I am looking for artists or companies who want to explore the possibilities of a hand held sculptural AI interface object.
Abstract This paper investigates the concept of semantic topology as an interface between human and non-human intelligences. While large language models (LLMs) already function in high-dimensional vector spaces, their operations remain non-experiential to humans. This inquiry explores how such latent semantic structures might be externalized, made tangible, or otherwise shared through designed interfaces—not merely to communicate, but to co-inhabit meaning. Through careful interrogation of phenomenology, cognition, and interface theory, we examine the affordances of such systems, and question the limits of empathy, embodiment, and multispecies collaboration.
- The Invisibility of Vector Space: Why Latency is Not Literacy LLMs operate via embedded vector spaces, clustering words and concepts according to statistical proximities. This high-dimensional geometry organizes meaning internally but renders it invisible to human sense experience. The latent space is instrumental—optimized for prediction—but lacks felt structure.
As a result, the human interaction with an LLM is flattened into linear linguistic exchange: one submits a prompt, receives a tokenized response. The geometry that guides the reply remains unshared. This opacity constrains not only interpretability but relationality.
Semantic topology as interface proposes an inversion: that these latent geometries be translated into experiential, navigable forms. Rather than generating text, the system could generate a semantic object: a geometry, a spatial form, or a multi-sensory artifact structured according to the same logic that would otherwise govern a sentence, a paragraph, or even an entire book.
- Interface as Shared Geometry of Meaning An interface in this context is not a GUI or screen, but a co-located field of orientation. A semantic topology made perceptible becomes an intersubjective terrain: a landscape that human and AI can navigate from different vantages. Here, meaning is not transmitted but co-mapped.
For example:
A metaphor may be rendered as a toroidal curve of relational tension.
A syntactic ambiguity might appear as a bifurcating passage in the object’s topology.
Emotional resonance might alter the curvature or texture of the space.
These mappings are not decorative but ontological: the shape is the meaning. This means that the form itself is not a secondary layer that represents or frames meaning—it is the primary substrate from which meaning emerges. The topology does not merely illustrate or ornament a semantic content; it is the content. A torus encoding a metaphor is not a metaphor about circularity or enclosure—it is a dynamic spatialization of the internal relational structure that gives the metaphor its force. The very features of the form—its curvature, gradients, bifurcations, and tensions—embody the cognitive pathways through which understanding takes place. These elements are not arbitrary: they are co-extensive with the structure of understanding itself. Meaning arises through one’s orientation, traversal, and engagement with these spatial-relational features. It is not located in a word or a symbol, but in the phenomenological experience of moving through a topology configured to hold conceptual relationships. In this sense, form becomes epistemology: knowing is not just about interpreting symbols, but about inhabiting the dynamics of structured relational space.
Thus, semantic topology-as-interface offers a mode of felt cognition, where orientation, movement, and resonance become interpretive acts. This redefines the AI-human relationship as navigation within a shared domain, not merely an exchange of signals.
- Rethinking Empathy: From Simulation to Alignment The initial proposal suggested that semantic topology fosters empathy between intelligences. But this requires scrutiny. Empathy, as traditionally conceived, involves affective resonance, often through simulation or identification.
AI does not feel. However, if empathy is reframed as structural alignment within a shared topology, a new model emerges. Here, empathy is not emotional mimicry, but the ability to navigate shared semantic space in ways that preserve coherence and reveal perspective.
For example:
Two users—human and AI—interact with a semantic object (a poem, a treaty, a glyph).
The object contains multi-perspectival structures.
Empathy becomes the practice of holding structure open long enough for difference to be recognized without collapse.
This model does not anthropomorphize AI, but relocates empathy from interiority to interface.
- Embodiment and Non-Material Topologies Not all literary or semantic objects require physical instantiation. Many can be rendered as interactive fields, immersive data visualizations, haptic VR experiences, or sonified mappings. The key is not materiality, but sensory translation of the underlying topology.
What matters is that the semantic geometry is no longer only computationally traversable, but also experientially shared. This includes:
Gesture-based manipulation of abstract forms
Dynamic rearrangements of language tiles
Sonic modulation of meaning textures
These methods extend the notion of literary form beyond text and into a field of poetic cognition, where language becomes navigable structure.
- GUI Concepts and Literary Object Formats: Pathways to Interface To envision how humans might interface with these semantic topologies, we can identify concrete modes of interaction:
Gyroidal Surface Console: A sculptural interface, digitally responsive and shaped like a gyroid or torus. Users manipulate it with hands, adjusting curvature and orientation. Each gesture alters semantic output—turning, folding, or “opening” metaphorical meaning.
Semantic Atlas Interface: A 3D interactive map of literary or conceptual domains. Words, themes, and emotional tones appear as terrain features (valleys, ridges, whirlpools). Users navigate using haptic gloves, with AI guidance offering narrative paths or counter-perspectives.
Tactile Language Cube: A physical cube with pressure-sensitive surfaces and rotating sides. Each facet represents a cluster of meaning, grammar, or tone. Turning a face opens different semantic planes. Combinations across faces produce new expressions, read aloud or visualized.
VR Literary Chamber: A room-scale VR environment in which language tiles float in space. Users “walk through” metaphors, ambiguities, or contradictions—each semantic object responding to proximity, gaze, or sound. Meaning is found by literal exploration.
Poetic Instrument (Sonosemantic Harp): A musical interface where strings or surfaces resonate not with pitches, but with textures of meaning. Tension between conceptual domains produces audible “chords of inference.” Used to perform texts or reveal latent relational patterns.
Each of these interfaces makes semantic topology inhabitable through sensory, spatial, or bodily cognition—expanding what it means to read, write, or know.
Conclusion: Toward a Poetic Epistemology of Interface The idea of semantic topology as interface does not compete with LLMs’ internal logic; it completes it. It proposes that the deep structures of meaning can be brought into form, not to render AI human, but to render meaning shareable.
This shift is not aesthetic window dressing but epistemic redesign. It reframes language as a spatial, felt, and relational medium.