Is a Song Just a “Data Object?” (Hint: No)
An OpEd by Ian Temple of Soundfly about the way we interact with songs as code-based entries in a data library, and what a “song” really could be?
Is a song just a data object trying to accrue data points? (Hint: No).
By Ian Temple of Soundfly Weekly
The most depressing definition I’ve ever heard for a song is a “data object.”
Perhaps unsurprisingly, it came from ChatGPT. I was asking it about the history of song forms, and we went deep. We started in prehistory, with those ritualistic flute jams you can find painted on cave walls. Who knows if these were songs or not, or what they sounded like. What they probably were is functional for the community in some way, whether supporting social bonding, ceremony, memory, education, or some other purpose.
From there, ChatGPT taught me about the Hurrian Hymns, the first examples in history of music notation we know about. These are a collection of lyrics and musical instructions carved into clay tablets in cuneiform, dated to around 1400 BCE, and found in modern day Syria. (I’d like to point out that’s the 2nd time in 3 weeks I’ve mentioned cuneiform in an essay. What do I get?)
Amazingly, some archeologists think they’ve figured out how to play some of this music, which describes specific intervals, a tuning system, and what strings to play on a 9-string lyre. Here’s a performance of the first song we know about, the oldest one we have a record of, as interpreted by archeomusicologist Prof. Richard Dumbrill:
From there, we sped through Sappho’s poetry and the Hebrew Psalms, Gregorian chants and the “love songs” of the Troubadours. ChatGPT pointed me toward a great Ted Gioia lecture about the many roles songs have played historically, beyond being a pleasant confection for the ears. He touches on how Aristotle considered music one of the four pillars of education and the work of anthropologist Frances Densmore documenting the different roles songs took on in Native American communities in the early 20th Century, from healing to storytelling to games.
We kept going from there into the early commercialization of songs in the 19th and 20th Centuries, first through sheet music and then through recordings, as songs became something owned, bought, and consumed.
And that’s when we arrived at the modern era and “data objects.” Here’s the exact quote:
“Today, the concept of ‘song’ is even more fluid: In streaming culture, a ‘song’ is a data object — a track, a streamable unit, detached from albums or even artists.”
A few thoughts. First of all, whatever — it’s a random ChatGPT quote. No point in reading too much into it.
But second, you know how AI is just a pattern-detecting machine that sucks up large swathes of the Internet and other human-made content and can piece it all back together in a way that offers a simulacrum of meaning? Yeah, well, sometimes that means that what ChatGPT spits out can be a fairly astute mirror reflecting back what we’re collectively doing here.
Factually, physically-speaking, ChatGPT is mostly correct. Most of our engagement with songs in today’s world is as code-based entries into the vast data warehouse of modern music consumption. And let’s face it: We do often treat songs as pieces of data. We rank songs by their listener counts, obsess over the numbers for our own releases, and listen to them on large, undifferentiated playlists with all artistic intention obscured.
The danger for us as living, breathing, meaning-seeking humans is: What if ChatGPT is also metaphysically correct? That’s what we should resist.
Today, I’ve got a new EP out with my group Sontag Shogun called In Cloudy States, and it is a triumph of artistic indecision.
Basically, I wrote a simple melody alongside some pretty chord voicings on the piano and sent it via voice memo to my bandmates. From there, the process kind of fell apart, as we played around with different versions and contributions and couldn’t decide on a final direction. What we ended up with is a four-track EP which is basically just the same song played in four different ways:
- Track 1: The main “song,” a melancholy piano idea swallowed up by tones, vocals, and washy effects.
- Track 2: A different version of that song but without the piano idea in it at all. Basically a reaction to the song without the song in it.
- Track 3: A third version of the song that’s only piano, but played differently, more loosely, an improvisation on the theme.
- Track 4: A fourth version of the song that begins with the original voice memo recorded to tape and ends with a foley-inspired beat and thumping synth bass.
This feels like a stupid idea.
In general, I’m a fan of decisiveness in all things, from ordering dinner to deciding on a strategic direction for your company. Making a decision, whether it’s right or wrong, pushes you down a specific path from which you can collect information and be better set up to make future decisions. It protects against endless tweaking and bickering. Being decisive is like blowing away whatever musty cloud of smoke is obscuring your vision.
I was reading this wonderful profile of the experimental jazz band SML in The New York Times yesterday. In it, the guitarist Josh Johnson describes how they wanted to steer clear of the studio because of how hard it is to make decisions as a group in an environment where anything is possible. That resonated. Whereas SML dealt with this decision paralysis by using the spontaneous compositional power of live recordings as the raw materials for their records, we dealt with it by simply refusing to decide on a single version of our song at all.
From a marketing perspective, it also doesn’t make any sense to release the same song four times — at the same time. I apologize in advance to our listeners.
But you know what? Now that we’ve done it, I’m actually pleased with it.
Because here’s the thing: I think we’ve gotten a little confused collectively about what a song is. Or at least a “song” has become synonymous with the recording of it. And in that, there’s a risk of forgetting that the fact of a song being a “data object” is one of the least relevant things about it.
Throughout history, songs have been so many things. They helped cement communities and created a sense of belonging. They served as acts of worship toward the divine and inexpressible. They hyped up scared young men for battle and intimidated enemies. They helped people find and express meaning in a multitude of ways.
A song is a source of connection and a social lifeline. It can be an outpouring of grief or a ceremonial undertaking, connecting us with lives past and future (as expressed memorably in that time-traveling scene in the film Sinners). It can be an emotional ignition, inspiring deep feeling and desire. It can be a vessel for understanding, or for falling down the indescribably deep wells of each moment that spread out like fractals when you focus in on it.
When I think of a song like that, it actually seems a shame that so often what’s taken for a song is the singular production printed and uploaded into the great data warehouse in the cloud — instead of the living, breathing, evolving chain of personal and community history it actually is. The data object is merely one snapshot of that song in action.
If that sounds like a post-release justification for our indecision, you may be right.
But what’s definitively also true is that the process of working through all the places this song could go was extremely playful and fun. It was an interesting challenge to pass a single pour of molten metal through multiple different molds and treatments. We allowed ourselves to try new things and stretch aspects of our arranging and production techniques in the process. At the very least, it was a valuable learning exercise. It even inspired me to create a coloring book based on the musical idea (which you can check out here).
I wrote the central piano idea for In Cloudy States thinking about how impossible it is to capture a single moment or experience precisely. In particular, with my young kids, that can feel tragic, like you want to grieve each moment with them you’ll never get back as it passes. But what you realize is that the echoes of each moment change as you play them back through your head over time, applying different filters and contexts that give them new meaning.
Maybe that’s what a song is at its core — a beautiful expression of a moment to be passed around, through different filters and contexts and interpretations. Maybe a song is a developing story, expressed in sound over time. Or at the very least, maybe there are creative possibilities that can open up for you in thinking of a song that way instead of trying to produce the best data object to pick up the most data points.
At least, that’s what we found. Whether others agree is out of our hands.
This week, if you’re looking for a fun musical exercise, I challenge you to take a single idea and explore multiple treatments of it. Or better yet, take our idea and explore more treatments of it. We’ve made the stems available for you (with some loose guidelines about how to use them). Go nuts with it.
You may even end up with a shiny new data object all your own in the process.
Ian Temple is the Founder and CEO of Soundfly. Follow his Substack, Soundfly Weekly, or join the growing community of musicians and educators on Soundfly for free today.
