AI & ML Collision

AI can now learn the laws of physics from the perspective of a single toddler instead of millions of video frames.

April 14, 2026

Original Paper

Zero-shot World Models Are Developmentally Efficient Learners

Khai Loong Aw, Klemen Kotar, Wanhee Lee, Seungwoo Kim, Khaled Jedoui, Rahul Venkatesh, Lilian Naing Chen, Michael C. Frank, Daniel L.K. Yamins

arXiv · 2604.10333

The Takeaway

Using a model (ZWM) that mimics child development, the system learns depth, motion, and coherence from sparse, first-person data. This shifts the focus from 'big data' to 'developmental efficiency,' showing that human-scale data is enough to understand the physical world.

From the abstract

Young children demonstrate early abilities to understand their physical world, estimating depth, motion, object coherence, interactions, and many other aspects of physical scene understanding. Children are both data-efficient and flexible cognitive systems, creating competence despite extremely limited training data, while generalizing to myriad untrained tasks -- a major challenge even for today's best AI systems. Here we introduce a novel computational hypothesis for these abilities, the Zero-