Volume 384 Issue 2320 | Philosophical Transactions of the Royal Society A | The Royal Society
Four instantiations of world modelling and self-modelling across biological and artificial systems: spatial cognition in rodents, reflective self-representation in primates, map-based planning in autonomous vehicles, and mirror-based self-recognition in robots. Credit: Victoria Klimaj, Adam Safron, and Amelia Thomley. How intelligent are language models, and how might their capabilities evolve in the coming years? How might we create artificial systems with the adaptive intelligence of naturally evolved organisms? To what extent are the remarkable abilities of data-driven foundation models similar to the context-sensitive causal world modeling of biological intelligences, or does dependence on learning from massive statistical associations mean such systems are fundamentally limited in their abilities to learn coherently integrated models of self and world? Through exploring diverse forms of world modeling ranging from agentic control to collective coordination, to the rich narratives