How Large Language Models Work: What Claude, ChatGPT, and Gemini Actually Do
You do not need to understand the mathematics behind large language models any more than you need to understand the internal combustion engine to drive safely. But just as a driver benefits from knowing that wet roads reduce braking distance, a clinician using AI benefits from understanding a few basic mechanics. These mechanics explain most of the surprising, frustrating, and occasionally dangerous behaviors you will encounter when you start working with these tools.
This course covers what you actually need to know, and nothing more.
The core idea: predicting the next word
A large language model, or LLM, is fundamentally a system trained to predict what word comes next given everything that came before it. That is the engine underneath Claude, ChatGPT, Gemini, and every similar system. It sounds almost too simple to explain their capabilities, but the scale at which this prediction operates produces something genuinely remarkable.
These systems have been trained on billions of documents: textbooks, websites, research papers, clinical guidelines, patient forums, novels, news articles, court documents, and more. Through exposure to all of that text, they have learned not just words but the patterns of reasoning, explanation, and argumentation that appear in human language. When you ask a medical question, the model is drawing on patterns absorbed from an enormous corpus of medical text to construct an answer that fits what a knowledgeable response would look like.
