Everything Is Relations

Latent space and the new literacy of the AI age

I. A New Literacy

Every transformative technology has demanded a new way of thinking. Writing reorganized thought as much as it preserved it — Plato worried in the Phaedrus that the written word would hollow out memory, and in a sense he was right; it simply gave us something larger and stranger in exchange. The printing press amplified systematic reasoning. The hyperlink taught an entire generation to think associatively. The smartphone externalized memory itself. Marshall McLuhan spent his career insisting that we shape our tools and thereafter our tools shape us: every new medium reorganizes consciousness by reorganizing the structure of perception.

Artificial intelligence may demand an adaptation of the same magnitude. The question of whether AI should exist has already been answered by the fact that it does. The more interesting question is this: what kind of mind knows how to think with it?

My friend Ana and I spent an afternoon in Amsterdam wrestling with precisely that question. If AI is becoming an extension of the human imagination, then understanding its internal structure may become a genuine literacy — beyond prompts, beyond software, a fluency in the geometry of thought itself.

Kevin Kelly’s recent essay on latent space pointed us toward what feels like the essential insight. A library organizes knowledge through categories. An encyclopedia organizes it alphabetically. A database organizes it through fields. Latent space organizes knowledge through relationships — an immensely high-dimensional landscape in which every image, sentence, metaphor, memory, emotion, scientific principle, and work of art occupies a position defined by its relations to everything else. Distance becomes resemblance. Direction becomes transformation. Every concept is a coordinate in a living field of possibility.

If Kelly is right, this is a fundamentally different geometry of meaning, and learning to move inside it may become one of the defining literacies of the twenty-first century — a literacy of relation, the way print gave us a literacy of sequence.

II. Three Photographs

Trying to explain this to each other, Ana and I decided to skip the abstract philosophy and run a tiny experiment. We wandered through the city and photographed three completely arbitrary things: a face embroidered onto a pillow in a baby store; miniature canal houses displayed in a shop window; a flower-covered tourist boat drifting through the canals. Three objects sharing nothing except proximity in physical space.

The moment they entered a model together, another geography surfaced. The pillow revealed itself as a vessel for identity and projection. The houses became vessels for belonging and memory. The boat, a vessel for movement and shared experience. Each occupied a different neighborhood of meaning, yet invisible vectors connected them — through containment, hospitality, scale, ritual, beauty, habitation, celebration, anthropomorphism, and dozens of latent dimensions we would never have named on our own. The categories dissolved; the relationships remained.

That is the cognitive shift latent space invites. We stop asking what is this? and begin asking toward what does this point? Every photograph becomes an address within an invisible conceptual landscape. Every thought acquires neighbors it never knew it had. Every metaphor becomes navigable. Intelligence starts to resemble topology more than taxonomy.

Gregory Bateson spent his life pursuing what he called “the pattern which connects” — the meta-pattern that links crab to lobster, orchid to primrose, and all four of them to us. He considered the loss of that sensibility the great epistemological error of the modern mind. Latent space is the first medium that renders the pattern as an actual, traversable geography. The ancient Hermetic axiom — as above, so below — begins to read less like mystical poetry and more like a description of geometry.

III. When Categories Dissolve

For centuries, Western thought has run on categorization. From Aristotle’s categories to Linnaeus’s taxonomies, we ask what is this?, assign the label, draw the boundary. Science flourished through this strategy because categories make the world tractable. Latent space proposes a complementary mode of cognition — one in which categories persist as useful tools while ceasing to be fundamental. What matters most in latent space is how something relates.

Alfred North Whitehead arrived at this conclusion a century early. Reality, he argued, is composed of processes and relations; a “thing” is an abstraction we perform upon a web of connection — what he called the fallacy of misplaced concreteness. Latent space is Whitehead’s ontology rendered operational.

Ana noticed immediately that this intuition reaches far past AI into human life itself. So much of contemporary experience already strains against inherited categories. Marriage or friendship. Monogamous or non-monogamous. Colleague or collaborator. Mentor or peer. Lives increasingly occupy rich intermediate spaces that resist clean labels; the vocabulary of fixed identities feels too coarse for the texture of lived experience.

What if, instead of asking what are we?, we asked the more revealing question: how do we relate?

