Music in the age of the algorithm
Get a shot of inspiration with the FT Weekend bulletin - the best in life, arts and culture. Delivered every Saturday morning.
Thanks to the Ngram Viewer, which searches 5m texts scanned by Google Books, it is possible to check the frequency of a word’s usage dating up to 2008. Thus we find that the slang term “sheeple”, meaning “sheep people” — a neologism mocking docile conformists — appears consistently but infrequently from the mid-1960s onwards before shooting up vertically in the 2000s. A further search of the 156m pages of US newspapers archived on Newspapers.com confirms that the spike continues into the present day. Decades after being coined, “sheeple” has finally caught on.
Why should this be? Two possibilities arise. Either we live in particularly conformist times, in which case there are greater numbers of sheeple to document. Or we have become more individualistic in outlook and thus more intolerant of conformity, in which case the luckless sheeple are being held up to a greater degree of ridicule than before.
In the past a useful, albeit deeply contentious way to find out would have been through music. “Nothing more clearly affirms one’s ‘class’, nothing more infallibly classifies, than tastes in music,” the French sociologist Pierre Bourdieu wrote in his 1979 book Distinction, which argued that perceptions of social difference are manufactured through lifestyle.
“Highbrow” music — the imagery comes from phrenology, the 19th-century pseudoscience that divined the level of person’s intelligence by the shape of their skull — is listened to by “posh” people while “lowbrow” music is listened to by “common” people. Appreciation masks snobbery. The more challenging a piece of music is understood to be, the more the listener claims cachet for choosing it. In 1984, when Bourdieu’s book first appeared in English, the distinction might be: I am interesting because I like The Smiths; you are not because you like Dire Straits.
Snap judgments like that, vital nourishment for New Musical Express readers of the period, are no longer so straightforward. Our acceleration since the 2000s into a world of almost limitless digital information — the kind that enabled me to look up the history of the word “sheeple” — has made almost all songs equally accessible. We now have centuries of music at our disposal, with more arriving at a dizzying rate: the streaming service Spotify’s 30m-strong catalogue has grown by 10m since 2013. The fantasy in the Argentinian writer Jorge Luis Borges’s 1941 short story “The Library of Babel”, of a library containing every conceivable book, is taking shape in music.
“When you can listen to almost any song in the world, how do you decide what to play and whether you like it?” Tom Vanderbilt asks in You May Also Like. The situation is liberating but also bewildering. Today’s listener is free to roam from The Smiths to Dire Straits and far beyond without fear of snobbish judgments. Notions of high and low culture have been challenged by new sociological theories of cultural omnivorousness. Distinction increasingly resides in the range of our listening, its eclecticism, not a narrow attachment to genre. Yet without a map, some sense of what we like and dislike, how do we make sense of the expanse of music stretching around us?
Vanderbilt, a US author and journalist whose previous book was about road traffic, examines the operation of taste in a number of fields, from food production to TripAdvisor star ratings. A recurrent theme is the mysteriousness of our tastes to ourselves and the corresponding ease with which they can be manipulated in order to sell us things.
In music, for instance, what we say we like and what we actually like might be quite different. It is a disparity that will be mercilessly exposed by the analytic tools that track our listening habits on streaming services or digital music players. I may think I am a sophisticate with a honed appreciation of John Coltrane’s sax tone, but the 29 times I have streamed Justin Bieber in the last week will tell a different story.
What is self-deceptive in our habits is transparent to the algorithms of big data, tirelessly interpreting the flow of biographical information we generate. “We’re capturing you as a location in taste space,” the chief data scientist of an online recommendation service tells Vanderbilt with sinister certainty. A “music intelligence” company he visits called The Echo Nest, now owned by Spotify, boasts of being able to apply a trillion data points from 35m songs and 2.5m artists to record listening habits.
The result of endless musical choice and a high level of sophistication at mapping our preferences has led to a curious situation. According to technological utopians, the digital marketplace should enhance choice, allowing a multitude of niche products to flourish in the place of a few mass-produced brands. That was the idea behind Chris Anderson’s influential 2006 book The Long Tail, which used music as a key indicator for the radically democratic age of consumption that supposedly lay ahead.
Hit songs, according to Anderson, were a function of old-fashioned constraints on supply and demand: shops would only stock CDs they knew they could sell. Digital music would end all that. Superstar dominance would be reduced and smaller acts would thrive in their place. Aggregate demand would replace blockbuster economics.
But the future has not panned out like that. Today’s A-List stars hoover up attention with the assiduity that their 1970s predecessors reserved for cocaine. In 2013, the top one per cent of artists accounted for over three-quarters of all revenue from recorded music sales. In that year 20 per cent of songs on Spotify had never been streamed (you can download an app devoted to playing them called Forgotify). The contrast with the gilded few at the other end of the scale is cruel. Last year UK singer-songwriter Ed Sheeran became Spotify’s most popular performer when he notched up his three billionth stream.
