platforms aim to zero in on the tastes of the individual listener. Like many other modern industries, the recording industry is doubling down on big data, giving their catalogs to the coders, and betting on a future of distribution and discovery dictated by quantification. Behind the interfaces of streaming platforms are vast databases of songs coded with pinpoint metadata and matched with freely provided listener taste preferences, an infrastructure designed to execute the recording industry’s century-long mission: suggesting with mathematical detail what a listener wants to hear before they know they want to hear it. Combing through a huge corpus of ever-expanding data for each individual song can be a vastly different undertaking compared to older forms of music marketing and distribution. What used to be a question of persuasion has become a problem of prediction.
Pop music was born and nurtured on radio’s non-stop streams of ad-supported music—the most prominent ancestor of the feed-driven landscape that shapes 21st century digital streaming.
Listeners are well-served by streaming platforms, but for artists, they cast the question of compensation in a stark new light. While the value debates that dominated the mp3 moment pitted fans against artists, the emergent streaming era has so far seen the return of corporate exploitation, with a speculative twist: The rich or soon-to-be-rich build innovative products, convince an ailing recording industry to sign over their catalogs, acquiring the bricks-and-mortar of their operations—digitized recordings—for fractions of a penny on the dollar. These operations are mostly funded by venture capital, periodic rounds of investments, or as cogs in vast empires of information, and they can feel overwhelming for fans and artists alike.
At the same time, streaming encompasses much more than locked-down platforms. At the most basic level, accessing digital files stored on distant servers through an online player is a form of music distribution capable of being deployed in countless ways. While the biggest streaming players can swallow small artists whole, all is not lost for those with little interest in big data. For every Pandora and Spotify, there are upload-friendly, embeddable platforms like YouTube, SoundCloud, Bandcamp—not to mention the embedded players omnipresent on countless music websites—which have proven incredibly useful for DIY-level promotion and retail.
Digital streaming’s predictive algorithms and on-demand capabilities can feel futuristic, but they are built from the fundamental ideas of several music ancestors. Remember, pop music itself was born and nurtured on radio’s non-stop streams of ad-supported music—the most prominent ancestor of the feed-driven Twitter/Facebook/Tumblr landscape that shapes 21st century digital streaming. When cultural critic Geoffrey O’Brien described a radio stream as "an artwork that manages the flow of information and music” within which “the thing coming up will always surpass what went before,” he could have been describing wading through a social media feed packed with audio embeds, letting YouTube sidebar recommendations guide a late-night deep-dive, or pruning a personal Pandora soundtrack designed to deliver the perfect songs, one after another. All new media inevitably bear traces of their forebears, if only metaphorically. Peel back the technological layers of streaming platforms and alongside databases, algorithms, wi-fi signals and smartphones, elements of radio broadcasts, jukeboxes, record changers, record stores, and telephone companies can be spotted amongst the sediments.
As streaming takes center stage for music commerce, questions with long histories must be reframed. In what ways are the non-stop interactions between databases and algorithms shaping our musical tastes? Do streaming platform business models inherently exploit artists when listener choice scales to infinity? Should speculative capitalism be the driving force for large-scale innovations in music technology, and is there a feasible alternative? Are we living in a technological golden age of creative possibility, cross-cultural communication, and sheer abundance, or a surveillance state controlled by privately-held brands promising endless access at the expense of imperceptible control? Answers to these questions are piloting digital music deep into the 21st century, but critically evaluating current technological developments means keeping an eye on the lessons of the past.
Next: Everyday Disc Jockeys: A History of Song Shuffling and Music Programming
II. Everyday Disc Jockeys: A History of Song Shuffling and Music Programming
Long before the iPod claimed ownership over “shuffling” music, and internet radio turned the practice into an algorithmic science, the practice of compiling and recombining musical works was a staple of music culture. Historian William Weber points out that as early as 18th century Europe, concert programs for the upper-classes reflected a broad diversity of compositions, reflecting the taste factions arising at the time. The dawn of mechanized music first emerged as player piano rolls, many of which were programmed to play not a single song, but a medley of well-known and current tunes.
As 78 rpm recordings took hold and a recording industry took shape, the early 1920s brought the automatic record changer, which allowed phonograph owners to create streams of uninterrupted music in the home, just as radio was starting to do the same thing. The jukebox took shuffling records to an industrial level—these boxes could be as aesthetically pleasing as a living-room Victrola, and as sharply designed as a new Oldsmobile. In the midst of radio’s Golden Age, everyday disc jockeys plugging nickels into machines were creating their own personalized stations in public spaces nationwide. By the late 1930s, music historian Elijah Wald estimates, between 50 and 60 percent of all recorded music was going directly into the 400,000 jukeboxes across the country. Because of their popularity, jukeboxes were also the site of mechanized innovations in understanding musical tastes: Via mechanized play meters built into the boxes, owner-operators quantified the aggregate tastes of local areas—the music popular in bars and soda fountains, at least—to exacting degrees.
As industry lore tells it, jukeboxes’ capacities for providing such detailed listener data redefined the possibilities of radio, which was hemorrhaging audiences to television by the 1950s. One night in the early 50s, Todd Storz and Bill Stewart of Omaha’s KOWH posted up at the bar across the street from their station to discuss marketing ideas. Over the course of a few hours, they noticed that patrons were playing the same jukebox selections over and over, all night long. Storz and Stewart merged this revelation into KOWH’s operating philosophy, creating the first Top 40 station. Instead of radio’s long-held focus on variety programming (which TV was doing much better), the tight, data-driven playlists of Top 40 stations drilled home the idea that the average music listener wanted their tastes flattered, and would happily listen to a format that delivered their preferences right back to them.
The celestial jukebox idea merged a beloved cultural form with the immeasurable scope of the heavens, giving notoriously tech-illiterate CEOs and politicians an idea around which to build a new industry.
Forty years after the KOWH revelation, and well after jukeboxes had entered the nostalgia phase of their cultural life, they were symbolically reappropriated to sell an idea for sending music through the most game-changing new communication network since radio. The phrase “celestial jukebox” emerged from a book by Stanford Law professor Paul Goldstein, in which he tried to convince the major entertainment conglomerates and their legislative associates that, with some tweaks to copyright law, they could monetize their back catalogs in an exciting new medium. As a metaphor for digital music distribution, the celestial jukebox merged a beloved cultural form with the immeasurable scope of the heavens, giving notoriously tech-illiterate CEOs and politicians a useful idea around which to build a new industry.
But 1998's Digital Millennium Copyright Act was, to put it lightly, not the legislative response that Goldstein’s celestial jukebox needed. Copyright regulations became much more restrictive under the DMCA, not less; the specter of universal access to music was undercut by the desire for strict control over digital music’s circulation. This was guaranteed by the DMCA’s allowance of Digital Rights Management (DRM) software to be embedded within music files sold online. Even at the time, it was clear that the celestial jukebox model represented the corporate-owned major labels trying to drag outmoded 20th century ideals of consumer control into a very different technological climate.
The celestial jukebox is mostly remembered today for how dramatically it failed. RealNetworks launched MusicNet in early 2001 with the EMI, Warner, and BMG catalogs, while Sony and Universal started the very similar service Pressplay. For between $9.95 and $24.95 a month, subscribers could navigate woefully incomplete artist catalogs; ugly, ad-laden interfaces; and bizarre usage restrictions. MusicNet limited listeners to 100 songs per month—they were streamable once each, or if downloaded, they were rendered unplayable after a month. Only Pressplay’s premium subscription allowed 20 tracks per month to be burned to disc, with no expiring downloads. Quietly, as these major label-led mistakes gathered press, the contemporaneous playlists-by-experts model used by Rhapsody—which, more than a decade later, claims to have passed one million paying subscribers—started catching on.
Understandably, music fans in the late 90s and early 2000s were much more fond of Napster’s peer-to-peer model of celestial jukebox, through which they could explore endless folders of mp3 files on strangers’ hard drives and download songs with no limits apart from storage space, bandwidth, and a fin de siècle strain of musical morality. Napster’s interface, merging a database architecture with a real-time search engine and chat capabilities, was far more innovative than the e-commerce portals favored by the major labels, though the exploration process was far from orderly. The early days of peer-to-peer downloading were a descendant of the early radio practice of DXing, or distance listening, in which, as radio historians Susan Douglas and Michele Hilmes have recalled, a generation of tech-savvy young men set up DIY rigs in their garages and tried to locate strange aural transmissions from far away. Like DX transmissions, peer-to-peer downloads were notoriously unpredictable and unreliable, and much of the pleasure in the practices came in the joy of technologically facilitated discovery and connection.
