My first AI adventure ☀

The Merlion and the missing rainbow 📖

Your mission: create a three-page mini story, with a beginning, a problem and a kind solution.

1. Child: name an imaginary Merlion. What makes it special? What does it want?
2. Grown-up: open ChatGPT and try this prompt:

Help us plan a funny, gentle story for a 7-year-old about a tiny Merlion in Singapore who finds a rainbow hiding in a kaya toast shop. Suggest three possible problems, using simple vocabulary. Ask us to choose one before continuing. Leave the ending for us to invent.

3. Together: choose one idea. Your child tells or writes what happens next. Ask “How does our hero feel?” and “What kind thing could they try?”
4. Child: draw three scenes on paper or try AutoDraw for a cover. Add your own words. Read your story aloud together.

My hero’s big idea…

One more spark: let the rainbow sneeze colours! What funny thing happens to the toast?

The hawker-centre change challenge 🔎

Mei buys 3 bowls of noodles. Each bowl costs S$4. She pays with S$20. How much change should she receive?

1. Child: draw three bowls and write the price below each one.
2. Grown-up: ask ChatGPT for a clue, using this prompt:

We are solving this maths problem together: 3 bowls of noodles cost S$4 each and Mei pays S$20. Give one short hint suitable for an 8-year-old. Do not give the answer. Wait for our working before the next hint.

3. Child: work out the total cost, then the change. Explain your thinking out loud.
4. Together: check by adding the cost and the change. Does it make S$20? Check the AI suggestion too.

My detective working…

Make your own mystery: change one number and solve again. Can you make a problem for your grown-up?

Can a computer become a leaf detective? 🌱

Your mission: see how AI learns from examples, and why it sometimes gets things wrong.

1. Draw 6 leaf pictures and 6 flower pictures. Try different shapes and colours. You can also use your own photos of objects, without people or identifying details.
2. With a grown-up, open Google Teachable Machine. Choose an Image Project → Standard image model.
3. Name the two classes “Leaves” and “Flowers”. Upload examples into the correct class. Keep two additional drawings aside for testing.
4. Select “Train Model”. Test the two pictures it has never seen. No webcam or model export is needed for this activity.
5. Try a tricky drawing. Did the computer look at the shape, the colour or the background? Add more varied examples and train again.

What surprised our family?

Science thought: a confident prediction isn’t proof. We test with new examples and explain what we observe.