top of page
World models
SPATIAL
For software to reason about a building, it needs an internal 'world model' — a learned picture of how the place is put together and how it behaves over time. We invest in this because a system that can simulate how a specific property will weather can plan its care in advance, rather than only reacting once a problem is visible.
A world model is an AI system's learned internal simulation of an environment — its structure, objects and dynamics — that allows it to predict outcomes and plan rather than merely react. In ATH, world models are the substrate for neuro-symbolic, spatially-grounded reasoning over the built environment (COSMOS provides exactly this kind of neurosymbolic grounding for compositional world models), enabling the engine to anticipate how a specific building will weather and colonise over time and to drive predictive stewardship.
spatial AI, path-dependent propagation
A world model is an AI system's learned internal simulation of an environment — its structure, objects and dynamics — that allows it to predict outcomes and plan rather than merely react. In ATH, world models are the substrate for neuro-symbolic, spatially-grounded reasoning over the built environment (COSMOS provides exactly this kind of neurosymbolic grounding for compositional world models), enabling the engine to anticipate how a specific building will weather and colonise over time and to drive predictive stewardship.
bottom of page