Build a real AI step by step, test how good it is, check it is fair — then explore more in the AI Lab.
💡 Remember about AI: AI is a clever helper, but it is not always right — if it is unsure, check with a teacher or grown-up. And never tell an AI your private things (your home, your password, or photos of yourself) — treat it like a stranger online.
1 · Choose
2 · Collect data
3 · Train
4 · Test
5 · Improve & be fair
Step 1 · Choose what your AI will learn
An AI learns to tell things apart. First make your groups (at least two). Ideas that work well: Ripe / Unripe banana, Leaf / Stick, Goat / Cow, Happy / Sad drawing.
Step 2 · Collect data (example pictures)
Give each group lots of example pictures. More examples = smarter AI. Try to give each group about the same number (that keeps it fair!).
Step 3 · Train the AI
Training means letting the AI study all your examples so it can find the pattern. Press the button and watch it learn!
💡 This AI learns by remembering what your examples look like (their colours and shapes). When you test a new picture, it finds the examples most like it — just like you know a goat because it looks like other goats you have seen.
Step 4 · Test your AI
Show the AI a new picture it has never seen. Can it guess the right group?
Step 5 · Measure how good it is, and be fair
Real AI builders always measure their model. Add a few test pictures (pictures you did NOT train with) to each group, then press "Score my AI". It tells you what percent it gets right — and warns you if your AI is unfair.
?
💡 To make your AI better: add more example pictures in Step 2 and train again. To make it fairer: give every group about the same number of varied examples. Then score it again and watch the number go up!
🔬 AI Lab
AI is not only about pictures! Try two more ways computers can be clever. Pick one:
Rules AI — sorting with yes/no questions
Some AI sorts things by asking yes/no questions about their features — like a game of "20 questions". Build a rule to tell two animals apart!
1. What are you sorting? (two things)
vs
2. Your yes/no rule
Pick a feature that is TRUE for one and FALSE for the other:
Does it? If YES, it is a
3. Test your rule
💡 This is called a decision tree — a real kind of AI that sorts things by asking questions about their features, not by looking at pictures.
Pattern AI — guess what comes next
Big AI learns patterns to predict what comes next (that is how it finishes your sentences!). Build a colour pattern and let the AI guess the next colour.
1. Build a repeating pattern
2. Let the AI predict
💡 The AI looks at the pattern you made, finds the repeat, and predicts the next one — the same idea behind how AI predicts the next word when it writes.
Next-Word AI — how AI writes sentences
AI that writes (like the text helper on a phone) reads lots of sentences and learns which word usually comes next. Build a sentence by tapping the AI's guesses — one word at a time!
1. Pick a word to start
2. Your sentence so far
…
3. The AI's next-word guesses — tap one
💡 The AI guesses from the sentences it has read. It picks the words that came next most often. Real AI has read billions of words — that is why its guesses are so good and it can write whole stories!
🏠 Home