The fond memories maker in my brain entered into overdrive many weeks ago when I chose to visit Pandora Radio after a three-year hiatus. While navigating the rush-hour press of tired bodies on the subway with buds in my ears, like almost every other commuter that night I came across a podcast segment detailing how Pandora’s brand-new CEO strategies to revive the struggling music service. I wondered about the methods which the music service I once depended on for my day-to-day individual soundscape had changed. Going to Pandora, I discovered, amidst its virtual dust-caked shelves wrapped in a modern interface, a record of my twentysomething sensibilities and existential stress and anxieties.
At the really bottom of this chronologically ordered rack were “Frou Frou Radio” and “The Way I Am Radio,” stations that I had created in 2009, languishing under 36 other stations I had produced and curated in between 2009 and 2015. A Pandora function called Thumbprint Radio billed as “a uniquely individual station influenced by all of your thumbs up,” felt like a reunion with an acquaintance I had not seen in years; all her recollections of me were true, however of a previous self. I winced and hit thumbs-down when Thumbprint began playing “Cape Cod Kwassa Kwassa” by Vampire Weekend and listened wistfully when “Such Terrific Heights” by the Postal Service and “Swansea” by Joanna Newsom came on.
After half an hour of cycling through varieties of fond memories, it occurred to me that unless I provided Pandora’s algorithm with new data points about my current musical fascinations and interests, it might never know to make the leap to Noname, or Natalia LaFourcade, or Kendrick Lamar, or Jóhann Jóhannsson’s initial soundtrack for the film Arrival. The algorithm wasn’t built to divine how the tastes of a lapsed user may have changed in the intervening years.
For much of my post-college twenties, Pandora was the music provider of choice in the two-bedroom home that I shared with my housemate in San Francisco. We seeded a myriad of stations with artists and songs we liked, and assiduously fed the algorithm with thumbs-up and thumbs-down feedback. Pandora’s secret sauce is asserted on the hypothesis that by decomposing a song into its musical fundamentals and gathering favourable and unfavourable feedback, its algorithm might identify particular combinations of attributes that you may discover irresistible, and subsequently recommend similar tracks to you. Here’s an example of the type of reasoning Pandora might expose if you asked why it suggested a specific Aimee Mann track, circa Magnolia: “Based upon what you have actually told us so far, we’re playing this track because it includes mellow rock instrumentation, excellent lyrics, a subtle usage of vocal consistency, groove-based structure and acoustic rhythm piano.” When I first heard about Pandora, before algorithmically driven suggestions like Netflix’s Cinematch became commonplace, I liked the idea that something as inscrutable as the ontology of taste might be demystified, broken down into its necessary components, and used to predict future sources of satisfaction.
However, eventually, the police verso esque judgment that Pandora asked of its users brought to the fore philosophical questions that my housemate to whom I owed my education in pre-Coltrane jazz and Brazilian pop and I would occasionally dispute over dinner. If I gave a particular John Mayer tune a thumbs-down, was I challenged it to him and his specific rendering of melody and consistency for eternity? Could I genuinely like Alanis permanently? Did Pandora’s algorithm assume that we corresponded and immutable agents, invulnerable to the day-to-day peaks and dips of state of mind or impulse or the development of individual taste? In addition to “thumbs-up” and “thumbs-down” buttons, Pandora’s designers had integrated a “avoid” button into the app’s user interface, as well as an embedded menu alternative that permitted users to signal to the algorithm that we were “tired of the track” (and hence to put that manifestation of John Mayer on the rack for a little while). These options only made the moments before possibly thumbing down a laden affair: I couldn’t merely act by the gut; I needed to question my impulses and equate my precise sensations toward a song into the most suitable virtual gesture.
Sometimes, an artist or a song that my housemate and I liked would creep into what we considered a musically incompatible context. One evening as we were making supper, Louis Armstrong, whom we loved and generally heard on the Dinah Washington station we had developed, suddenly sounded his look on my housemate’s Luisa Maita station. My housemate abruptly swung around with a knife in hand and tears in her eyes from chopping onions walked throughout the kitchen area to the laptop computer on the table, and mashed the thumbs-down button. Satchmo had shown up with his gleaming trumpet in the incorrect music hall, led astray by the algorithm’s associative reasoning.
In August 2010, the Onion released a parodical homage to the music service entitled “Desperate Pandora Employees Rushing to Find Song Location Male Likes.” The piece conjures a fictitious 32-year-old Boston customer called Dave Lipton who skips or thumbs-downs every track, flummoxing the business’s musicologists; in the Onion’s absurdist legend, Pandora’s personnel then go to great lengths to source obscure or genre-bending music to calm Lipton, just to be fulfilled by more futility. For me and my housemate, our constant source of comic frustration was the opposite experience: Each of our Pandora stations, when left running ignored for several hours, almost always seemed to degenerate into a single genre lounge music. Lounge music is, to me, the lowest common denominator, so dull and inoffensive regarding being an affront; it is the indistinct grey that emerges from blending too many colours on the paint scheme. Were an exceptional artificial intelligence to take over the world, definitely lounge music would be the Huxleyan drug that the makers would release to numb minds and quell rebellions.
I eventually wearied of tending to Pandora’s algorithm; my stations became stale and greatly convergent on a cluster of overplayed favourites. By the time I moved out of the two-bedroom in late 2015, I had stopped utilizing Pandora entirely and proceeded to Spotify.
Today, Spotify’s algorithmic suggestions are incorporated into the service’s interface in such an understated style that if you didn’t look hard enough at the suggested music at the end of your manually curated playlist, or if you weren’t sufficiently curious about checking out the service’s Discover Weekly feature, you could set about your week, even your entire year of music-listening without understanding them. However, what I find exhilarating about Spotify’s instantly generated weekly mixtape is its tolerance for riskier, wider-ranging suggestions; the misses are worth the statistical gamble for the hits. While Pandora’s algorithm relied on preferences broadly mentioned and slowly fine-tuned through experimentation, Spotify’s recommendations are based upon probably more accurate and more specific signals, especially tunes that you’ve listened to through their on-demand service when, how much of each, how frequently.
You may also love to read:
It’s undoubtedly unfair to contrast the memory of an old service with a recent experience of a newer one. The thing that I will always enjoy about Pandora is how groundbreaking it was, as one of the very first internet services to bring algorithmically driven classification and suggestion into public life. When Pandora was established back in 2000 and made it through the dot-com bust, algorithms and artificial intelligence weren’t yet buzzwords, and AI wasn’t as sophisticated, nontransparent, or poised for increasing application in critical worlds like medication and warfare as it is today– to the point that a new discipline called XAI, or explainable AI, has emerged in an attempt to make available and comprehensible the complexity of what goes on in contemporary AI’s proverbial black box.
In such a way, Pandora was a new design for explainable AI. It informed you precisely why it chose to serve up that specific Childish Gambino track or that Hilary Hahn recording of Sibelius’s violin concerto. Someday, we may be sentimental for a time when algorithms were more straightforward and more intelligible when they appeared more like fresh, sometimes quixotic or cheesy attempts at understanding and translating art and taste, and less a matter of life or death.