Unit Title: Understanding Large Language Models (LLMs) and Generative AI
Level: Introductory–Intermediate
Duration: ~90–120 minutes (flexible)
🎯 Learning Objectives
By the end of this week, you should be able to:
- Explain what a Large Language Model (LLM) is and how it works conceptually.
- Understand what Generative AI means (and how it differs from traditional AI).
- Identify the differences between language generation, completion, and classification.
- Recognise the role of tokens, probability, and training data in generative tools.
- Evaluate 3 examples of generative AI use and explain how they work.
🧭 Lesson Flow
| Segment | Duration | Format |
|---|---|---|
| 1. What Are LLMs? | 20 min | Concept + Analogy |
| 2. How Do They Work? | 25 min | Step-by-step logic + Diagrams |
| 3. Use Cases of Generative AI | 20 min | Examples + Reflections |
| 4. Risks and Misconceptions | 10 min | Guided Reading |
| 5. Exercises & Knowledge Check | 30–45 min | Hands-on activities |
🧑🏫 1. What Are Large Language Models (LLMs)?
📖 Teaching Script:
A Large Language Model (LLM) is a type of AI that has been trained on vast amounts of text to predict what comes next in a sentence — one word (or token) at a time.
LLMs don’t “know” meaning like humans do. Instead, they’ve seen so much language that they can generate incredibly accurate, creative, or useful completions based on statistical patterns.
🔍 Analogy: “AI as a Super Autocomplete”
Think of an LLM as a supercharged autocomplete tool.
When you type “Can you help me with…”, it looks at millions of similar sentences and chooses the most likely continuation.
It’s not “thinking.” It’s predicting.
🧠 Simple Working Definition:
A language model is an AI system trained to understand and generate human language by predicting the most probable next word or phrase.
🧩 2. How Do LLMs Work?
📘 Step-by-Step Logic
- Training
- The model is trained on billions of pages of text from books, websites, articles, etc.
- It learns to predict the next token (word or part of a word) given previous tokens.
- Tokenisation
- Text is split into tokens (e.g. “ChatGPT is amazing” → “Chat”, “G”, “PT”, “is”, “amazing”).
- Each token is given a numeric ID.
- Probability Assignment
- For each new token, the model calculates the probability of every possible next token.
- It then selects the highest-probability one — or samples creatively.
- Output Generation
- The AI strings together tokens to form coherent sentences.
- It continues until you stop it, or it reaches a limit.
🖼️ Diagram Prompt:
Draw or imagine this process:
User Input ➝ Tokenised ➝ Pattern Search ➝ Probability Assigned ➝ Best Token Chosen ➝ Next Token ➝ Output
🎨 3. Use Cases of Generative AI (with Examples)
| Use Case | Description | Example |
|---|---|---|
| Text Generation | Creating articles, stories, or responses | ChatGPT writing a product description |
| Summarisation | Condensing long text into short summaries | A legal summary of a 10-page contract |
| Translation & Paraphrasing | Changing language or tone | Rewriting a formal email into casual tone |
💬 Reflection Prompt:
For each of the three above, write:
- Where you’ve seen it in your life
- A benefit and a possible risk
⚠️ 4. Risks, Misconceptions, and Limitations
| Misconception | Reality |
|---|---|
| “AI understands language like us” | No — it predicts based on patterns, not meaning |
| “LLMs are 100% accurate” | No — they can hallucinate or fabricate false content |
| “Bigger = always better” | Not always — size helps, but training quality and design matter more |
🚨 Critical Concept: “Hallucination”
This is when an AI confidently produces an incorrect or fictional answer.
Example: “The Eiffel Tower is in Berlin.”
LLMs may “guess” wrong if the patterns suggest the wrong context.
🧪 5. Exercises & Knowledge Check
✅ Exercise 1: Predict the Next Token
Try to complete these sentences based on probability:
- “The sun rises in the ____.”
- “Albert Einstein is famous for the theory of ____.”
- “I would like to book a ____ for two people.”
Reflect: These aren’t “right” — they’re most likely. That’s how LLMs work.
✅ Exercise 2: LLM Use Case Mapping
Match each tool with its primary function:
| Tool | Function |
|---|---|
| ChatGPT | ? |
| Grammarly | ? |
| DeepL | ? |
✅ Exercise 3: Explain in Plain English
Without using technical terms, write:
- “How does a language model work?”
- Use 100 words or fewer.
- Imagine explaining it to a 10-year-old.
🧠 Knowledge Check (10 Questions)
- What is a token?
- What does an LLM predict?
- What is generative AI?
- Name one example of a generative task.
- What is the difference between summarisation and generation?
- What is hallucination in AI?
- Why can LLMs make factual mistakes?
- What’s an example of LLM use in everyday life?
- Do LLMs understand meaning?
- What’s one strength and one weakness of an LLM?
📝 Wrap-Up Assignment (Optional)
Title: “My First Encounter with a Language Model”
Write ~250 words describing:
- A tool you’ve used that’s powered by an LLM
- What it helped you do
- What you found surprising or confusing about how it responded
📦 End-of-Week Deliverables
- ✅ Diagram of the LLM process
- ✅ 3 use cases (benefits + risks)
- ✅ Plain-English explanation of LLMs
- ✅ Answered knowledge check
- ✅ Reflection journal or short essay