Dezhi Luo

Spatial world models, representations that support flexible reasoning about spatial relations, are central to developing computational models that could operate in the physical world, but their precise mechanistic underpinnings are nuanced by the borrowing of underspecified or misguided accounts of human cognition. This paper revisits the simulation versus rendering dichotomy and draws on evidence from aphantasia to argue that fine-grained perceptual content is critical for model-based spatial reasoning. Drawing on recent research into the neural basis of visual awareness, we propose that spatial simulation and perceptual experience depend on shared representational geometries captured by higher-order indices of perceptual relations. We argue that recent developments in embodied AI support this claim, where rich perceptual details improve performance on physics-based world engagements. To this end, we call for the development of architectures capable of maintaining structured perceptual representations as a step toward spatial world modelling in AI.

Despite excelling in high-level reasoning, current language models lack robustness in real-world scenarios and perform poorly on fundamental problem-solving tasks that are intuitive to humans. This paper argues that both challenges stem from a core discrepancy between human and machine cognitive development. While both systems rely on increasing representational power, the absence of core knowledge-foundational cognitive structures in humans-prevents language models from developing robust, generalizable abilities, where complex skills are grounded in simpler ones within their respective domains. It explores empirical evidence of core knowledge in humans, analyzes why language models fail to acquire it, and argues that this limitation is not an inherent architectural constraint. Finally, it outlines a workable proposal for systematically integrating core knowledge into future multi-modal language models through the large-scale generation of synthetic training data using a cognitive prototyping strategy.

Self models contribute to key functional domains of human intelligence that are not yet presented in today’s artificial intelligence. One important aspect of human problem-solving involves the use of conceptual self-knowledge to detect self-relevant information presented in the environment, which guides the subsequent retrieval of autobiographical memories that are relevant to the task at hand. This process enables each human to behave self-consistently in our own way across complex situations, manifested as self-interest and trait-like characteristics. In this paper, we outline a computational framework that implements the conceptual aspect of human self models through a modified version of the joint-embedding predictive architecture. We propose that through the incorporation of human-like autobiographical memory retrieval and self-importance evaluation, the modified architecture could support machine agents with significantly enhanced self-consistency, which could be applied to deliver more believable simulations of human behaviors.