Nima Ajam Gard•January 22, 2026•13 min read

What's Missing from Robot Learning (Part I)

In this series, I examine what is missing from today's dominant approaches to robot learning. My focus is pragmatic: what is the next meaningful step toward scalable, reliable physical AI systems that actually work in the real world.

Artistic representation of learning to snowboard - Physical AI robot learning analogy — Learning a physical skill like snowboarding requires more than observation—it requires embodiment.

Three components of physical AI robot learning: Imitation Learning, World Model, and Reinforcement Learning — The three components of robot learning map to different aspects of skill acquisition.

https://path-robotics-1.wistia.com/medias/i13jzgw6ot

Scott McCloud Understanding Comics illustration about self-awareness and extended identity in physical AI world models — From Scott McCloud's "Understanding Comics" [11] - our awareness of self flows outward to include objects of our extended identity.

NVIDIA physical AI simulation ecosystem diagram showing Physics Simulation, Asset Generation, and Skill Training axes — The NVIDIA ecosystem spans three axes: Physics Simulation (Newton/Warp, Isaac Sim), Skill Training (Isaac Labs), and Asset Generation (Replicator).

Lucid Sim generating diverse simulated training scenes for physical AI robot learning — Lucid Sim [14] generating simulated scenes - both approaches can generate effectively unbounded data for robot learning.

Physical AI world model training workflow: Build Scenes, Interaction, and Train in parallel — A world model training workflow: Build scenes from real-world videos, add robots for interaction, and train in parallel with various textures, scenarios, and lighting.