May 17, 2019

When the Owls Cry in the Night

Recently I’ve admitted to myself that I’ve basically given up on discovering new music. It no longer circulates as social currency so much among the people I know, and I find I don’t have the time to absorb anything unfamiliar. It’s easier to dismiss what I happen to hear as reworkings of stuff I already know. I never thought this waning of curiosity would happen to me, but now that is another lost illusion from youth. Confronted with the infinite playlists at my fingertips, the endless shuffle, I feel ready to surrender, either by letting music wash over me indifferently or by switching to podcasts.

I’ve reached a point in life where I find I have to invent rationales to listen to anything, a strategy for choosing specific songs among the thousands and thousands in my library. Going running is one of them — I have an iPod Mini for that purpose, and last week, when it was finally nice enough to run (I am on a fair-weather fitness program), I broke it out and found myself immersed again in the person I was last fall, and the musical preoccupations I had been cultivating then.

When I am in the swing of running regularly, I spend about as much time adding and subtracting songs from the iPod as I do exercising. These swaps usually reflect the various listening projects I have going on at any given time; last summer these included listening to Aretha Franklin (she had died in August and I was systematically listening to her entire catalog), the Fall (I had downloaded the Complete Peel Sessions and was working my way through them), and various Britpop acts (I skipped the genre almost entirely when it was happening and was trying to educate myself — I’m still occasionally capable of summoning some historical curiosity).

Listening to the iPod on my run, most of the songs made me think mainly of whatever protocol I had devised that led to them ending up on the iPod. In a sense, I was listening more to my own thought processes and obsessions than the music. To me, it was a collection of my intentions more than the musicians’. And when a song came on that wasn’t associated with any project I could remember — in this case, “Four Sticks” by Led Zeppelin — I immediately began to conceive a project that I could build around it. Maybe it was the endorphins, but the song seemed so much better than I remembered it to be. Maybe it was actually the best Led Zeppelin song, or at least the most overlooked. I started to think about how much I must be misremembering the rest of the Led Zeppelin catalog, which I have been absorbing since before I can remember. (I have no memories of first hearing songs like “Stairway to Heaven” and “Whole Lotta Love”; they just have always been in my history, like the Beatles and a few other bands.)

Still running, I started to daydream about what I needed to do: listen to every Led Zeppelin song, in chronological order, both forward and backward, and then make a comprehensive list ranking them, along the lines of those lists that occasionally appear on magazine websites. Typically I think those lists are trash when I run across them; they always remind me of Scharpling and Wurster’s “Rock, Rot, and Rule,” which is about a unqualified rock critic promoting a book of his arbitrary opinions as “the ultimate argument settler.” Of course, the point of these lists is to start arguments, to outrage, and I usually end up taking the bait and clicking through.

Such lists seem like a response to abundance: There is so much music that it makes less and less sense to evaluate songs or even albums in isolation. Instead, entire careers should be swallowed whole and assessed as such. It reflects the way people can get into a particular artist now, not one album at a time but by downloading the discography in a single torrent.

A lot has been written about the effects of abundance on music consumption, particularly if one was raised in conditions of relative scarcity. Often abundance is equated with a kind of aesthetic cheapening, as if having an infinite number of songs to stream makes them infinitesimally significant to the listener. I’ve probably waxed nostalgic somewhere about having to save my pennies as a kid to buy some album from a cut-out bin — let’s say, Tormato, by Yes — and then listening to it dozens of times to force myself to like it. You know, the good old days.

But it seems one of the main effects of abundance is to shift the frame of reference to larger units, from songs and albums to artists and genres. When I was writing about the algorithmic death-metal generator last week, I was thinking about this too: The next level beyond having access to an abundance of existing songs is a means of producing infinite amounts of new music according to one’s tastes or whims. It’s more interesting to fixate on how you navigate the ocean by your own compass than to pay careful attention to the water. Music is a raw material for articulating your own trajectory. Why care about particular songs in and of themselves at that point?

Along the same lines, it’s possible to become more interested in the capabilities of a model to produce interesting variations than in any particular instantiation in a particular composition. Any given “song” can be seen as just one note in a larger unfolding composition of what a particular set of algorithms is capable of creating at a certain point of their training. In a sense this would point back to how music must have been experienced before the advent of recording, with every performance constituting its own unique thing, regardless of whether it was produced from a score. But rather than be taken with the ephemeral singularity of a performance (these can be instigated on demand with a generative AI), we can consume the model’s learning process as the composition, and not the particular sounds it makes at any given time. It could be like listening to birdsong that becomes progressively more complex and interpretable. Its emerging capabilities are more interesting as a trajectory, much in the same way our own emerging tastes can be to ourselves.

Forget repetition. (Sorry Mark E. Smith.) Every song could be seen as single-use, completely disposable. You could produce music on the spot for whatever occasion you need it for, and why should it have to be something you already know well? When I was a teenager, I remember being picked up one afternoon by the girlfriend of one of my friends to go to the mall, and in her car stereo, she had a tape that was just same song recorded over and over again to fill the side. I remember this mainly because the song struck me then as an extremely preposterous one to choose for this treatment: “Burning Flame” by a band called Vitamin Z. But it also sort of scared me; it seemed almost pathological. I struggled to understand it. For me the whole point of listening to music was to acquire more and more knowledge of it that I could show off, and I couldn’t grasp that people might associate specific songs with specific feelings, and certainly not with that sort of monomaniacal intensity. If musical scarcity has taught us to fetishize that kind of familiarity and repetition, maybe the future of generative aesthetic abundance promises a different kind of enthrallment. We could become unstuck on things.

I still tend to treat songs as information, as data. That seems to harder to resist when there are so many of them available to sort through, when every time I listen to anything it produces more data about my listening profile, adds to the scoreboard of play counts. I increasingly find myself leaning into that sort of “prosumption” — I need all my listening to feel productive on my own terms. With the Zeppelin list I was imagining on my run, I fantasized about coming up with a spreadsheet assessing each song on a series of qualities (riffs, lyrics, vocal performance, guitar solo, rhythm section, etc.) according to a one-to-five rubric, as if I needed to come up with a methodology that could render my own opinions defensible to myself. The prospect of this scorekeeping made the future listening seem more purposeful and exciting; I was going to revivify this overfamiliar music to myself with metrics.

In this interview with Holly Herndon, whose new album is made from music generated by machine-learning algorithms, she laments how many AI music projects attempt to re-create already existing styles: “It gets us in kind of a feedback loop culturally which does not move us forward,” she says. “It doesn’t respond to what’s happening now and music should be responsive to the politic and the material world around it.” She posits the possibility — probably an inevitable one at this point — that algorithms would be used to create new material by dead artists:

If we can extract the kind of logic and aesthetic of a dead person and then reanimate them in future versions of themselves, they weren’t able to opt into that. We could write an entirely new catalog of Tupac music that he could perform in this voice model and we could maybe give it really fucked up lyrics that Tupac would have hated.

For better or worse, this is a bit like a distillation of what I decided to do with Led Zeppelin. While on the surface it would seem like I’d be really concentrating on specific songs and their specific qualities, I would also be trying to extract something from their music about their aesthetic in general that could be programmable as data. I can imagine an AI producing new material according to my rubric; I could adjust some sliders and produce new songs that are more of the same but different. I want to be stuck not on the same song but on the same band, only in a way that was about the idea of Zeppelin and not restricted to their actual legacy. Actually it would be a new way to be stuck on myself. I would turn my understanding of their history into an open-ended process. This feels like a way of taming the threat of abundance, not by rejecting it exactly but by converting it into a set of rules. It is a way to navigate the infinite without sailing over the edge.