Personally, I never did much with Napster. But I was introduced to its descendant Soulseek in 2003. While nostalgia likely clouds my impressions, the practices of digging through strangers’ folders and leaving slow-bandwidth downloads running overnight only to wake up to entire out-of-print discographies in the morning were perks, not bugs. This was not the case for Brian Whitman, who began playing with peer-to-peer in the years before starting his PhD in Media Arts and Sciences at MIT. Napster “was a turning point for music access, but probably a step back for music discovery,” Whitman wrote in a 2013 blog post. “The search was abysmal... and there was no discovery beyond clicking on other users’ names and seeing what they had on their hard drives. I would make my music available but, of course, no one would ever download it because there was no way for them to find it.”
So Whitman went ahead and tried to solve such problems. While most music fans had probably never heard of the Echo Nest before Spotify recently bought the self-described “music intelligence platform,” anyone who has used Rdio, Vevo, iHeartRadio, Spotify Radio, and countless other streaming platforms has been guided ever-so-gently by the database-driven music profiling company Whitman launched with fellow MIT PhD Tristan Jehan in 2005. As I type this sentence, a ticker at the bottom of the Echo Nest’s website claims that it currently catalogs more than a trillion data points about 35 million songs by 2.6 million artists—an ever-expanding corpus of big data that is powering significant portions of the 21st century’s celestial jukebox.
Next: More Like This: Algorithms and Musical Taste
III. More Like This: Algorithms and Musical Taste
“The record listener is a child of the supermarket,” wrote Evan Eisenberg in 1987's The Recording Angel, the ur-text on the social and philosophical implications of turning music into consumable things. For Eisenberg, music fandom is inextricable from shopping, and “self-expression is almost entirely a matter of selecting among packages that someone else designs." He adds: "That kind of freedom can be tyrannical.” Music’s ongoing digitization into libraries of songs numbering in the tens of millions requires the consistent development of technological solutions to this tyranny of choice, to alleviate the minor-but-pervasive stress of simply deciding what to listen to amongst infinite options. Increasingly over the last couple decades, this work is being done by algorithms.
The technological systems that circulate and contextualize music have long been powered by algorithms of various levels of complexity. When jukebox owner-operators consulted play-counter data to restock jukeboxes, they were working with a primitive music recommendation algorithm, as are program directors at Top 40 stations, those overseeing Billboard’s rankings, and the engineers who code iTunes’ shuffle feature. Each of these practices necessitates a series of data-crunching operations designed to produce a specific result.
These pale when compared to the complex big data algorithms powering social media and other platforms based around ceaseless flows of information and recommendations, which collectively have come to be called simply "the Stream." Facebook’s news feed, for instance, queries thousands of factors every time the page is reloaded, interpreting your past behavior and current desires to determine whose updates—and which ads—are most important to you at that moment. Google’s PageRank, Twitter’s trending topics, Netflix’s recommended films, and countless other modern cultural conveniences are powered by impossibly huge recommendation systems dictating not only the cultural choices we make, but the lists of options we’re aware of, and the tastes we develop.
The early moments of product recommendations took shape as collaborative filtering, a technique dating back to the late 1980s that was popularized by the widely-circulated 2002 book Word of Mouse. Collaborative filtering, the authors contend, is a revolutionary update to demographic-driven market research, through which individuals are categorized based on quantifiable data like gender, race, age, and geography. Through collaborative filtering, consumers are understood not as members of social groups, but as individuals with fickle tastes, determined by their product ratings and purchase histories, and this data is matched with the tastes of like-minded others. As the Echo Nest's Whitman tells me, while collaborative filtering was all the rage in the late 1990s and early 2000s, its modern utility—largely visible in “those who bought this also bought this” recommendations—has reached a limit.
“Collaborative filtering will tell you that Radiohead and Coldplay sound sort of similar—popular stuff and other popular stuff are similar,” he says. "But it ignores everything about the thing itself. It’s only looking at the data about purchase or play. The object could be anything—underwear, or a song.”
More recently, computer engineers have looked to content-based recommendation as a way to address music-as-music, not simply as a generic commodity. Under this heading falls what’s long been called “machine listening”—epitomized most popularly by the Shazam app—in which songs are scanned for musicological factors and matched against those of other songs in infinite configurations. The Echo Nest uses machine listening, but it’s far from the company’s most important innovation. That would be its unique process of data retrieval and curation, which entails scraping information from social media platforms, Wikipedia entries, album reviews, and blog posts, which employees then shape into metadata, attached to songs and artists. When describing this labor-intensive aspect of the coding and recommendation process, Whitman suggests the Echo Nest is a living creature with an endless appetite: “If there’s a new artist, we’ll ingest it and try to learn about it.”
These ingestions lead to new ways of imagining the users of streaming platforms, and determining what they want. “The fact that the first song you type into your streaming service is indie rock, and not classical, tells us so much about you as a person,” Whitman says. The more a user interacts with an Echo Nest-powered platform, the more their own taste profile is matched up against countless metadata points the Echo Nest has already compiled.
“We don't just see that you have liked a song, we know about that song," Whitman continues. "To us, a song is not just a database entry, it’s the key, the tempo it’s in, the instruments.” There are also notoriously subjective cultural factors, as well. “Beyoncé will have [data points like] ‘sexy female vocalist,’ and ‘married to Jay Z,’” he explains. What fills the most important slot—the next song to be played—Whitman views as “a huge prediction problem.” Ideally, the more a listener uses a platform, the more of her taste data she allows the system to ingest, the more the Echo Nest gets to know her.
For decades, musical tastes were understood as deriving from one’s economic position and access to education, and were correspondingly slotted into “high” and “low” categories. The privileged classes preferred the elegance of Bach's The Well-Tempered Clavier, while the working classes preferred the winsome pop of Petula Clark, sociologist Pierre Bourdieu outlined in his widely disseminated 1960s study Distinction. In his 2007 book Celine Dion’s Let’s Talk About Love: A Journey to the End of Taste, music critic Carl Wilson finds many limits to this idea.
“In a hyper-mediated, mass-production culture, a lot of reference points are shared across classes,” Wilson concludes, when trying to evaluate his distaste for the Canadian chanteuse. According to an article published by sociologists Richard Peterson and Roger Kern in 1996, Wilson reveals, the “omnivore” is the new model for the music connoisseur, and one’s diversity of listening across the high/low spectrum is now seen as the social signal of refined taste.
Through his PhD research into the engineering end of music recommendation platforms, UC Irvine’s Nick Seaver discovered that Peterson and Kern’s work was familiar to Pandora’s advertising department. This is far from the sociologists’ intention, but is very interesting for its reapplication. “They can score someone's omnivorousness by applying a kind of diversity metric,” Seaver tells me. “A higher diversity score should indicate a higher social status, which means that these listeners can have more expensive ads sold against them.” Seaver, a cultural anthropologist by training, is impressed by the theories used to justify the work of algorithmic music recommendation. “Every Echo Nest blog post is like a chapter in some new sort of social theory textbook,” he admits, only half-jokingly.
Do taste topiaries run the risk of sealing listeners into self-flattering cocoons? Is the tendency toward diversity mitigated by the
safety of algorithmically-determined similarity?
Like many music fans, I consider my listening habits to be fairly diverse. But I also realize that algorithmic-determined internet radio streams are not designed for those of us who already spend large portions of our free time (and work time) researching, acquiring, and sorting music. This has long been the divide between the “lean-back” approach to radio listening and the intense, deep-dive approach of aficionados.
As I started playing with Pandora, iTunes Radio, Rdio, and Spotify Radio, however, I started becoming more intimately familiar with the listening labor required to construct algorithmic taste profiles. Providing positive or negative feedback for each new song was both irresistible and slightly maddening. Chiding and rewarding these systems felt like a mix between topiary gardening and a Rorschach test: a non-stop process of pruning my tastes, spurred by my instant reaction to new information. Such acts of pruning, I realized, are perhaps the defining social activity of life in the Stream. Voting “thumbs down” on a track for not immediately satisfying me is akin to liking a status update or a tweet, or even voting on the likeability of friends themselves by “hiding” certain folks in my Facebook feed, or periodically treating my list of Twitter followees like an overgrown bonsai tree.
