Learning World Models through Actionable Representation for Next-Generation AI
Inspired by neuroscience, this AI approach mimics how the brain builds internal world models through interaction. This shift from passive pattern recognition to active learning aims to create more efficient and transparent AI with a deeper, causal understanding, enabling true planning for next-gen robotics.
Description
Recent neuroscience studies indicate that the human brain constructs internal models of the world through (inter-)actions, a process likely to fundamentally enhance the generalization, transparency, and efficiency of current AI systems.
This project aims to incorporate world model learning into AI to enable planning and reasoning, with significant applications in robotics and autonomous systems.
Key data
Projectlead
Project status
Start imminent, 12/2025
Institute/Centre
Centre for Artificial Intelligence (CAI)
Funding partner
Digitalisierungsinitiative der Zürcher Hochschulen DIZH / DIZH Fellows 2025
Project budget
198'500 CHF
