ORCID iD
Document Type
Conference Proceeding
Subject Area(s)
Artificial Intelligence
Abstract
World models are being built twice, from opposite ends, without a shared theory of how the two halves should meet. One lineage grounds the world model in perception: a self-supervised, latent-predictive encoder – exemplified by Joint Embedding Predictive Architectures (JEPA) – that learns the structure of sensory experi-ence. A second, older lineage grounds the world model in cognition: an explicit, inspectable structure of entities, rules, and constraints, ranging from knowledge graphs to formal logic to physical law. Neither lineage alone has produced a world model that is simultane-ously adaptive and auditable. We argue this is not solved by picking a side, but by theorizing the seam between them. Borrowing the System 1 / System 2 distinction from cognitive science, we cast the perceptual and symbolic world models as two distinct groundings of the same underlying world, connected by a pair of bridge opera-tors we call uplift (the promotion of a statistical regularity into a labeled, discrete entity) and downlift (the relaxation of a discrete constraint into a continuous bias on a probabilistic manifold). We further argue that the “symbolic” side of this seam is not one thing but at least three – knowledge graphs, formal logic, and physical law – each with its own grounding semantics and bridge mechanics. Most existing neurosymbolic world models implement only one register and one direction of the bridge. We propose that future world models will increasingly be defined not by choosing between neural and symbolic representations, but by how effectively they maintain consistency between perceptual experience and cognitive knowledge.
Publication Info
Postprint version. Published in ACM AI LEADERSHIP SUMMIT: Defining AI's Next Chapter, 2026.
APA Citation
Sheth, A., Thareja, M., Pawar, A., Rawal, N. One Size Does Not Fit All: Revisiting World Models and Neurosymbolic AI: ACM AI LEADERSHIP SUMMIT: Defining AI's Next Chapter; August 30 – September 2, 2026, Atlanta, Georgia. https://aisummit.acm.org.
Accepted as Vision Paper - Preprint version.
Rights
© Sheth, A., Thareja, M., Pawar, A., Rawal, N.; ACM 2026. This is the author's version of the work. It is posted here for your personal use. Not for redistribution. The definitive Version of Record was published in ACM AI LEADERSHIP SUMMIT.