Having failed to discover anything new or train the platform to make a stream that could match one of the shuffled playlists in my mp3 library, I tired of my ad-laden experiment with algorithmic taste pruning. But I didn’t stop wondering about its effects. Do taste topiaries run the risk of sealing listeners into self-flattering cocoons? Is the tendency toward diversity—determined on internet radio stations by a knob that slides between familiarity and “adventurousness”— mitigated by the safety of algorithmically-determined similarity? Is this something to fear, or just the big data version of niche marketing tactics that took hold on radio decades earlier? Does internet radio create a virtually cloistered local scene for the age of infinite “curation”? A Top 40 mobius strip to each individual?
As I conducted my ad hoc experiment with internet radio taste topiaries, I realized that these “stations” are much closer to sentient versions of the digital music playlists I'm much more fond of creating on my own. By constantly presenting me with computer-generated reflections of my own preferences, they're automating the pleasurable act of music recombination that I’ve been doing since making my first pause-record mixtapes in middle school. In the digital landscape, music obsessives are more likely to use publicly shared playlists to proclaim the diversity of their tastes, and there’s no place better than Spotify to do so (though Rdio, launched in 2010 by the inventors of Skype, comes awfully close).
With a catalog boasting more than 20 million songs, with 20,000 added daily (20% of which go unheard—a bug the Echo Nest will ostensibly help fix), Spotify’s primary appeal is as a massive data dump organizable into playlists: They claim more than 1.5 billion have been created so far. Spotify is the coolest and most useful platform to emerge in music streaming’s second wave. But because of its focus on providing free access to a massive catalog of material, it's also become the most controversial digital music service since Napster.
Next: Spotify: The Post-Piracy Platform
IV. Spotify: The Post-Piracy Platform
During the 1930s and 40s, playing records on the radio and in jukeboxes was a highly contested practice. Particularly for artists who made their living performing live—playing on the radio and in clubs—the fear was that radio royalties wouldn’t come close to compensating for the cannibalization of live audiences, and sales of sheet music and recordings. Back then, many artists decided to take public stances to protect their livelihoods. Bandleader Fred Waring went as far as stamping his Victor recordings with the phrase “Not licensed for radio broadcast” in the mid-30s, and when he learned a Philadelphia station was playing them anyway, he took it to court. In 1939, ASCAP pulled all of its material from radio for several months, unable to settle on a higher royalty rate with broadcasters. In response, broadcasters formed the BMI licensing agency, which filled its ranks with the hillbilly and R&B music that ASCAP wouldn’t touch. In the 40s, indie label Decca Records pulled its records from radio in favor of stocking singles in jukeboxes. While performers’ unions claimed that jukeboxes were putting live musicians out of work, many smaller acts that appealed toward the young, minority audiences at soda fountains and bars were reaping the promotional benefits of jukebox play.
Pulling one’s catalog from Spotify—or, at least, vociferously airing grievances about their fairness to artists—has become the 21stcentury update of those gambits during radio’s first decades. In a piece published on Pitchfork in 2012, musician and writer Damon Krukowski critiqued the landscape for DIY and indie artists in the 21st century, noting that “industrial capitalism on a [small] scale” had given way to a model of “financial speculation” that strives to turn profits for investors on the backs of cheaply streamed content provided by artists. The former Galaxie 500 drummer also opened the books on his streaming royalties, demonstrating how paltry the payouts are for decidedly non-paltry numbers of plays.
At a much higher level of visibility, Thom Yorke pulled the Atoms for Peace catalog from Spotify in mid-2013, taking to Twitter to condemn the platform’s compensation model for up-and-coming artists; he’d later memorably classify Spotify as “the last desperate fart” of the recording industry’s “dying corpse.” A few days later, David Byrne extended Yorke’s Spotify critique to “the internet,” which he claimed "will suck the creative content out of the whole world until nothing is left." In October 2013, Jana Hunter of Baltimore indie act Lower Dens— the exact sort of band that Byrne and Yorke aimed to support with their public stances—took to Tumblr to argue: “Music shouldn’t be free. It shouldn’t even be cheap.”
Spotify’s attempts to balance free access to its vast catalog with fair compensation to artists derives in large part from a particularly Swedish way of thinking about digital music. Stockholm-born Daniel Ek, a precocious, slightly cocky entrepreneur, founded Spotify in 2006 while CEO of µTorrent, one of the world’s most popular BitTorrent clients. In its early form, Spotify became instantly popular in Europe as what communications professor and author Patrick Burkart calls “a rogue index,” offering vast access with no licensing.
“The labels played ball,” Burkart explains, “because the Swedish government was advocating on their behalf, and everybody was trying to find a solution to [BitTorrent platform] the Pirate Bay.” When Swedish police—under significant international pressure—raided the Pirate Bay’s hosting service in 2006, the politically minded Swedish Pirate Party suddenly gained international attention. Burkart, whose new book focuses on Pirate politics, explains that the movement is “informed by the experiences of music fans, independent artists, and tech-savvy Internet activists who collectively reject the unilateral terms of participation in the celestial jukebox.”
As the Spotify concept developed, Ek shaped it as a piracy-killing Napster-with-permission that could turn an estimated half-billion illegal downloaders into paying subscribers. In 2010, former Napster investor Sean Parker invested $15 million in Spotify, then helped broker a crucial partnership with Facebook. Parker’s uniquely American sense of cocky tech-entrepreneur rebelliousness (see: The Social Network) fit well with Ek’s Scandinavian-bred sense of openness, coupled with his salesmanship and skill at coding.
For years now, Sweden has become synonymous with music piracy. The reasons, Burkart tells me, lie in the country’s cultural makeup and its recent legal and political histories with respect to copyright. “Sweden lagged behind other EU countries in terms of implementing directives on copyrights and data retention,” he explains. “Swedes had a little bit longer to experiment on the net and develop alternate delivery systems.”
Sweden is a socialist democracy with a proud history of entrepreneurialism, as well, and Burkart describes a Swedish legal tradition called allemansrätten, a property statute that translates to "every man's right." “It’s most often applied to hiking, boating and skiing across private property,” Burkart says. “As long as you don’t disturb it, you can even camp for a night, collect berries and mushrooms and do non-destructive things in this commons.” To tech-savvy Swedes like Ek, who came of age in the heady times of the late 90s and 00s, entrepreneurialism was filtered through this model of infinite access—the celestial jukebox-as-cultural-commons.
Like many 21st century tech startups, Spotify itself has yet to turn an actual profit—because of its business model, the more Spotify expands in the U.S., the more money it loses.
Several years after Napster’s inability to strike deals with the major labels, Spotify needed the participation of the multinational corporations who control around 80% of the world’s recorded music in order to launch its ostensibly piracy-killing platform. The solution was one that went directly against the commons idea: offer the majors equity in the company, an agreement that means that somewhere between 18 and 20 percent of Spotify is owned by the three remaining major labels. (The exact number isn’t known because the deals were struck under binding non disclosure agreements.) For many artist advocates, this is one of many troubling aspects of the service.
“The major labels are able to build a new business on the backs of artists without telling them very much at all about how their compensation actually works,” Casey Rae, an attorney and label-owner, tells me. Rae is the interim executive director of the Washington D.C. nonprofit Future of Music Coalition, founded in 2000, which is devoted to musicians' rights. Rae and many others within the FMC aren’t inherently predisposed to hate platforms like Spotify, but they worry about the sustainability of the streaming model, for artists as well as other independent platforms. “There are only three major labels left, so it's not even really like Spotify can have a competitor enter the marketplace,” Rae explains, "because any competitor would have to be extraordinarily well-capitalized in order to even get to the negotiating table.”
Even though Spotify’s effects on revenue were an unknown, it was too compelling an opportunity to pass up for most indie labels. “Early on, Spotify was this weird foreign threat,” remembers Chris Swanson, co-owner of the Secretly Label Group, which includes Secretly Canadian, Jagjaguwar, Dead Oceans, and the Numero Group. “It was this cloud in the distance that we were all reading about, seeing how it was impacting overseas markets in some positive and negative ways.”
The payout numbers Swanson and other label owners were being offered were significantly lower than the 70 cents-per-mp3 that iTunes paid. Nonetheless, the Secretly Group signed up, and were able to negotiate decent terms. “We just didn't want to be missing from a platform,” Swanson says. With a few years of hindsight, he surmises that holding music off of Spotify “is almost like it's not on the grid, which maybe forces people to go buy records, but it actually means people are talking about the music less.” But Spotify’s ties to peer-to-peer—a huge catalog, an on-demand interface, and a free, ad-supported version—means it threatens profits more than any other platform.
