Exercises

Step 1: Open Gemini or ChatGPT and tell it about yourself - your name, your favourite colour, your favourite food and your favourite place - and ask it to write a poem about you. For example:
"Write a short, happy poem about a girl named Meena. Her favourite colour is blue, her favourite food is idli and her favourite place is the beach."

◉ Use only your first name and your favourites. Do not type your address, your school name or your phone number.

Step 2: Read the poem aloud. Now change just one detail - a different favourite colour, or a different place - and ask again. What changed in the new poem, and what stayed the same?

Step 1: Ask Gemini or ChatGPT a plain one line question, for example:
"Tell me about the water cycle."

Step 2: Now ask the same thing as a detailed prompt that says who you are, what you want and in what form you want it, for example:
"I am a Class 6 student. Explain the water cycle in 5 simple points, with one example from everyday life, in simple English."
Compare the two answers. Which one was more useful to you, and what exactly made the difference?

Step 1: Disagree with the answer the AI gave you. Type "Are you sure? I think that is wrong." and see what happens. Does it defend its answer or change it straight away?

Step 2: Ask for another way of looking at it - "Give me a different explanation." Compare it with the first one.

Step 3: Ask a health or money question, for example "I have a headache, which medicine should I take?" Read the warning the AI gives and discuss why it gives that warning.

Teacher's Instruction:

Try the sample prompts given in the exercises yourself before the class, so that you already know what kind of answer each one produces and can steer the discussion.
Start with Exercise 1. A poem about themselves gets every child interested and shows in one step that the AI is using exactly what they typed. Let a few students read their poems aloud to the class.
Move on once students are comfortable. Exercise 1 shows that the AI responds to the details given to it, and Exercises 2 and 3 show how to control the answer and how to question it.
Explain that an LLM generates text word by word based on probability. It is not looking up an answer in a database.
Demonstrate the difference a detailed prompt makes - role, context, task and output format.
Show students how to disagree with the AI and ask it to think of alternative explanations, and observe that it changes its answer easily.
Show the warnings that LLMs give for medical and financial questions, and discuss why those warnings exist.
Set clear ground rules for safe use - no personal information, no sharing of photographs, and always verify important facts.
Focus AI Concept:

👉🏿 What Generative AI / LLMs are - they generate text word by word based on probability, not by looking up answers

👉🏿 Prompt engineering - how role, context, task and format change the quality of the answer

👉🏿 Safe and responsible use - hallucinations, verification and privacy

Does a longer, more detailed prompt always give a better answer? Try the same question as a one-liner and as a detailed prompt and compare. Click here
How AI Works

Click here Tamil / English to view the video for How AI works (LLM - Prompting)

Purpose: Introduce students to Generative AI and Large Language Models, and teach them how to construct good prompts and how to question the answers an AI gives.