Course 1, lesson 8 of 100, Ages 3+
Practice makes better
Learning by trying again
Like I’m 5
When you learn to throw a ball, you miss at first. You try again and again, and you get better. AI learns the same way, by practising lots.
The big idea
Every time you practise, you notice what went wrong and change a little. Threw too far? Next time throw softer. Bit by bit, your throws land closer.
AI practises on examples thousands or millions of times. Each time, it checks how wrong it was and nudges itself to be a little less wrong. That's why AI gets better with more practice and more examples.
Examples
- Riding a bike: Wobbly at first, steady after lots of tries.
- Handwriting apps: They practised on thousands of handwritten letters, so they can read yours.
- Game-playing AI: It played millions of games against itself to get good.
How it works
- Try the task.
- Check how close you got.
- Change a little and try again.
Check your understanding
- How does AI get better?
- Options: By practising on lots of examples; By sleeping; By getting a new coat of paint.
Answer: By practising on lots of examples. Practice with feedback is how AI improves. - What should you do after a miss?
- Options: Notice what went wrong and try again; Give up forever; Blame the ball.
Answer: Notice what went wrong and try again. Noticing mistakes and adjusting is how people and AI both learn.
Remember
Try, check, change, repeat. That's how you and AI both get better.
Talk about it
What's something you got better at by practising?
Go deeper
This try-check-adjust loop is the heart of training. The 'check' is a loss function that measures the error, and the 'adjust' is an optimiser such as gradient descent.