Still, like many 21st century tech startups, Spotify itself has yet to turn an actual profit itself—its value is based on investors’ speculation of its future success as a business model. This leads to what appears like a paradox, or at the very least, terrible PR. On one hand, Daniel Ek is a very rich man. On the other, Spotify’s yearly revenues are exceeded by its fixed costs—marketing, payroll, and especially licensing fees and royalty payments, which themselves account for 70% of revenue. (Spotify claims to have paid $1 billion in royalties to artists as of the end of 2013.) Because of its business model, the more Spotify expands in the U.S., the more money it loses.
In December 2013, Spotify took a small step toward transparency by releasing an information-laden website detailing many aspects of its operation, including its royalty payments to artists. One of the more interesting revelations was that during July of last year, a “niche indie album” collected $3,300 from plays on the platform, and a “breakthrough indie album” pulled in $76,000. I reached out to several labels trying to discern the origin of such categorizations. Unsurprisingly, no one had ever heard of them. Spotify’s “niche indie” classification, as it turns out, is about as useful in a descriptive sense as “classic hits” are for radio listeners. Its purpose for Spotify, perhaps obviously, is pure public relations. “Niche indie” artists are what Spotify wants them to be.
Talking about these Spotify classifications and numbers, Jeremy DeVine, owner of NYC-based indie label Temporary Residence Ltd. (Explosions in the Sky, Eluvium), says, “We get paid almost exactly $3,300 a month right now from all on-demand streaming services. For the entire label—215 titles—we make between $3,300 and $3,500 a month.” As for the “niche indie band” categorization? “That’s a band who can sell as many of one record in a month as an entire independent label that does fairly well sells for their entire catalog in the same timeframe,” DeVine estimates, citing Beach House and Vampire Weekend as possible examples.
Over the past few years, DeVine has devised his own experiments at Temporary Residence to understand Spotify’s threats more directly. “We don’t release records to on-demand streaming platforms on release dates anymore,” he explains. Such a delay—which other labels have adopted—echoes the past decade’s label strategies devoted to avoiding mp3 leaks ahead of release dates. Keeping new records out of certain channels around the time of release—treating Spotify as the new MediaFire, in a sense—is widely viewed as the best way to translate release-day promotion and hype into actual sales.
DeVine had a different, somewhat grander experiment in mind a couple years ago. When he noticed that his label’s digital and physical revenues were declining after offering them on Spotify, he devised a plan: pull the entire Temporary Residence catalog off on-demand services for one calendar year and then compare those revenues to other years. But DeVine decided to drop this experiment altogether after sending an email to the acts on his roster. “There were a handful of artists who understood and liked the idea of it,” DeVine remembers, “but they just didn’t want to be attached to a statement. There would suddenly be articles like, ‘Explosions in the Sky Pulls From Spotify’ or ‘Pinback Pulls From Rdio'. It’s not their fight.” DeVine agrees with Yorke’s principled stand, but notes that many smaller artists fear that their music would stop speaking for them, in lieu of the politics of their label.
“I'm like a small farmer who interacts with people who consume what I make, and the Spotifys of the world, which are like McDonald’s, make people less aware of how the thing gets
made and of its value.” — Singer/songwriter David Bazan
The battle being fought, which many label owners and artists understand, is one of perception. When music becomes big data, the act of listening to a single song—something that was never trackable before—is capable of generating a small amount of monetary value. While Spotify’s promotional affordances for new artists have yet to be proven, the platform has redefined the possibilities of micro-monetizing back catalogs over a period of years. This angle on the Spotify Problem illuminates what is perhaps the biggest shift in the economics of popular music occasioned by streaming music royalty schemes: the shift away from the more established procedure of collecting (more significant) royalties for point-of-sale music recordings to a lengthy series of micropayments.
The point-of-sale royalties model, which drove music consumption in the 20th century, banked on radio play to drive purchases of albums, concert tickets, and other merchandise. This is a very pop and rock-centric model, to be sure. There’s a reason that rappers, whose income is structured through a different equation of live performance, brand tie-ins, merch sales, and recordings, aren’t complaining much about streaming micropayments. Under the streaming micro-royalties model, however, people pay fractions of a penny every time they stream any song, for as long as Spotify is around. Of course, the disappearance of Spotify is a very real possibility—as much as Daniel Ek publicly desires to create an experience for artists and fans, he’s under much more pressure to turn a profit.
For Swanson and his labels, the focus with Spotify is more on monetizing catalog than promoting new releases. He uses Drag City—a label that was a formative influence on the launch of Secretly Canadian and does not offer its music on Spotify—as an example: "Drag City’s back catalog is the sort of stuff that people are going to keep playing over and over,” Swanson claims. “I'm convinced that they would make more money on Smog records streaming over 10 years than they would by selling individual downloads.” In this way, one of the economic logics of Spotify for labels is the logic of the classic rock or oldies station, with exponentially more depth. For Swanson, the bottom line is twofold: return on investment, but also what he calls “cultural market share”—the continued circulation of music that has outlived its initial hype—for which Spotify works admirably.
For DIY artists, there’s no need to make public statements about financial stations. It’s more a matter of staying away. Singer/songwriter David Bazan, who is currently playing concerts in his avid fans' living rooms, says Spotify and its ilk offer his career very little. “The majority of my interaction with my audience is people who buy records and go to shows,” he says. “Outside of concerts, it’s just disposable media.” Bazan’s albums with Pedro the Lion, released by Jade Tree between 2000 and 2004, are available to stream on Spotify, but his two most recent solo albums, released by Barsuk, are not. As Bazan tells it, the process was rather simple: He asked Barsuk to withhold his solo albums from the platforms, and they agreed. A longtime proponent of sustainable, DIY music-making, Bazan uses locavore food movements to explain his distaste for Spotify. “I'm like a small farmer who interacts with people who consume what I make and tend my little patch of ground," he says, "and the Spotifys of the world, which are like McDonald’s, are going to make people less aware of how the thing gets made and of its value.”
Inherent with the “slow food” mentality as applied to music is the necessity of the tangible fetish property, the physical music object. “I’m willing to lose hundreds of impressions here or there in the publicity game in order to have people get my record and hold it in their hands,” says Bazan. “When it costs them something, they’re going to have a slightly deeper relationship with it.”
Vinylphilia and other sorts of affective relationships with musical objects are nothing new, but they take on a new level of importance when contrasted with a transmission from a distant server rather than an mp3 on a hard drive. It’s often hard to understand the affordances of music media until the next one comes along. Mp3 files felt incredibly intangible, even ephemeral, compared to CDs, cassettes, and records, but at least you could carry around an object that you were certain contained those files. With the new crop of streaming platforms, there’s even less a sense of ownership, only the procedure of remotely licensing a digital file to start playing each time you click. It’s understandable that many artists worry about the implications of streaming music: Access isn’t nearly the same phenomenon as ownership. Internet radio platforms like Pandora highlight the long history of this idea.
Next: Pandora: Internet Radio Is Cleaning Up the Nation
V: Pandora: Internet Radio Is Cleaning Up the Nation
Pandora more or less invented algorithm-driven internet radio in the mid-2000s, but Tim Westergren’s project didn’t start off as anything close to radio. After studying computer acoustics and audio recording at Stanford in the late 1980s, Westergren found himself adrift. He worked a variety of jobs in and around the music business before falling in with the Silicon Valley crowd in the late 90s. His experience as film score composer—a job that required him to break music down to its component parts, then and match those with narrative hooks—combined with advice from musicologists and coders, led to the Music Genome Project, an expansive database of songs given a scientific veneer. Song characteristics are called “genes,” which were collected into “chromosomes,” which make up larger datasets called “genomes.”
In order to try and bring the Genome to market, Westergren and his team, funded by $1.5 million in angel investor capital, launched Savage Beast, an e-commerce company, at the peak of the celestial jukebox gold rush in 2000. AOL Music tried Savage Beast as its recommendation engine, and Best Buy and Barnes & Noble tried it for in-store kiosks, but money was scarce in the post-dot-com bubble era. A venture capital influx of $9 million in 2004 allowed Westergren to pay back his long-suffering employees, hire an actual CEO, and by September 2005, launch the Pandora site as a standalone internet radio application. In late 2005, Pandora sold its first ad. In 2008, millions downloaded the iPhone app. By 2010, Pandora struck a deal with Ford. “Think about what made AM/FM radio so accessible,” CEO Joe Kennedy told The New York Times in 2010. “You get into the car or buy a clock for your nightstand and push a button and radio comes out,” he said. “That’s what we’re hoping to match.”