The fastest-growing niche in music does not support tech utopianism either. A revival in popularity of vinyl records represents an attempt to turn the clock back to the analogue era. Even giant UK supermarket chains such as Tesco, the epitome of “middlebrow” cultural tastes in phrenology-speak, have begun stocking them.
Vinyl’s booming sales point to an anxiety about our new age of plenty — that we are becoming bad listeners. Songs are streamed at low-definition bit-rates, “the worst quality in the history of broadcasting or any other form of distribution”, in the affronted words of audiophile rock veteran Neil Young, who yanked his songs from Spotify and Apple Music last year.
Attempts to lure the public with better sound quality, such as Young’s Pono music player, meet with indifference. Tidal, the streaming service launched by the rapper Jay Z with the promise of CD-quality sound, has 3m paying subscribers compared to Spotify’s 30m. “More songs” is the maxim of today, not “better sound”.
The problem is personified in the opening pages of Ben Ratliff’s Every Song Ever, when the author, a music critic for The New York Times, encounters a teenage boy on a bus in the Bronx listening, with a mix of “near boredom and absolute confidence”, to an R&B song on his mobile phone’s tinny speakers. To the analogue-era snob, this youth with his tintinnabulating phone is the epitome of today’s easily distracted music fan, a bitty consumer of low-bit songs. But Every Song Ever disagrees. “He’s the great listener of now,” Ratliff declares, waving aside the issue of sound quality.
He is right to do so. A good song may be enhanced by a good sound system but it does not automatically follow that it becomes a bad song when played on a bad sound system. In an age when thoughts about the content of music risk being drowned out by an obsessive focus on its distribution, Ratliff’s insistence on disregarding bugbears about degraded audio standards is refreshing.
The book welcomes the vast archive of music that has been opened up to us. But in the background lurks a shadowy desire to avoid the fate of Borges’s librarians, who are driven mad by the unreadably limitless store of information surrounding them. “Algorithms are listening to us,” Ratliff writes, as though in a paranoid whisper. “At the very least we should try to listen better than we are being listened to.”
His method involves jumping between different songs, tracing affinities. A discussion about the sense of space in music leads from Miley Cyrus’s “Party in the USA” to Pink Floyd’s The Dark Side of the Moon. Musings on loudness go from The Who to avant-garde Japanese rocker Keiji Haino. The digressions cross cultures and genres. They reflect an age when songs are replacing albums as music’s main unit of currency.
Albums were a product of technological restraints on how many tracks you could fit on to a record or CD. Today we live in the playlist age, able to mix and match songs from whichever source we want. Ratliff insists we should approach the task with greater rigour than the analytic tools observing us. We have a head start: the human brain is better at recognising melodies and genres than computers.
Every Song Ever is a useful call to action but flawed in execution. Ratliff’s determination to avoid thinking about music in terms of genre or social context leads to overburdened interpretations (“Is slow music in general a sound of dying?”). Flimsy proposals mount up, many exposed in the pages of John Powell’s pop-science guide, How Music Works.
Genre, according to Ratliff, “is a construct for the purpose of commerce, not pleasure, and ultimately for the purpose of listening to less”. But Powell explains that sorting music by type is automatic (“One of the first things we do when listening to a new piece of music is classify it”). The “truest kind of improvisation”, Ratliff asserts, is done “for its own sake, or one that doesn’t follow obvious patterns”. The statement looks jejune next to Powell’s interesting descriptions of the cultural differences in improvisation across western and Arabic music.
The relentlessly jocose Powell, a British physicist and classically trained musician, has deficiencies, too. His insistence on valuing music according to its physiological and psychological properties leads to a disregard for other artistic motives. Serialism, the compositional method invented by “shiny pated intellectual” Arnold Schoenberg, is dismissed as “bonkers” and “completely out of step with human psychology”. The idea that Schoenberg might have had deeper intellectual ambitions for his music goes unacknowledged.
Each of the above books raises fascinating questions about the opportunities and pitfalls of our glut of music. None supplies a firm answer.
One practical course of action might be to extend musical education at school, the quality of which in the UK was recently bemoaned by the London Symphony Orchestra’s new conductor Simon Rattle. Another is to recover notions of “good” and “bad” in music — to ask not whether a song succeeds in achieving its aims but whether it should have those aims in the first place.
Is catchiness good if what is being broadcast is not? Can we like a song while disliking its message? They are hard choices; but if we are to be more than “sheeple” then we must solve them for ourselves.
You May Also Like: Taste in an Age of Endless Choice, by Tom Vanderbilt, Simon & Schuster, RRP£12.99/Knopf, RRP$26.95, 320 pages
Every Song Ever: Twenty Ways to Listen to Music Now, by Ben Ratliff, Allen Lane, RRP£20/Farrar, Straus and Giroux, RRP$26, 272 pages
Why We Love Music: From Mozart to Metallica — The Emotional Power of Beautiful Sounds, by John Powell, John Murray, RRP£16.99, 304 pages
Ludovic Hunter-Tilney is the FT’s pop critic
Get alerts on Life & Arts when a new story is published