Recent work in neuroscience indicates that the human brain constructs internal models of the world through interaction rather than observation alone. That distinction matters for AI. A system that only recognises patterns can interpolate, but it cannot plan, because it has no model of what its own actions would do.
This fellowship incorporates world-model learning into AI systems so they can plan and reason, with the obvious applications in robotics and autonomous systems. The shift is from passive pattern recognition to active learning, aiming at models that are more efficient, more transparent, and carry something closer to a causal understanding of their environment.
It is the research line that runs directly into what we build at Binabik.