By mid-2013, Pandora had succeeded at competing with radio, with 76 million users and an 8.6% share of total U.S. radio listening. Indeed, the connection to radio is more than just a metaphor for Pandora. The platform is licensed through the DMCA as a pure play webcaster, which means limited song skips per hour and no rewinding. (In contrast, platforms like Spotify and Rdio are classified as “on-demand” services). The webcasting designation qualifies Pandora for a statutory license like broadcast radio stations, which means that it doesn’t have to strike deals with individual record labels to clear music, as long they pay the government-set royalty rates.
One crucial difference between Pandora and over-the-air radio stations, however, lies in who gets paid. Whereas broadcast radio stations pay royalties via ASCAP and BMI only to composers and publishers—a deal that goes back decades—Pandora is required to pay performers, who collect directly via the non-profit performance rights organization SoundExchange.
“Pandora actually is a very powerful engine for revenue generation in an area where performing artists didn't make any money until the mid 90s,” according to the FMC’s Casey Rae. Radio stations continue to resist paying performance royalties, having contended for decades that playing songs over the air is inherently promotional. (The U.S. is the only country in the world in which music performers are not directly compensated for broadcast radio play.) In contrast, Rae contends, “the statutory royalties that webcasters like Pandora pay can be looked at as a major step forward in terms of clear, transparent, equitable artist compensation.”
In May 2007, spurred in large part by Pandora’s quick success, the Copyright Royalty Board significantly raised the mandatory royalty rates for webcasters. Before the ruling, webcasters paid a fixed annual fee plus a percentage of their profits. After the ruling, they are required to pay per-song fees plus annual fees per channel. What defined Pandora technologically had come back to bite it economically. One of the primary advantages of internet radio is infinite choice, and many other stations much smaller than Pandora still had thousands of algorithmically-derived channels. The new fee structure made basic operations a mathematical impossibility for some of these smaller webcasters, for whom royalty payments exceeded total revenues.
To combat the rising royalty rates, Westergren quickly hired a PR firm and started encouraging Pandora subscribers, artists, and managers to contact their representatives in support. Bills were introduced—the Internet Radio Equality Act in 2007 and the Internet Radio Fairness Act in 2012—but neither have gotten very far. The fight transcended partisan lines, though in the public eye, it boiled down to a battle between those who supported tech developers’ rights to a market without crippling economic regulations versus the non-exploitative compensation of artists.
The solution is somewhere in between these poles: a sustainable internet radio model with enough scope and choice to satiate consumers but that can also earn enough revenue via subscriptions and advertising to pay the artists a fairer wage for the products of their labor. Of course, much of the battle is fought in the arena of public perception, where Westergren has faltered. Much like the very wealthy Daniel Ek, it doesn’t help Westergren’s case as a put-upon public servant that Pandora’s 2011 IPO resulted in a $3.2 billion market cap.
Internet radio breaks sharply from the broadcast model, building itself around the individual, not the mass, and scraping
activity data into a hyper-personalized experience.
Webcasters pay artists directly while broadcast radio doesn’t, but the numbers are still incredibly low. The explanation as to why royalty payments are so low lies in internet radio’s unique conception of a listening public. In a blog post from June 2013, a day or so after an editorial by Pink Floyd used some fuzzy math to justify an anti-Pandora stance, Westergren wrote a long post on Pandora’s blog explaining the company’s approach to royalties. He started by blaming the RIAA—which, to be fair, has a terrible reputation for inflating, if not imagining, profit losses—for the bogus numbers that Pink Floyd passed along. Then, he clearly conceptualized the difference between Pandora and broadcast radio.
“Each spin on Pandora reaches a single person, compared to a ‘play’ on FM radio that reaches potentially millions of people,” he explained. “In other words, a million spins on Pandora might be equivalent to a single play on a large FM station.” This difference, Westergren concluded, explains why Pandora payouts are so small. “If major market FM stations paid the same rates as Pandora, based on audience,” he continued, “some would be paying thousands of dollars for every song they played. How much do they pay performers right now? Zero.”
Much of the debate about Pandora—and to a degree, all internet radio—centers on such a conceptual dilemma. The dominant form of transmission during the 20th century was broadcast, in which a single transmission reached millions of people simultaneously. For radio historian Susan Douglas, broadcast radio added “a new cognitive dimension” to social life that could bind “utterly diverse and unknown people together as an audience,” forging “powerful connections between people’s inner, thinking selves and other selves.” What is often called a “monoculture” nowadays is described by Douglas as an “imagined community” of strangers, united not through shared space, but because they were hearing the same programming (and ads) at the same time.
Internet radio breaks sharply from this model, building itself around the individual, not the mass, and scraping activity data into a hyper-personalized experience. The cognitive effects of interacting with internet radio are profoundly different than the broadcast model. Pandora is programmed to “learn” individual tastes, and then deliver music that flatters those tastes—the Top 40 model applied to the Long Tail, perhaps. Westergren may not be the most sympathetic figure, but he’s right: Internet radio’s promotional value is significantly less at the level of a single “spin” than the broadcast model’s. Directing a song at a single person is valuable, but of a very different sort than directing a song at a million people at the same time.
The manner in which Pandora’s algorithm delivers individual songs to individual listeners is itself unique amongst its streaming-platform contemporaries. Pandora is much snobbier than the Echo Nest, using a team of musicologists to code each song based on hundreds of criteria, from purely musical factors such as tempos and tones to more subjective categories like emotion and busyness. Using the terminology developed by the Echo Nest, Pandora is the epitome of a recommendation service with tons of care, but little scale. The best streaming platform will rank highly in both the number of songs it offers and the ingenuity with which it sorts them (naturally, the Echo Nest ranks highly on both, while Whitman notes that Pandora has “only” one million songs coded over more than a decade, compared to the Echo Nest’s 35 million).
The extreme care of Pandora’s approach allows Westergren to make some fairly outlandish claims about the objectivity of the algorithm’s results—he's said that the derivation of musical taste is nothing more than “social jockeying” and snobbery, withering in comparison to Pandora’s genomic approach. But Westergren’s arrogance in eliminating the social construction of taste in favor of his Music Genome is countered by an anecdote from a 2009 New York Times profile by Rob Walker, in which a violinist who works as a Pandora song decoder dubbed an Indian raga a three-and-a-half out of five on the scale of “exoticism,” leading to a brief debate about the subjectivity of that term. While it’s questionable to think that a country comprising 1.2 billion people could in any way be “exotic” to someone from Earth, this coder’s evaluation underscores the fact that Pandora is far from “celestial” in its scope—it’s only available in the U.S., Canada, and New Zealand. Moreover, such subjective coding draws attention to the fact that recommendation algorithms, like tastes, always arise from intricate networks of incredibly advanced technologies and human beings.
Next: Human After All: Beats Music and Google Play
VI. Human After All: Beats Music and Google Play
This past January, the long-rumored result of a collaboration between Trent Reznor and Beats—as in Beats by Dre headphones, the most successful music accessory since the iPod—turned out to be a smartphone-based music streaming platform, launched in collaboration with AT&T. As part of its opening branding blitz, Beats promised to put flesh-and-blood music-experts—ahem, “curators”—ahead of the cold, robotic algorithms that guide the music selections of its competitors.
“Music is much more than digital files,” Reznor purrs in one ad. “It breathes, and bleeds, and feels,” and “integrates the best technology with friendly, trustworthy humanity.” Reznor’s claim is, of course, factual, but that’s beside the point when it comes to marketing lingo. What matters for Beats Music is that its users believe they’re taking part in a streaming experience driven by personalities rather than algorithms.
I use the word “experience” purposefully, because it comes up a lot in discussions of streaming media platforms. Over the past few decades, advertising has embraced the idea of not simply selling things to consumers, but turning all manner of pleasurable activities into sponsored “experiences.” The same holds for developers, who work in concert with the people tasked with selling products to a general public. When I asked Beats Music CEO Ian Rogers what made Beats Music’s deal with AT&T and its smartphone-first philosophy unique, the “e” word came up immediately.
“Your phone comes with a pair of headphones, but you can still build a billion dollar business off of people upgrading that experience,” Rogers bluntly (and correctly) admits about the market that Beats Audio cornered. “Your phone comes with music too, but the experience is pretty crappy.” As is the basic function of modern branding, crafting a user experience via a streaming platform is a way of engendering loyalty to a commodified idea through a series of pleasurable activities—all based around the individual, not the mass. It’s fitting that the most exciting new streaming platform since Spotify would emerge from a billion-dollar electronics brand based around isolating the music listener.
