← Back to Curriculum
⚙️
Year 9 • Lesson 1

How AI Really Works

Under the hood of large language models

You can't use a tool critically until you understand what it actually does.

🎯 Learning Objectives

  • ✓Explain how large language models (LLMs) generate text using pattern prediction
  • ✓Describe what training data is and why it creates bias
  • ✓Identify three structural limitations of current AI systems

📚 Key Vocabulary

Large Language Model (LLM)
An AI trained on massive text datasets to predict likely next words
Token
A chunk of text (roughly a word or part of a word) that AI processes
Temperature
How creative/random vs. focused an AI's output is — a setting you can control
Training data
The massive corpus of text used to train the model — full of human biases
Context window
How much text an AI can "see" at once — like working memory

⏱️ Lesson Timing (50 mins)

🪝
Hook
5 mins

"Finish this sentence: The best thing about school is..." Students guess what AI will say. Teacher enters it live and shows result. Then repeat with an unusual context. How does AI decide? That's today's lesson.

👨‍🏫
Teach
15 mins
  • •LLMs: trained on trillions of words from the internet, books, code
  • •How text generation works: probability prediction, one token at a time
  • •Temperature: low = focused/predictable, high = creative/random
  • •Training data problems: internet text is not balanced — skews toward English, Western, male
  • •Context window: AI has no true memory beyond the current conversation
  • •Key companies: OpenAI (ChatGPT), Microsoft (Copilot), Anthropic (Claude), Meta (Llama)
  • •Multimodal AI: text, image, audio, video — all converging
🛠️
Create
35 mins

Investigation task: students run a set of structured experiments to observe AI behaviour. (1) Test temperature by asking the same question multiple times — is it consistent? (2) Find an example of AI reproducing a cultural or gender bias. (3) Test the context window by giving a very long context and seeing if AI remembers detail from the start. (4) Find a task where AI fails completely — and write a hypothesis for WHY. Document with screenshots in a "How AI Works" investigation report.

Tools:
Microsoft CopilotWord Online
📢
Share
5 mins

What surprised you most? What would you want AI developers to fix?

🔧 Tools & Resources

Microsoft Copilot
✅ Free
https://copilot.microsoft.com

🤖 AI Angle

This IS the AI lesson — building the critical literacy foundation for all Year 9 AI work.

📝 Evidence of Learning

How AI Works investigation report with screenshots and observations

📚 Homework

Watch a 10-minute explainer video on how LLMs work (e.g. 3Blue1Brown or Computerphile on YouTube). Write 5 things you learned that weren't in today's lesson.

⏱️ 20 mins

♿ Differentiation

Support (for students who need help)

Structured experiment sheet with exact prompts to use

Extension (for advanced learners)

Research how AI training has changed from GPT-2 to GPT-4. What were the key breakthroughs?

← Back to All Lessons📝 Take the Quiz