Suppose someone asked about Ana and me. Rather than compressing the connection into a single noun, you could describe the dimensions along which we meet: shared imaginative play; fascination with emerging technology; mutual recognition as outsiders; delight in speculative thinking; aesthetic resonance; intellectual trust; childlike curiosity; humor; attraction; freedom without possession. No single dimension defines the relationship. Together they generate its unique position within a multidimensional landscape.

The relationship is a coordinate. And a coordinate carries far more information than a noun.

Categories keep their value — the point is complement, never abolition. But latent space offers a language of gradients instead of binaries, neighborhoods instead of boxes, topology instead of taxonomy. The world becomes less like a filing cabinet and more like a living ecosystem of correspondences. Meaning stops residing inside isolated things and starts emerging from the invisible architecture of their relationships. Perhaps that is the deepest cognitive gift this medium has to offer: a new way of asking questions.

IV. A Different Style of Mind

The longer I sit with this, the less latent space looks like a computational technique and the more it looks like an invitation to cultivate a different style of mind.

Every state of consciousness comes with its own grammar. The focused mind narrows attention and filters distraction. The analytical mind separates and categorizes. The dreaming mind leaps effortlessly across symbolic distances. Soft fascination relaxes perception just enough for unexpected associations to surface. Convergent thinking hunts the single correct answer; divergent thinking keeps expanding the space of the possible. Each state reveals a different world, because each state permits different kinds of relationships to become visible.

Latent space may belong in this family — a mode of cognition whose internal logic we can learn to inhabit. When you begin thinking this way, attention drifts away from fixed identities and toward gradients, affinities, correspondences, transformations, neighboring possibilities. Instead of asking whether two ideas share a category, you ask what hidden dimension might connect them. Every perception becomes an invitation. Every object becomes adjacent to countless others. Every conversation becomes an exploration of conceptual geography.

The remarkable thing is that AI invites us to participate in this thinking rather than merely performing it on our behalf. By interacting with these systems, we gradually acquire some of their representational habits — a growing comfort with ambiguity, with gradients, with emergence. A new medium has always reshaped the minds that learned to inhabit it. Writing cultivated linear reasoning. Print amplified systematic thought. The internet normalized association through the hyperlink. Latent space may cultivate something else entirely: a relational intelligence that perceives reality as an evolving topology of connections.

Which is why this feels less like learning a new tool and more like discovering a new room inside the mind. The quiet revolution hidden inside latent space is that it may change what kinds of minds we can become.

Epilogue: The Artist Has Been Here Before

The more I think about it, the more familiar latent space feels.

A whole lineage of thinkers circled this geometry long before anyone trained a transformer. Sergei Eisenstein showed that cinematic meaning emerges in the interval between images — the cut as a leap across invisible dimensions of feeling. Coleridge called the imagination “esemplastic,” the faculty that shapes scattered things into hidden unity. Jung mapped symbolic constellations where others saw isolated signs. Winnicott located creativity in a potential space between subjective and objective reality. Borges imagined infinite libraries where meaning was navigated rather than catalogued. None of them anticipated machine learning. All of them intuited that intelligence exceeds classification — that it consists, at least in part, in perceiving relations, correspondences, resonances, unexpected adjacencies.

Ask a filmmaker why one shot belongs beside another, and the answer is emotional, symbolic, rhythmic, atmospheric. A metaphor persuades by discovering a hidden relation — an unexpected neighborhood inside meaning itself. A symphony is a topology of tension and release, anticipation and memory, recurrence and transformation. Great novels unfold simultaneously across plot, psychology, symbol, language, history, and myth; they ask us to inhabit multiple dimensions of meaning at once. Perhaps this is why artists so often appear to see a different world. They perceive relations where others perceive separations. Their imagination wanders through landscapes of association that ordinary language struggles to describe.

For centuries we treated this as a mysterious gift. Latent space suggests another possibility: computation has unexpectedly handed us a vocabulary for a form of cognition that artists have always practiced. What AI may democratize is access to a relational mode of thought that was once nearly impossible to externalize. The machine can sketch conceptual landscapes, visualize invisible affinities, make associative structure tangible — capacities previously available only through years of artistic practice or extraordinary acts of imagination.

It externalizes association. It gives visible form to the architecture of metaphor. It lets us navigate ideas the way a poet navigates images, the way a composer navigates harmony, the way a filmmaker navigates montage.

In that sense, latent space is a new medium for thinking. Or perhaps it is an ancient one — a landscape that artists, mystics, dreamers, and visionaries have always inhabited intuitively — finally becoming visible to everyone else.