Generative AI for Healthcare (Part 1): Demystifying Large Language Models

Unlocking the true potential of generative AI starts with understanding how it works. This video—the first in a new educational series—introduces healthcare professionals to large language models (LLMs) like ChatGPT: what they are, how they generate responses, and how to use them thoughtfully. Join us as we explore: • How LLMs fit into the broader landscape of AI in healthcare • What actually happens behind the scenes when you submit a prompt • The core techniques that shaped today’s most powerful models — and what the future holds Drawing from both foundational literature and the latest developments, this series translates complex AI concepts into practical insights—no computer science background required. Shivam Vedak, MD, MBA - https://medicine.stanford.edu/profiles/shivam-vedak Dong-han Yao, MD - https://med.stanford.edu/profiles/dong More about the speakers: Shivam Vedak, MD, MBA, and Dong Yao, MD, are physicians and clinical informaticists at Stanford Medicine. Their work focuses on the practical application of generative AI in healthcare, bridging system-level implementation and frontline clinician education. They have been invited to present and teach on this topic at academic institutions and conferences nationwide, reaching a diverse audience of physicians, healthcare IT professionals, and other clinical leaders. Chapters: 0:00 — Introductions and Disclosures 2:50 — Why Is Prompting Hard? 6:55 — The Three Epochs of Healthcare AI 18:01 — Tokenization and Embeddings 27:58 — Transformer Architecture and Self-Attention 34:45 — Pre-Training and the Evolution of LLMs 44:01 — Post-Training: Making the Model Helpful and Aligned 49:29 — The Reasoning Era: Scaling Test-Time Compute 54:18 — Summary: What Is an LLM?