World Labs Establishes World Model Taxonomy: Spatial Intelligence Route Begins to Systematize
World Labs releases a functional classification framework for world models, organizing the hierarchy and capability dimensions of spatial intelligence.
World models move from vague slogans to a discussable functional framework, providing a common language for robotics, AR, autonomous driving, and physical world agents.
Physical agents require, beyond language capabilities, persistent spatial representation, causal prediction, actionable simulation, and error correction mechanisms after execution.
Spatial intelligence may form a new platform layer independent of pure language scaling, driving investment in data, simulation, sensors, and robotics ecosystems.
Investment leaders should distinguish between generating 3D content and world models that support decision-making/action; the latter has higher technical barriers and industrial value.
Observe standard datasets, actionable world models, simulation-to-reality transfer, and complete hardware task demonstrations.