Though coverage of Beats Music’s launch focused on Reznor, Dr. Dre, and music biz legend Jimmy Iovine, it’s Rogers’ low-key presence and extensive experience with online music that shapes the platform’s experience. Whereas Daniel Ek is an ex-BitTorrent CEO and Tim Westergren is a Silicon Valley-connected musician and composer, Rogers started as a tech-savvy mega-fan, active for much of digital music’s online history. After the Beastie Boys took note of Rogers’ comprehensive fansite—in 1993, it was one of the first of its kind—and hired him as a consultant, he pushed them to the forefront of fan interactivity and technological integration for the next decade.
Then, after trying and failing to turn Yahoo! Music into a worthwhile platform, he took over Topspin, a service designed to help up-and-coming artists market themselves online. Rogers’ recent hire of ex-Gang of Four bassist and longtime brand strategist Dave Allen as Beats’ first artist advocate makes sense given Rogers background in artist marketing and platform development. Allen was one of the few artists to take a public “pro” stance on Spotify and its ilk, claiming in one well-circulated 2013 blog post that “intelligent men like David Byrne and Thom Yorke… do not appear to understand that we are in the midst of new markets being formed” through streaming platforms, not the death of the recording industry or creativity.
Beats Music promotes its experience as a human one, but like all streaming platforms—like all recorded music—it’s a combination of humans and technologies. Beats’ “Just For You” homepage is an oft-pruned reflection of activity data and preferences, and “The Sentence” feature reframes the internet radio model as magnetic refrigerator poetry. Through its AT&T-powered connection to flesh-and-blood humans, Beats echoes a short-lived fad of the 1940s, when Rock-Ola’s Mystic Music Jukebox and AMI’s Singing Towers inspired the 1944 film “Swing Hostess”, in which a down-on-her-luck singer takes a job filling requests as a “jukebox operator.” Seventy years later, Beats’ jukebox is driven not by distant music operators, but by veterans of radio label A&R and music journalism (former Pitchfork Editor-in-Chief Scott Plagenhoef heads programming and editorial). Its experience combines the expert paradigms of Pandora and Rhapsody with Songza’s activity-and-mood-based playlists and the Echo Nest’s scraping of the web’s music discourse. Under the Beats brand umbrella, Rogers’ is crafting the musical software to complement Beats’ upscale $300 headphone hardware.
Aside from “experience,” the key to understanding Beats Music lies in “curator,” a word that came up as often in my interview with Rogers. That term—beloved by the Echo Nest’s music coders as well—has reached a post Web 2.0 saturation point. It makes sense: At a time when digitized cultural objects, from high art to .gifs, circulate widely for anyone to sort through, recombine and display, a title formerly reserved for the organizers of art galleries and museum exhibitions now applies to the omnivorous activities native to the Stream. “Our playlists have a consistent quality and length across all of our curators,” Rogers told me, lumping the Beats staff in with the platform's “third-party” curators, including “Lone Star, Ellen DeGeneres, Pitchfork, and Decibel.” For Rogers, curation equates to a level of distinction and definite shape that internet radio and Spotify don’t offer, while framing the playlist as the platform’s default music commodity. With Beats Music, Rogers tells me, “you know what you're getting in terms of quality and length, and that makes it a consumable thing.”
“Online music services need bushwhackers carving paths from
one starting point to another. We’re not gatekeepers. We’re not tastemakers. We’re park rangers.” — Google Play Music's Tim Quirk
What Ian Rogers is for Beats Music, Tim Quirk is for Google Play Music. In the late 1980s, Quirk formed the band Too Much Joy, which was briefly signed to Warner as part of the major labels’ Buzz Bin-era land-grab of the early 1990s. In 2009, he inaugurated the trend of musicians-examining-their-digital-royalties, via a blog post titled “My Hilarious Warner Bros. Royalty Statement.” He’d been working at Rhapsody for several years at that point, and used his recording industry know-how to ask some pointed questions in a self-effacing, humble-bragging sort of way. His band’s quandary was that any digital (or physical) royalties they received were only chipping away at the $400,000 advance they still needed to recoup for Warner Bros. Quirk’s plight underscores the fact that artists signed to major labels often receive much worse streaming deals than those signed to indies—many of which offer 50/50 profit splits after advances are recouped—but also draws attention to Google Play’s plans to build-out its streaming and retail platform. Quirk, the outspoken public face of Google Play, is an ex-artist who knows from paltry digital royalties. But he also built the streaming stalwart Rhapsody, and knows how to market algorithmic recommendation as a flesh-and-blood operation.
Quirk and Rogers were co-presenters on a panel at South by Southwest’s 2014 conference titled “Man vs. Machine: The Curation Dilemma”, and while the title suggests a battle, both Rogers and Quirk know better—it’s the work of people using increasingly advanced technologies that has driven music recommendation for decades. Quirk also pitches Google Play Music as a quirky, flesh-and-bone team of music nerds. “I believe strongly that Google Play's music-discovery algorithms are way more geeky and subjective and weirdly human than others,” he tells me.
But his descriptions don’t stop there. At the 2013 Future of Music Summit—the Future of Music Coalition’s annual conference—he coined a metaphor for music recommendation in an age of infinite abundance that reframes digital curation in a unique way. “Telling the entire world what it should and shouldn’t listen to has become far less important than simply making this overgrown musical jungle navigable,” he claimed. “Online music services need bushwhackers carving paths from one starting point to another. We’re not gatekeepers. We’re not tastemakers. We’re park rangers.”
And if any single company is cut out to carve careful paths through the morass of digital music, it’s Google. Their skill at clearing through clutter is, in a way, what made its search engine so powerful; more than any other single entity, Google’s PageRank algorithm organized and sanitized the web. Like Apple with iTunes Radio, Google waited for the field to settle before making its big push, and Google Play Music aims to corner the streaming and digital retail markets by offering a digital music retail outlet as well as a paid subscription option that competes directly with Spotify (and pays better royalties to artists than its on-demand competitors). Like Spotify and iTunes, Google Play Music subscribers can incorporate their libraries into the platforms, but Google allows the uploading of one’s mp3 collection—illegally obtained or not, everyone has long given up caring—to their servers, or to have tracks matched by Google’s 20-million-song library.
At this point, the possible pirate origins of the files don’t matter as much as the fact that you’re listening to music while immersed in Google’s universe. Unlike Spotify, Beats Music, and Pandora, which are all single-entry portals offering access to large bodies of music, Google Play Music is built into a vast search, communications, and software ecosystem, co-opting the trust and reliance that hundreds of millions ascribe to its ever-expanding software, hardware, entertainment, and communication offerings. Where Beats and Apple are offering content to complement their lifestyle products, Google aims to re-route digital music into a retail-focused music experience contained within the infrastructure of the world’s largest public information utility.
Quirk ruffled a few feathers during that same FMC talk with his bold claim that “music is priceless.” He didn’t mean it literally, of course—the chutzpah required to take that stance while working for what is still the world’s leading directory for pirated music that is now selling music would be too much for anyone to bear—but as a way to address the continuities between music’s pre-digital era and the streaming present. “The same song will always be worth different things to different people at different times,” Quirk told the audience. “The online music revolution hasn’t changed that. It’s simply made the fact glaringly obvious.” As it happens, one of Google’s other properties has been making that fact obvious outside the gaze of the recording industry for years.
Next: Life Inside the Infinite Library: YouTube and SoundCloud
VII. Life Inside the Infinite Library: YouTube and SoundCloud
Jace Clayton doesn’t care to talk about how companies and artists are profiting from streaming music platforms. He’s much more interested in the cultural implications of a life lived inside what he calls music’s “infinite library.” As part of his research for a 2012 Fader article, Clayton met groups of Egyptian kids who were obsessed with 50 Cent after watching his videos on YouTube and Dailymotion. He imitates them for me: “We watch 50 Cent videos and we can't understand what he's saying, but we know what it's about because we understand the visuals.” The music they make from this influence is called Mahraganat, which means “festival.”
“It's basically an updated version of Shabbi, or populist Egyptian roughneck music,” Clayton explains. “They're incorporating structures from hip-hop, maybe a turnaround in the beat every four bars, Auto-Tune, and a certain sort of exuberance and attitude. It's a bit parochial, but then there's this weird overlay, which comes as a result of watching 50 Cent's ‘In Da Club’ over and over again, and getting access to what hip-hop culture is about.”
Clayton is particularly attuned to young people making use of technologies to reimagine styles from the other side of the world through local languages. He's perhaps more widely known as DJ/rupture, the globetrotting artist who treats his world travels as part-gig, part-field research expedition. Clayton’s DJ mixes, which intersperse R&B and hip-hop smashes with music from around the world, reflect the unpredictable paths and recombinations of music at a time when scarcity and a lack of access feel like relics of the past. In 2009, he wrote an article about “world music 2.0,” the term at once a knowing nod to the corporate-created music genre and an acknowledgement that decades later, the landscape is much different.
