Circumlocution 🌀
On spiraling towards meaning
The following is a short piece I wrote about AI and meaning in March of 2023.
Describe your favorite song to me, only through words. Go ahead. Take as many words as it takes to tell me about it. I could suggest you explain its history, quoting the artist, maybe you even go into what genre it is and which genealogy of scene of musicians it belongs to, vividly describing its moment of time. Maybe you go on for days and days trying to describe it to me. Maybe you start to use increasingly more detailed similes and metaphors: you tell me “it smells like the first moment you step out the door.” And you keep telling me metaphors, increasingly more obscure yet relatable, and in doing so you circle in on the song, perhaps I start to get a clearer if still hazy idea of it. But however deep you dig, you cannot touch the song: the rhythm, harmonies, the cracks in the voice of the artist which provoke a certain sentimentality, it all remains behind an obstinate veil. One that’s so very opaque.
And because it’s opaque, this is what we see! We see beautiful metaphors, we see lore and history, and paths traveled to be told. I still don’t know what the song sounds like but I no longer see a veil but the new elaborate tapestry which hangs in front of it. This idea of spiraling through language towards something which can never be touched is what I call “circumlocution.” Literally it means speaking in circles, but it is also a condition of some unreachable real.
So how does one peek behind it? In the case of a song you could, well, listen to it. Is that sufficient? Put on your headphones and listen. Maybe invite some friends into the room, listen to it together. Then go outside and listen to it. Perhaps do so again in a crowd, at a concert. Perhaps watch someone else listen to it. Watch a thousand people dance to a snippet of it. Each time you trace and retrace, alone or together, its contours dissolve and reappear. Did they change or stay the same? Does it feel closer or further away? Listen to it a year from now, is it what it used to be?
But ultimately the music speaks for itself, on its own terms. Each time. The music speaks, and you listen and, if you’re open to it, it flows through you and deposits a fine sediment within you. Perhaps the sediment takes the shape of a melody you hum for days or hides itself as a vibration, a feeling that only hums when you play the song. Ever the mixologist, you decide to make a cocktail of the song and another experience, and together they flow through you and the sediments, as they channel through you, react like little molecules and leave a deposit of some third thing. Perhaps this thing returns to you as an image, vividly accessible only upon playing this song. It would almost seem as if the song became the tablet upon which this memory was etched, only to be gently stowed away within you and to be carefully retrieved each time the song plays.
What we’re talking about in this case is an encounter where the human allows the world to pour in. This makes the human contingent, as new residues are deposited and metabolized. Some are stored away and others are immediately purged. Over and over, the world comes inside as a mixture, highly contingent and at times volatile. A volatile mixture could be overstimulating and unpredictable, like an uncontrolled, unintended chemical reaction. Other times, it comes in and readily slides into little cookie cutter holes perfectly placed in a membrane, ready to receive. How the steady flow of flecks, phenomenology, blends inside of us is a crucial concern for understanding our encounter with Large Language Models. Of highest concern is the phenomenology mediated by words, which are extremely potent, sticky, and at times contagious, carriers of meaning.