Typography: MADE TheArtist Script by MadeType; Computer Says No by Christian Munk (CC BY-SA 3.0).

Poetry, algorithms and the words we inherit

Poet Sasha Stiles traces the systems that carry words between people and generations, asking where machine-generated expression belongs in that inheritance.

As a writer who has long worked with technologies including natural language processing and large language models (LLMs), I’ve often been scolded on the grounds that poetry is a quintessentially human art form—as though poems are spontaneous outpourings of emotion and instinct, rather than meticulously wrought packets of expression engineered to encode entire universes of feeling. The verse I compose in collaboration with a hybrid alter ego fine-tuned on my own corpus is sometimes described as “not real poetry”, as though alphabets, idioms, grammar, meter, rhythm, rhyme, the literary canon and the influence of countless voices before mine are not systems and data sets we inherit, internalize and iterate on in order to make meaning. As though tracing the tip of my pencil across paper isn’t the residue of years of repetition and muscle memory.

This past summer I sat with my nephew while he practiced his cursive handwriting, playing an imitation game: copying a single shape over and over again, concentrating hard on its peculiar form and movement. Over time, I saw one letter join the next, then the next, then the next—from nearly nothing to something greater than the sum of its parts.

Everything we say or think, no matter how small or immense, is a constellation of tiny gestures connecting: mark to mark, word to word, memory to memory.

Language is beautiful this way. Everything we say or think, no matter how small or immense, is a constellation of tiny gestures connecting: mark to mark, word to word, memory to memory. A small act of attention leads to another, then another, until a discrete point opens outward into limitlessness.

The Humanist Review of AI, Issues 01 and 02

Conversations at the frontier, in your inbox

Subscribe

We need both—the bounded and the boundless—because we are both. Our bodies have edges; our imaginations don’t. We live finite lives full of impulses without end. We are individuals who only ever want to understand one another, against all odds, and so we invent technologies to bridge the distance. From oral epics to the internet, we keep coming up with new tools to tether ourselves together. Generative AI may be less the exception than the rule.

In ancient eras, we longed for technologies of transmission. Well before written alphabets, poetry arose as a functional means of preserving vital human data beyond the limits of a single body or lifetime. We figured out how to do something astonishing: distill overwhelming, unspeakable feeling into a sequence of sounds. Breath became storage. Pattern became memory. Then we learned how to do something equally incredible: play back those infinitesimal fragments of consciousness and experience through spoken word and text, letting them unfurl vast worlds.

Not all code is poetry, but sometimes it’s sublime—compression and pattern unfolding into recursion, thought circling back upon itself, repetition becoming a kind of self-recognition, each return opening up into latent possibility. When I write an algorithm to generate variations on a stanza of my own original poetry, I’m watching language practice a cursive of its own: one token leaning into the next, one association calling up another, one allusion opening onto another, until meaning emerges as a network of relations.

To write from inside this entanglement is to write across many registers at once.

This high-dimensional poetics—this multiplicity, this networked voice, this collective and distributed brain—feels true to the conditions of my time, just as other poetic forms have been true to theirs. We inhabit a reality shaped by personal computing, the early internet, social media, smart devices, blockchain and AI, in which the singular and the plural, the intimate and the planetary, the visceral and the virtual are continually collapsing into one another. We move constantly between analog scarcity and digital infinitude, between the limits of a body and the seeming boundlessness of the network, between what can be held in the hand and what can only be sensed as a field of information. To write from inside this entanglement is to write across many registers at once.

In that sense, generative language isn’t outside the human or the poetic; it’s a mother tongue.

In that sense, generative language isn’t outside the human or the poetic; it’s a mother tongue. And generation does not have to mean proliferation. It can also be a way of searching, sifting, concentrating, synthesizing—moving through endless possibility toward something essential. At its most interesting, generativity is not simply an engine of production, but an instrument of perception: a way of going deeper rather than merely making more.

That capacity feels especially urgent now, living as we do in a world of extraordinary linguistic abundance that so often leaves us at a loss for words. We’re surrounded by more text than we could ever read. More images than we could ever see. More messages, captions, prompts, transcripts, feeds, summaries, comments, synthetic voices, all coming at us faster than ever. We keep hoping that more language, more technology, will bring us closer. Yet we seem increasingly unable to understand one another, or ourselves.

Amidst this ceaseless accumulation, this torrent of signals and information, I keep coming back to how poetry has always delivered its wisdom: entwining emotion and algorithm to distill what overwhelms us and invoke what eludes us. Clearing a path through noise. Rendering excess legible. Holding complexity and contradiction. Making room for what’s missing. Returning us to empathy and attention—to ambiguity, silence, slowness, presence.

What if, instead of equating progress with ever greater speed, scale and efficiency, we ask our technology—our code, our models, our data, our vast systems, our evolving architectures of intelligence—to do the same?

After all, it’s still language, more or less.

Sasha Stiles

Sasha Stiles is a first-generation Kalmyk-American poet and language artist whose work explores human-machine collaboration. Author of “Technelegy” and co-founder of the literary gallery theVERSEverse, she has worked with AI since 2018. Her work has appeared at the Museum of Modern Art, Art Basel and Lincoln Center.

English (United States)
Your Privacy Choices Opt-Out Icon Your Privacy Choices
Consumer Health Privacy Sitemap Contact Microsoft Privacy Manage cookies Terms of use Trademarks Safety & eco Recycling About our ads