“World music 2.0 is a bottom-feeder, obsessed with microcultures breeding on YouTube, invite-only chat rooms, and other obscure corners of the internet where the line between producer and consumer blurs," he says, "and most songs end up being given away for free.” Instead of the center-periphery mindset of broadcasting—whether “center” stands for corporations or governments—Clayton describes a network in which “edges look to edges.”
For Clayton, focusing on the cultural implications of streaming platforms is far more interesting than tracking their economic impacts. “Once you get money out of the equation and you're not worried about labels and large media corporations and artists getting paid," he says, "the infinite musical library becomes a kind of cultural patrimony.” The differences between the metaphors of celestial jukebox and infinite music library are subtle but important: one is driven by micropayments, the other is a civic good, equally accessible to all.
Through his creative work as DJ/rupture, Clayton’s focus is on telling sonic stories by putting diverse cultures in dialogue with one another. In a very real way, DJs are the original human engines for music recommendation and recombination, the flesh-and-blood progenitors of the algorithmic features built into streaming platforms. “DJs begin because we're obsessed with music, because we're listening to it all the time, because we're excited about discovery, we're excited about connections,” Clayton tells me. “Then I think, 'How can I turn all these sounds around me into a narrative?'”
Clayton’s perspective subverts the too-frequent tendency amongst tech and music journalists to equate technological transformations of “music” to its effects on “the recording industry.” Yet while YouTube streams are powering the non-stop hybridization of exciting, off-the-mainstream-grid global musics, they’re also being translated into statistics for the most centralized, corporate purposes. In the wake of “Gangnam Style”, “Call Me Maybe”, and “Harlem Shake,” Billboard changed its popularity methodology fairly significantly. “The notion that a song has to sell in order to be a hit feels a little two or three years ago to me,” former Billboard editorial director Bill Werde told The New York Times last year. “The music business today—much to its credit—has started to learn that there are lots of different ways a song can be a hit, and lots of different ways that the business can benefit from it being a hit.”
YouTube has long been the default global site for music
discovery and sharing, and it’s not even a close race.
YouTube is easily the largest, most influential, and most difficult to peg of all the music streaming platforms; It's long been the default global site for music discovery and sharing, and it’s not even a close race. An early 2011 Nielsen report claimed that three times as much music was streamed on YouTube than was downloaded from online retailers. Partially because of its embeddedness within Google searches and partially because of its openness to accept uploads from any registered user, YouTube has become the default online locale for finding an old song you can’t remember the words to, an ever-expanding archive of global popular music, a source of live concert transmissions, the MTV of “viral” clips complete with its own awards show, a space to make playlists, and a resource for people to teach others how to produce or compose music. It’s a promotional tool for one-hit wonders to earn advertising revenue (just ask Psy), a "Star Search"-style platform for seeking fame (just ask Justin Bieber), and a public space for commentary (just stay out of the comment sections—unless you treat them as art). Via its Content ID algorithm, YouTube is a DMCA-regulated seeker, finder, and deleter of corporate copyright violations, incapable of considering "fair use." The Echo Nest powers the recommendation features for Vevo, the major labels’ corner of YouTube. YouTube is top-down, bottom-up, and all points in-between. Like the company that owns it, YouTube has become a verb.
Depending on one’s perspective, YouTube is an accessible, ever-growing archive that is reversing pop music’s orientation toward the future, or it’s an accessible, ever-growing resource for artists to explore and recombine new sounds from around the globe, ceaselessly spinning off new micro genres (and memes). YouTube was a central figure in Simon Reynolds' 2011 book Retromania, responsible for “the astronomic expansion of humanity’s resources of memory,” and thus largely to blame for the fact that “the presence of the past in our lives has increased immeasurably and insidiously.” While longing for the boredom of his youth and the scarcity of out-of-print records, Reynolds worries about the capacity of the open, virtual archive to turn creativity into simple mimicry, when YouTube renders otherwise foggy memories freely available for detailed study.
Whether these memories are based in the brain or on the web, they're prone to disappear without a trace. Sometimes it’s as simple as a user deleting an account, but increasingly it’s the more sinister version of corporate owners “disappearing” large swaths of cultural production to avoid having to deal with copyright claims. In the wake of Rupert Murdoch buying Myspace and “nuking” the imeem streaming service in 2009, ethnomusicologist and blogger Wayne Marshall, a longtime annotator of the microtrends popping up every second on any number of online streaming platforms, wrote an extensive blog post, spurred by the very real fear that “entire media ecosystems” might suddenly “succumb to the sudden slash and burn of corporate logic, which cares little for what we might celebrate as cultural vitality.” I’ve been using the word “platform” throughout this article as linguistic shorthand to describe a variety of streaming services, but as Marshall notes, the term can disguise as much as it describes. YouTube and other services use “platform” as strategic PR, Marshall contends, to cover up the much more precarious technological and political realities that underpin their use. Calling YouTube and other streaming services “platforms” creates the image of an elevated space on which one might communicate to a large audience, strategically eliding the fact that uploads can vaporize at any point, often without warning.
The term "cloud" acts as a form of linguistic wallpaper masking
the fact that digital copies of “your” music are in fact somewhere amidst vast server farms around the world, subject
to disappearance without notice.
Over the past few years, SoundCloud has filled the role that imeem and its ilk once did, but not without similar worries about cultural vanishing. Started in Berlin by two transplanted Swedes, SoundCloud launched in 2008 with a simple, user-friendly interface allowing artists (particularly DJs) to create pages and easily share their work by embedding streams in other sites. As the platform grew, its developers started incorporating algorithms, similar to YouTube’s Content ID, that could detect copyrighted work being re-appropriated and circulated. Marshall took note of the myriad DMCA-sanctioned automated takedown notices that artists were receiving from the platform, and filed it under the category of what he dubs “platform politricks,” citing the “chilling implications for everyday musical practice, global popular culture, ‘fair use,’ and the public domain.”
Clayton has mixed feelings about SoundCloud as well. “They courted DJs in the beginning when they could afford to be more lax about copyright enforcement, using that to hype and reach a larger mainstream crowd,” he recalls. “And then they say, ‘OK, we're going to start pulling a lot of those early mixes that we pretty much tacitly condoned.’ It's this awful bait-and-switch.”
The bigger problem with such practices, Clayton argues, extends to cloud services offered to everyday music fans: “What does it mean if your own personal music is being stored in a platform that's hard for you to access, hard for you to download, has source code you can't tweak, you can't port it to something else if it gets bought by Murdoch's tentacles?” The word “cloud” fills a similar function for SoundCloud, Apple, Amazon, and Microsoft as “platform” does for YouTube. It acts as a form of linguistic wallpaper masking the fact that digital copies of “your” music are in fact somewhere amidst vast server farms in places like Maiden, North Carolina; Ashburn, Virginia; Singapore; New South Wales, and Victoria, Australia; and Dublin, Ireland, subject to the terms of impossibly long end-user licensing agreements and to disappearance without notice.
One of the things that SoundCloud does particularly well, however, is allow user-created metadata to reshape the pathways of music discovery. As Quirk describes it, Google Play Music’s park rangers are well-versed in the skills of tagging and organizing. “We understand that metadata is merchandising,” he says. “It's not enough to just be in the system, you have to be constellated with other things that are like you.” But while Google Play and its contemporaries rely on top-down tagging, SoundCloud use bottom-up, user-created metadata as the communicative seeds of a vast music platform based around the idea that genres can be a series of endlessly recombinable word games.
On platforms like SoundCloud, user-created tags allow for genre constellations on the model of Twitter hashtags: a publicity mechanism for introducing strangers to one’s work, a navigation tool for listeners trawling the site for specific sounds, and a way to meet possible collaborators. Clayton refers to SoundCloud genre tagging as a “folksonomy,” a sort of DIY linguistic discovery interface that couldn’t be more distant from the musical connections served up by “a James Blake Pandora station,” in his dismissive comparison.
