Tutorials · Chapter A (1/4) · ~8 min

Data fuels AI

Try it → see it → read → next

Models don’t invent magic — they learn from examples you (or someone) gathered.

Try yourself

Playground

Fuel the learner

Label each message yourself — watch the fuel gauge (accuracy) rise. Data is the ingredient.

Tap a card, then Spam or Not spam

Recap

What you just did

You practiced teaching with examples (spam vs not). That’s the heart of modern AI: show many cases, let the model approximate the rule.

Teach

How it works

See it

Data → pattern → guess

PhotosMessagesClicks

Modelfinds patterns

Guesson new stuff

Thin or skewed data = thin or skewed learning

See it: examples → model finds patterns → guess. Labels (“spam”, “liked”) and variety matter. One recipe copied fifty times → false confidence; many kitchens → better new-dish guesses.

Use it

When you'd use this

Judging a product claim (“trained on medical images”) — ask whose data, how recent, how labeled.
Building anything — plan how you’ll get clean examples before you pick a model.

Watch out

More data isn’t automatically fairer. Biased collections bake bias into the learner. Privacy: collecting data has rules and people attached.

Try next

Try this next

List three datasets you interact with today (photos, playlists, map routes). For each, guess what label the app is optimizing.