
Deep Generative Models I: Lecture 1 - Introduction
Stefano Ermon, Associate Professor of Computer Science at Stanford, opens CS236 by laying out what generative models are and why they matter. He frames the core problem as building simulators for complex, high-dimensional data such as images, text, and video, then works through the probability distribution concepts that make those simulators possible, including inverse problems and natural control signals for generation. Ermon surveys applications in medicine, healthcare, and robotics, and traces how fast the field has moved since the course started. The lecture mixes conceptual grounding, the philosophy of generative versus discriminative modeling, with practical course mechanics: project expectations, grading requirements, and the balance between theory and hands-on coding the course demands. It closes with his pitch for why mastering these models, both the math and the implementation, is worth the course's workload.