🌐 Spatial Web Foundation

Spatial Web: Concepts of Space from AI

Artificial intelligence is transforming humanity's understanding of space itself — expanding from the geographic and physical to hyperspace, tensors, and the high-dimensional worlds of machine cognition. The Spatial Web is the infrastructure that integrates them all.

📐 IEEE P2874 Working Group 🔬 Hyperspace Modeling Language 🤖 Agentic AI Integration

Space and time are humanity's primary tools for organizing information. From the earliest maps of the ancient world to GPS satellites and digital twins, spatial reasoning has been foundational to how we understand, communicate, and act on knowledge about the world.

But artificial intelligence — particularly the large language models, graph networks, and tensor-based systems that define modern AI — is introducing radically new concepts of space. These are not the Cartesian coordinates of classical geography, nor the three-dimensional physics of Newton. They are high-dimensional, relational, and abstract: tensor spaces in which concepts like "apple" and "orange" can be compared, semantic spaces where similarity is a kind of distance, and graph spaces where non-coordinate relationships define structure.

The Spatial Web, defined by IEEE 2874-2025, accommodates and unifies these expanded notions of space. Its concept of Hyperspace — entities connected by relationships — is deliberately general enough to encompass geographic space, physical 3D space, tensor/vector space, semantic space, and more. This generality is what makes the Spatial Web a foundational infrastructure for the age of agentic AI.

George Percivall, Vice Chair of the IEEE P2874 Working Group, has argued that the Spatial Web's most important innovation is precisely this generalization: "Space and time help us organize information. We've got the most general version of space and time in hyperspace."

Core Insight
AI doesn't just operate in space — it creates new kinds of space. Neural networks build high-dimensional vector spaces where meaning can be computed geometrically. The Spatial Web extends this insight: every entity in the world can be represented as a node in hyperspace, with relationships as the geometry that connects them.
Word-to-Vector → World-to-Vector
Large language models transform words into vectors, enabling conceptual comparison in high-dimensional space. The Spatial Web generalizes this principle to the entire physical and digital world: cyber-physical entities — people, buildings, sensors, agents, and objects — are represented as vectors in hyperspace, enabling integration, reasoning, and interoperability at unprecedented scale.
IEEE 2874-2025
The IEEE Spatial Web Standard defines Hyperspace Modeling Language (HSML) and Hyperspace Transaction Protocol (HSTP) — providing a Rosetta stone for integrating 3D worlds, AI vector spaces, business analytics, and geographic information systems within a single coherent framework.

Expanding Concepts

A History of Space — From Euclid to Hyperspace

Each era of science and technology has expanded the concept of space. AI represents the most recent — and most radical — expansion yet.

Classical — c. 300 BCE

Euclidean Space

The foundational geometry of points, lines, and planes. Three-dimensional, flat, and absolute — the space of everyday physical intuition. Remains essential for engineering, construction, and physical modeling.

3D geometry surveying architecture
Enlightenment — 17th c.

Cartesian Coordinate Space

Descartes united algebra and geometry, enabling precise numerical representation of location. This foundation underpins all of geographic information systems (GIS), GPS, and location-based services.

GIS GPS coordinates mapping
Modern Physics — 1905

Spacetime

Einstein's relativity unified space and time into a four-dimensional continuum, curved by mass and energy. Space is no longer absolute but relational — dependent on the observer's reference frame.

relativity GPS corrections cosmology
Machine Learning — 2010s

Vector / Tensor Space

Neural networks embed concepts, words, and objects as high-dimensional vectors. Semantic similarity becomes geometric proximity. Word2Vec, embeddings, and transformer attention all operate in this abstract space of meaning.

word embeddings LLMs tensors
Graph AI — 2010s–2020s

Relational / Graph Space

Graph neural networks bring non-coordinate relationships to the fore — connections between entities define structure without requiring Cartesian placement. Knowledge graphs, social networks, and ontologies all inhabit this relational space.

knowledge graphs ontologies GNNs
Spatial Web — 2025

Hyperspace

The Spatial Web's Hyperspace unifies all prior concepts of space. Entities — physical, digital, biological, computational — are nodes; relationships of any kind define the geometry. Geographic, semantic, temporal, and physical spaces are all instances of Hyperspace.

