Module 1
Module 1: AI Concepts & Terminology
Master the vocabulary of AI and understand how LLMs actually work
Welcome to Module 1!
Before we jump into writing complex medical notes, we need to speak the same language. This module is designed for absolute beginners. If you’ve never used ChatGPT or an LLM before, you are in the right place.
Learning Objectives
By the end of this module, you will understand:
- What a “Prompt” actually is (Input vs Output)
- Zero-Shot vs. Few-Shot Prompting (The power of examples)
- The Context Window (How much can the AI remember?)
- Hallucinations (Why AI lies and how to catch it)
Lesson 1: The “Hello World” of Prompting
An LLM (Large Language Model) is like a very advanced autocomplete engine. It predicts the next word based on the text you give it.
The text you give it is called the Prompt. The text it generates is called the Output.
The quality of the Output depends entirely on the quality of the Prompt. This is called “Garbage In, Garbage Out.”
Exercise 1.1: Your First Prompt
Let’s start with something simple. You have an email from a practice manager, Sarah. She wants to schedule a meeting. Your job is to extract the available times.
Exercise 1.1: The "Hello World" of Prompting
Your Challenge
Write a prompt to extract the available meeting times from Sarah's email.
View Source Email
Generated Output:
Lesson 2: Zero-Shot vs. Few-Shot
In AI terms, a “Shot” is an example.
- Zero-Shot: Asking the AI to do something without giving any examples.
- Example: “Translate this to Spanish.”
- Few-Shot: Giving the AI one or more examples of what you want.
- Example: “Translate this to Spanish. Example: Hello -> Hola. Text: Goodbye.”
Few-Shot is almost always better. It shows the AI the pattern you want it to follow.
Exercise 1.2: The Power of Examples
Now, let’s try to get that same email data into a specific format: JSON (a format computers use). This is hard to describe in words, but easy to show with an example.
Exercise 1.2: Zero-Shot vs Few-Shot
Your Challenge
Extract the meeting times into a JSON object. Try providing an example (Few-Shot) to get the format exactly right.
View Source Email
Generated Output:
Lesson 3: The Context Window
The Context Window is the “short-term memory” of the AI. It’s the maximum amount of text the AI can consider at one time.
If you paste a 50-page medical record, the AI might “forget” the beginning by the time it reads the end. Or, it might hallucinate details because it’s overwhelmed.
Key Concept: You must be specific about where the AI should look.
Exercise 1.3: Finding the Needle
Now we switch to a medical transcript. It’s long. Your job is to find one specific detail without making the AI hallucinate.
Exercise 1.3: The Context Window
Your Challenge
Find exactly what the patient said about their diet. Do not include other details.
View Patient Transcript
Generated Output:
Module 1 Complete! 🎉
You’ve mastered the basics! You now understand:
- Prompts are just instructions.
- Examples (Few-Shot) make prompts much more powerful.
- Context is key to accuracy.
Next, we’ll apply these concepts to build full clinical notes.