My family left Los Angeles because of the smog. There may have been more to it than that—a job offer for my dad, his longing for something new, a general feeling that we had endured more than our fair share of traffic. But in my memory it’s always only about the smog, the way it would make the sunsets blush, the way it would shroud the skyscrapers as if to provide the city a semblance of decency.
I was back in LA this summer, on assignment for a magazine but otherwise wasting time, getting reacquainted with the city I was born in. And on certain days, driving to work or the beach or wherever, I would have to admit that my father had a point—there’s nothing quite like the sight of a twelve-lane superhighway packed with cars up to and well past the edge of visibility, into the smog, miles of people sitting alone in their cars, a forlorn procession that inches up every minute or two.
At times like these I was certainly glad I had bought an AUX cord at a RadioShack going out of business sale. Traffic, the nonfunctional AC in my grandma’s old Element, drought, smog, apocalypse, whatever. On July 17th it stopped mattering—from that day on, I had Tame Impala’s Currents blasting from the speakers.
I won’t bother reviewing the album—if you haven’t listened to it, you should, and if you have, then you don’t need to be told how good it is. I bring up Currents only because it represents the first time in a while that I felt completely satisfied for any serious length of time with the music I was listening to.
That’s probably because I listen to a lot of music. There’s no standard measurement for this sort of thing, but I know that I probably spend more time listening to music than, like, other people. There are some upsides to this: the cultural cachet, the admiration of my friends and peers, the occasional compliment from my mom. But there are also some downsides, especially the short shelf life of even my most beloved tracks. To me, decent songs remain listenable for about four days after my first exposure, and great songs for maybe a couple of weeks.
With the proliferation of Spotify, Apple Music, and other streaming services, which allow nearly any song ever recorded to flow through a listener’s headphones with just one click, it seems reasonable to say that people now have more options for choosing the soundtracks to their lives than at any other time in history. So why do I always feel like I’m listening to the same ten songs on repeat?
No one has quite yet found a cure for boredom, but the tech geniuses and venture capitalists behind those above-mentioned streaming services are working to treat it, by ensuring a steady stream of new music is always available. Beginning roughly with Pandora’s unveiling in 2005, people in and around the music industry began to think that the real money was to be made in the precise and artful curation of songs, rather than in their distribution (and, it goes without saying, not in their production).
They were, generally speaking, wrong. Spotify, the Swedish company founded in 2006 and by this point undoubtedly the biggest streaming service out there, has still yet to turn a profit, in large part because the more successful they’ve become at recruiting subscribers, the more they’ve had to pay out in royalties. But financial insolvency aside, these companies have permanently altered the way we listen to music. According to Nielsen’s latest report, streaming is up, with 135.2 billion streams so far this year (as compared to 70.3 in the same period last year). Meanwhile, album sales declined another four percent, and downloads of individual tracks are down more than ten percent.
But look, there I go, making a discussion of how we listen to music sound like a market report. I can’t help but think of this as a sad byproduct of these streaming services, which have gotten so good at suggesting songs by knowing nothing about them. Take Spotify for instance. In the wake of Apple Music’s launch this June, Spotify rolled out a new feature called “Discover Weekly,” which every week gives each user a new playlist of songs “chosen just for you.” What Spotify doesn’t mention is that no one is sitting in some darkened room in Sweden, combing through your playlists like an NSA employee might comb through your phone records, hoping to know you—your likes and dislikes, your favorite shower music, etc.—so that they can predict the songs you’ll love. Instead, the Discover Weekly playlist is “chosen” by a proprietary algorithm owned by Echo Nest, a Spotify subsidiary.
The algorithm works through a process known as collaborative filtering, in which computers analyze your listening habits, find other users with similar habits, and then offer you some of the other person’s favorite tracks and artists with whom you’re unfamiliar, while doing the same for them. It gets more complicated than that, but the essential fact is that, to the computers doing the curating, all songs sound the same. That is to say, they don’t know the difference between the songs you love and those you loathe—they just know which ones you always listen to and which ones you always skip.
In some sense, music selection algorithms are nothing new. When jukeboxes were still a thing, those who owned them used simple algorithms based on play counts to figure out which records to keep in the box and which to change. This technique grew more sophisticated as the options for distributing music became more diverse—algorithms provided the basis for Top 40 stations, for Billboard rankings, for something as seemingly simple as the shuffle function. Until recently, that was about the extent of their reach: they could tell corporations which songs were being listened to the most, and radio stations could adjust their playlists so that no one had to go more than ten minutes without hearing a Taylor Swift song.
But these modern music suggestion algorithms are different. They’re not targeting the masses—they’re targeting you, willing to engage with your unique tastes and produce a playlist equally unique. You may feel flattered by this at first. Certainly, part of me is happy to wake up every Monday with a fresh playlist waiting for me, even if another part remains vaguely uncomfortable with the fact that something as personal and specific as my taste in music can be predicted, with surprising accuracy, by machines.
At this point, it’s quite possible that my computer has better taste in music than I do. Every week it presents me with a wonderfully diverse playlist, everything from unfamiliar artists to classic tracks from before I was born to deep cuts from bands I already like. These individualized playlists certainly seem to cater to everyone’s innate eclecticism, but paradoxically, they may in fact move us closer to a homogenization of music tastes. After all, since collaborative filtering works by suggesting to you what others like you already enjoy, each successful suggestion makes your own tastes a little less yours, and a little more the herd’s.
