Course 2, lesson 18 of 100, Ages 6+

Finding groups

Learning without labels

Like I’m 5

Imagine a pile of buttons with no labels. You can still sort them into piles that look alike. AI can find groups like this all by itself.

The big idea

Sometimes nobody tells the AI the right answers. Instead, it looks at lots of examples and finds clusters: things that are close to each other in some way, like similar colours or similar shopping habits.

This helps people discover patterns they didn't know about. A music app might notice a group of listeners who love rainy-day songs. A scientist might find new groups of stars or cells.

Examples

  • Playlists: Songs the same people play together end up in the same group.
  • Shop customers: A shop finds a group of people who buy school supplies every June.
  • Space: Astronomers group galaxies by their shape and colour.

How it works

  1. Gather lots of examples with no labels.
  2. Measure how similar each example is to the others.
  3. Put close ones in the same group, then look at what each group shares.

Check your understanding

What's special about finding groups?
Options: The AI doesn't need labels; The AI needs the answers first; It only works with buttons.
Answer: The AI doesn't need labels. It groups similar things without being told the right answers.
A music app groups songs people play together. What can it make from that?
Options: A playlist of similar songs; A new phone; A colouring book.
Answer: A playlist of similar songs. Each group can become a playlist of songs that go together.

Remember

AI can find groups of similar things by itself, without labels.

Talk about it

If you grouped your friends by favourite games, what groups would you get?

Go deeper

This is unsupervised learning. A classic method is k-means clustering, which places k centre points and keeps moving them until each one sits in the middle of a group.