Course 2, lesson 11 of 100, Ages 6+
Teach Pip
Learning from examples
Like I’m 5
If you show a friend lots of apples, they learn what an apple looks like. AI learns the same way, by seeing lots of examples.
The big idea
To teach an AI what an apple is, you don't write a rule like 'apples are round and red'. Instead, you show it lots of pictures labelled 'apple' and lots labelled 'not apple'.
From those examples, the AI works out the clues by itself: the shape, the shine, the little stalk. Then it can recognise apples it has never seen before, even green ones, if it saw enough variety.
Examples
- Fruit sorting: Shown thousands of labelled fruit photos, an AI learns to tell apples from pears.
- Your phone keyboard: It learned from millions of typed sentences which words usually come next.
- Plant apps: People sent photos of plants with their names, so the app can name new ones.
Check your understanding
- How did Pip learn what an apple is?
- Options: By seeing lots of examples; By eating one; By guessing at random.
Answer: By seeing lots of examples. Pip learned from labelled examples. That’s how lots of AI learns. - Why should the examples include green apples too?
- Options: So the AI learns apples can be green; So it learns apples are always red; Colour never matters.
Answer: So the AI learns apples can be green. Variety in the examples helps AI recognise things it hasn't seen before.
Remember
More examples, and more different ones, mean better guesses.
Talk about it
How did you learn to recognise your favourite fruit?
Go deeper
This is supervised learning. Labelled examples go in, and the model learns what makes an apple an apple. More examples, and more varied ones, usually mean better guesses.