Other companies have tried different strategies. Pandora, for example, doubled down on data-based explanations as to why you like a certain song, by creating a semi-scientific “music genome.” And Apple Music uses so-called experts to make playlists for broad swaths of people, based on genre and some sort of activity, like “country fishing” or “indie picnic,” both of which seem designed to ensure only white people ever use Apple Music. Regardless, even if these strategies don’t have the same pitfalls as collaborative filtering, they’re not nearly as good at suggesting new music. Despite my apprehensions about the whole thing, I have to admit that Discovery Weekly routinely features one or two dozen good songs, and occasionally a great song (such as, in recent weeks, “Pains” by Silk Rhodes and “Camelblues” by Mndsgn.)
That’s the thing about these streaming services—they all offer more or less the same unlistenably large library of music, meaning they differentiate themselves mainly by how they guide users through the loads of shitty songs to get to the one or two good ones. And with a user base numbering in the tens of millions, it’s impossible to give any one individual reliable, personalized suggestions without resorting to algorithms, programs that predict your future behavior based on your past decisions, as if a human being, like a math equation, were merely something to be solved. Spotify’s algorithm may have me pretty much pegged, able to deliver me songs I deem worthy of adding to my virtual library with greater consistency than even I can achieve. But I can’t help feeling that the easier it is to find new songs, the easier it is to forget them.
Back when people still paid for music, they would have to make decisions about which songs were most worth buying. People write in reverent tones about the heyday of record stores, when loyal listeners would spend hours digging through the stacks to find the one album worthy of purchase, an album to which they could afterward devote themselves. I wasn’t alive back then, and don’t presume to know if music really sounded better when it was scarcer. But I do know that part of the joy of music comes from actually searching for it. Features like Discover Weekly may be good at finding new songs that I like, but no matter how good they get, they can’t mimic the devotion that comes from hours spent digging through forgettable crap until I find one song that makes me forget the meaning of the word forgettable.
These days we have an infinity of listening options. There’s the right song for every conceivable situation—the problem is that you probably haven’t found it, lost among a sea of equally plausible alternatives. Given this, it’s not unreasonable to let our computers do our scouring for us. After all, they already know how much I’ll enjoy a movie on Netflix before I’ve seen it, they know which ads I might conceivably find relevant, they can recognize my face in pictures—who’s to say they don’t know me better than I know myself?
But computers can only view music in terms of commodity, of data and averages and ratings, of the soundlessness of numbers. Sure, they may be able to figure out that because I like one indie rock band I’ll like another, that I listen to approximately the same stuff as every other college student, that human beings aren’t all that unique, and if they are their tastes are probably the least unique thing about them. But that all misses the point that we like the songs we do for strange and inexpressible reasons—a mental link with a specific time, or place, or person, an ability to makes us feel in ways we had forgotten—and sometimes for no reason whatsoever (“Call Me Maybe” anyone?). The problem with taking the algorithms that predict my purchasing behavior on Amazon and using them to predict my listening behavior on Spotify is that music speaks to us in ways that possessions don’t, in languages that can’t be coded. If we forget that, we risk misclassifying music as just another commodity. And just as the contents of our shopping carts are unlikely to deliver us any lasting satisfaction, so our music libraries become just as expansive—and just as empty.
Maybe that’s why I always feel like I’m listening to the same songs over and over again. Not because I get bored too easily, but because that boredom is too easily cured. With an endless queue of music and an increasingly sophisticated army of computers willing to guide me through it, why bother devoting myself to one song, or one band, when I’ll get suggested another one just as good in a minute? Or maybe these algorithms are just tools, and I’m just overreacting. Somehow, music always finds a way to stay the same, which is to say: it keeps changing, improving, getting newer and weirder and more human.
Not entirely unlike the algorithms popping up to shepherd me through it all. This summer convinced me that there’s no meditative space quite like the peaceful sanctity of your car, mired in traffic. Trapped in a car in a sea of cars, I couldn’t help think that maybe the music I was listening to wasn’t so different from the smog I was staring at. Nowadays, listening options stretch far beyond the hours in a day, beyond anyone’s attention span. And sometimes it’s nice to have some computer-operated assistance to guide us through the smog, GPS or whatever else, to know where we’ve been and to guess where we’re going. For what it’s worth, they’re pretty good at that. But sometimes it’s nice to shut all that off, to steer by instinct, to get lost in the sprawl of the smog until you can’t see where you started from.
The night before Currents came out, I went to see a decent DJ perform at a famous LA music venue, but left at precisely 11:45 so that I could get home by midnight and listen, front to back, to the new album. Something about the whole evening struck me as strange. There I was, missing out on what was surely a pretty good conclusion to the show I had bought tickets to, on the physical experience of attending a concert (the bass resonating in your stomach, the smell and stick of sweaty bodies, the cleansing warmth of cheap beer) so that I could go home, sit alone in a hammock in my darkened back yard, slip on my headphones and slip away from the world, for a little while at least. And somehow, it was the right decision. Before there was recorded music, there was just music, free and ephemeral. Now, by some odd fluke of circularity, we’ve come back to the start. With Spotify and other streaming services, it can’t even be said to physically exist on our phones. It’s in the cloud now, or in the smog, as I sometimes think of it—maybe it’ll be safe there.