AI World Models
A comprehensive course for technically curious learners who want to understand how AI systems learn internal representations, imagine futures, plan actions, generate interactive worlds, and support physical AI.
4 sections ยท 16 lessons
Course outline
Modeling Foundations
- Mental Simulators
- Latent State Compression
- Action-Conditioned Dynamics
- Planning by Imagination
- Audit a World-Model Claim
Control Algorithms
- Model-Based Control
- MuZero's Planning Targets
- DreamerV3's Generality
- Bad Dreams and Uncertainty
- Choose the Control Loop
Video World Models
- Visual Patch Scaling
- Coherence and State
- Genie's Latent Actions
- V-JEPA's Embedding Predictions
- Judge a Video Simulator
Physical AI
- World Foundation Models
- Digital Twins and Synthetic Data
- Benchmarks for Causality
- Roadmap and Limits
- Build a World-Model Scorecard