In this episode, we discuss Critiques of World Models by Eric Xing, Mingkai Deng, Jinyu Hou, Zhiting Hu. The paper critiques existing approaches to world models by emphasizing their role in simulating all actionable possibilities for reasoning and acting. It proposes a new general-purpose world model architecture featuring hierarchical, multi-level, and mixed continuous/discrete representations learned via generative and self-supervised methods. The authors envision this model as enabling a Physical, Agentic, and Nested (PAN) AGI system.