HSML IEEE 2874 world-to-vector unified AI
"AI and big data systems can efficiently handle terabytes and petabytes of information using tensors — very high-dimensional spaces. We generalize the notion of space to hyperspace. Space and time help us organize information. We've got the most general version of space and time in hyperspace."
George Percivall — Vice Chair, IEEE P2874 Spatial Web Working Group
Hyperspace Tensors IEEE 2874 Agentic AI World Models

Core Concept

From Word-to-Vector to World-to-Vector

Large language models transformed language into geometry. The Spatial Web extends this transformation to the entire world.

The Spatial Generalization of AI
Words as text
pre-2013
Word2Vec
semantic vectors
LLM Embeddings
context + meaning
World-to-Vectors
Spatial Web / IEEE 2874

Word2Vec and its successors demonstrated a profound insight: language could be understood geometrically. Words are points in high-dimensional space, and their relationships — synonymy, analogy, association — are vectors. "King minus Man plus Woman equals Queen" is a geometric fact about the structure of meaning.

Large language models extended this principle dramatically. Transformer architectures build attention-weighted representations in which every concept is locatable in a shared semantic space. This is not merely a metaphor — it is the operational reality of how modern AI processes meaning.

The Spatial Web extends this spatial metaphor beyond language to the full world. Just as Word2Vec maps words to vectors, the Spatial Web maps cyber-physical entities — sensors, buildings, agents, digital twins, geographic features — into hyperspace. This "World-to-Vectors" principle enables AI systems to reason about physical reality with the same geometric power they currently apply to language.

This generalization is not merely theoretical. The Hyperspace Modeling Language (HSML) provides the formal vocabulary for representing entities in hyperspace — defining their properties, relationships, and behaviors in a machine-readable form that any conforming AI agent can interpret.

The Hyperspace Transaction Protocol (HSTP) defines how agents exchange information about entities in hyperspace — enabling the real-time synchronization of world models across distributed AI systems.

Together, these specifications create the infrastructure for collective spatial intelligence: many AI agents, robots, and human users sharing a single, coherent, continuously updated model of the world — grounded in the generalized spatial framework of hyperspace.

Where traditional GIS handles geographic space and AI handles semantic space, the Spatial Web handles all spaces simultaneously — making it the natural integration layer for AI systems that must act in the physical world.

GeoRoundtable Application

Engineering for Expanded Concepts of Space

Understanding how AI transforms the concept of space has direct consequences for how we design and govern socio-technical systems.

🗺️

Geospatial Integration

Traditional GIS represents geographic space as coordinates, layers, and features. AI-driven systems require dynamic, relational, high-dimensional representations — and the Spatial Web provides the bridge between these paradigms.

🤖

Agentic World Models

AI agents need coherent world models to act purposefully. Hyperspace gives agents a common spatial framework — enabling coordination, shared situational awareness, and interoperability without requiring a central authority.

🔗

Multi-Domain Interoperability

When space means hyperspace, systems from radically different domains — enterprise analytics, autonomous vehicles, environmental sensors, social graphs — can all be positioned in a shared representational framework.

📐

Standards as Spatial Grammar

IEEE 2874 provides the formal grammar for hyperspace — the shared language that allows AI systems, robots, and people to describe their understanding of the world in mutually interpretable terms. This is the prerequisite for collective intelligence.

Related Work

Explore Further

GeoRoundtable work connecting concepts of space, AI, and the Spatial Web.

Work with GeoRoundtable on the Spatial Web

GeoRoundtable brings together expertise in geospatial standards, agentic AI, and philosophy of engineering to help organizations navigate the Spatial Web ecosystem.

Get in Touch