Though large portions of both platforms are commercially-supported, SoundCloud, YouTube, and their myriad online counterparts are much more vibrantly populated not by artists striving to gain fame, but simply to make music. As Brown University ethnomusicologist Kiri Miller explains in her book Playing Along, YouTube’s most prominent role for many is the proliferation of music lessons. She coins the phrase “amateur-to-amateur” as an update to “peer-to-peer” to describe many YouTube musicians who don’t claim to be experts, but who use the platform to find like-minded dilettantes. Similarly, University of Texas historian Karl Hagstrom Miller is writing a book that considers YouTube as the new “parlor piano,”a space for anonymous amateur musicians for whom playing popular music is, like using YouTube, a part of their everyday lives. Though they’re too often overlooked in lieu of digital music’s economic impact on the recording industry and artist royalties, these projects illuminate the myriad non-commercial forms of musical activity that populate the Stream.
Next: Streaming Pasts and Futures
VIII. Streaming Pasts and Futures
The chasm between YouTube amateurism and the high-level debates over Spotify and Pandora royalties makes clear that streaming is far from just one thing. It's everywhere: DIY musicians use Bandcamp’s malleable platform to allow curious browsers to sample before buying, and sites likePitchforkand NPR Music have built out streaming widgets and standalone pages for free, pre-release previews. The Hype Machine, once the Dow Jones of blogged mp3s, has morphed into a streaming platform. This Is My Jam, a single-track service created by an ex-Last.fm employee, reimagines the bedroom-wall style music fandom of peak-period Myspace for single-song doses.
While the major media corporations are using the Stream to crowdsource A&R (Lyor Cohen’s 300 label striking a partnership with Twitter, Warner collaborating with Shazam) and expand operations into new platforms (the Tribune Company buying music metadata giant Gracenote), fans and artists are morphing their everyday activities to account for streaming advances. I’ve seen vinyl DJs use Spotify and SoundCloud to supplement their crates with the new pop tracks clubgoers drunkenly scream for (and witnessed the awkwardness of a mid-song wi-fi hiccup), and watched apartment get-togethers evolve into dance parties collectively soundtracked by people playing YouTube clips off their phones. Anyone who pays a monthly fee for server space and learns a bit of coding can set up their own personalized streaming site; anyone with limitless money and cultural influence can strike a deal with a wireless provider to turn their new album into a smartphone app and data-mining portal.
“What we saw in the 20th century was an anomalous blip when music had a physical form. Music has this intrinsic pull
towards the dematerial, towards the unbuyable.
It's a slippery, ghostly thing.” — Jace Clayton
In the years preceding the turn of the 21st century, the musical objects that were introduced scarcely a century earlier underwent a radical micro-materialization into mp3s, upending rituals that had been so firmly established as to seem a natural part of musical life. Seemingly overnight, highly-regulated commodity networks were augmented from the bottom-up as mp3 files burrowed new conduits that proved impossible to regulate. “Piracy” became insufficient to describe digital musical transactions; they were malleable and quick enough to count as acts of everyday communication.
For Jace Clayton, as digital files give way to streams—dictated as much by the everyday activities of fans and artists as well as the top-down logic of the corporation—music is returning to its originary, pre-industrial form. “What we saw in the 20th century was an anomalous blip when music had a physical form,” Clayton surmises. “That was very unusual in the course of human history and it will soon be very unusual again. Music has this intrinsic pull towards the dematerial, towards the unbuyable. It's a slippery, ghostly thing.”
From the dawn of radio to the ascent of the celestial jukebox, guided by the market’s invisible hand or just coursing through the air, music has fluctuated between states of ghostliness and tangibility. At the same point that phonography was reifying music into collectible things in the early 20th century, radio broadcasts sent it through the electromagnetic spectrum. Early radio listeners—whose imaginations were colored by the mechanized devastation of the Great War and were subject to the strong pull of spiritualism—often thought they were hearing the voices of the dead. The industry built around recordings and technological advances in radio were future-oriented products of modernity, but Susan Douglas reminds us that broadcasting “carried people back into the realms of pre-literacy” through a mode of communication “reliant on storytelling, listening and group memory.”
Several decades later, with radio’s ethereal signals long the subject of countless musical odes and its commercial imperatives the subject of bitter musical attacks, music’s ephemeral nature and object status were symbolically reappropriated as the “celestial jukebox.” And now, a decade after the recording industry’s first efforts rotted on the vine, a slew of well-capitalized startups are rerouting the insurgent musical currents of the digital age, with inexhaustible resources powered by databases and algorithms, disguised by elaborate interfaces, piped through unseen cables that, diverted by wireless routers, push music through the air to the attractively designed objects capable of receiving radio signals now called “wi-fi.” Listeners aren’t mesmerized by the miracle of transmission anymore; used to being plied with access to everything, they’re more likely to carp if an artist’s work isn’t represented in streaming catalogs.
Though the move from properties to portals has the capacity to engender thrilling new forms of musical innovation and cross-cultural collaboration, it is also cause for new anxieties. Streaming makes music feel more invisible than it’s ever been, but it still circulates through pathways that are much more controllable than those of mp3 files. There’s no need for the RIAA to file gaudy suits against its own consumers anymore, when fans are willing to exchange infinite accessibility for mundane surveillance and invisible corporate control. Digital streams take shape via hardware and operating systems that are subject to capitalism’s planned obsolescence, and they’re subject to laws crafted by politicians, corporations, and lobbyists who don’t have creators’ interests in mind. For the FMC’s Casey Rae, the stakes of streaming are rooted in its sustainability. “The health of an ecosystem is best judged by how many independent operators it can sustain, and it would be a tremendous tragedy if we ended up in a winner-takes-all environment.” In late 2012, Spotify received $50 million in funding from noted vampire squid Goldman Sachs, which doesn’t bode well for the long-term survival of the dozens of playlists I’m currently following on the platform.
As I write this article, I’m obsessing over Beats Music, trying to learn every nook and cranny of the platform. I’m beaming its playlists from my phone to my stereo receiver through my wireless router—so far, “Best of Trackmasters,” “Peter Murphy: Deep Cuts” and “Intro to Ryuchi Sakamoto” are my favorites. My giddiness is nothing new. Over the past few years, I’ve similarly obsessed over Spotify and YouTube before moving on to the new thing.
For communications professor Patrick Burkart, the pleasures we derive from each new platform represent a central problem to modern consumerism: “We think they satisfy all of our consumption desires, but they actually get closer and closer and closer and never quite satisfy, which is why we can be convinced to go on to the next one with the same sort of promises.” Such obsessions come complete with a moral quandary unique to our current moment, in which technological fetishism directly impacts the sustainability of artistic creation. After a decade during which millions of music fans became used to unfettered mp3 access while quietly ignoring the economic consequences for artists, certain mental calculations come into play. In an email, a friend put it succinctly: “Spotify is the best-yet middle ground between putting an artist in financial difficulty and putting myself there.”
Digital music has merged seamlessly into the 21st century information stream, but streams have been an organizing musical metaphor for much longer—traceable back to Edward Bellamy’s mystifying “music rooms” of 1888. Or maybe the origin is 11 years earlier, two weeks after Alexander Graham Bell filed his patent for the telephone, when a New York Times writer predicted the decline of live music when listeners had the option of streaming operas at home through phone lines. Or perhaps the imaginative origin of streaming lies in Francis Bacon’s utopian novelNew Atlantis, published in 1627. Bacon described “sound-houses” found on the mythical island of Bensalem, where people “practise and demonstrate all sounds and their generation,” and which feature “all means to convey sounds in trunks and pipes, in strange lines and distances.” Music is the most ephemeral art, and the most artful form of human communication. People have been imagining new ways of ambiently circulating it for centuries. The only certainty for the future of streaming music, history tells us, is a future of streaming music.
Photo by Jimmy Giambrone
DeMarco and his girlfriend, Kiera McNally
Mica Levi; photo by Steven Legere
Photo via 
Photo by Henrik Källvik
Persson (center) playing with Harvester, photo by Jakob Sjöholm
Photo by Henrik Källvik
Photos courtesy of Laraaji
Gang Wizard: Rob Woodworth, Michael Landucci, Linus Landucci, Christopher Breedon
Lavender Country live in the 70s
Haggerty playing the Seattle Gay Pride Parade in 2000
Haggerty running for Washington State Senate with Nation of Islam running mates in 1988
Haggerty with his daughter Robin in 1981
The Haggerty family in 1945
Masha Alyokhina and Nadya Tolokonnikova at a recent Voice Project event in L.A. Photo by John Tyler Curtis.
I realized fairly quickly that flying south to claw around for Ralph Rinzler’s index cards and Mike Seeger’s bootleg tapes was a useless errand. I told a lot of people not to worry about it, and nearly all of them appeared relieved. I no longer knew why I was so preoccupied with gathering or in any other way quantifying Harry Smith’s 78s—what I thought they could tell me about music or art or humankind, how I thought they might augment or guide my own experience of collecting.