Learning Goals
5 minBy the end of this lesson you can:
- Name the three main ways machines learn.
- Give one everyday example of each.
- Match a task to the learning type that best fits it.
Warm-Up · How Did You Learn?
8 minLast lesson we learned that data must be good and fair. Now, how does learning actually happen?
Match how you learned each of these:
- Times tables — with flashcards that show the answers.
- Sorting your toys into groups nobody named for you.
- Getting better at a video game by playing and chasing a high score.
Reveal the link
These are the three ways machines learn too: with answers, by finding groups, and by reward.
New Concept · Three Ways to Learn
18 min1. Learning with answers (supervised)
We show the machine examples and the right labels — like flashcards with answers. This is what we've done so far: cat/dog photos, smile/frown sketches.
2. Finding groups alone (unsupervised)
We give the machine examples with no labels and ask it to find groups by itself. It might cluster shoppers into "snack buyers" and "drink buyers" without being told the groups.
3. Learning by reward (reinforcement)
The machine tries things, gets a reward for good moves and a penalty for bad ones, and improves over many tries — like levelling up in a game.
Knowing the type helps you understand any AI: did it get answers, find groups, or learn by reward?
Worked Example · Which Way Did It Learn?
18 minLet's match real AIs to their learning type.
| AI | Learning type | Why |
|---|---|---|
| Spam filter | With answers | Trained on emails labelled "spam" or "not spam". |
| Grouping shoppers by habit | Finding groups | No labels — it discovers the groups itself. |
| Game boss that improves | By reward | Tries moves, scores wins, gets better. |
| Photo "is this a face?" | With answers | Trained on photos labelled face / not face. |
Ask: were there labels (answers)? No labels but groups? Or trying for a reward? That tells you the type.
Try It Yourself
20 minUse your worksheet.
Label each "with answers", "finding groups", or "by reward": a robot learning to walk by trying; sorting songs into moods with no labels; predicting pass/fail from labelled past results.
Hint
Look for: answers given, groups discovered, or a reward for trying.
Write one example of each of the three types — they can be from real life, not just AI.
Hint
Flashcards = with answers. Tidying into your own groups = finding groups. Practising for a score = by reward.
Mini-Challenge · Pick the Right Way
12 minFor each task, decide which learning type fits best and say why.
- Teach a drone to land smoothly.
- Split 1,000 unlabelled customer reviews into natural groups.
- Tell ripe from unripe mangoes using labelled photos.
It works if each choice matches the clue: reward for trying, groups with no labels, or answers given.
Show the answers
- By reward — it learns smooth landing by trying and scoring.
- Finding groups — no labels, discover the clusters.
- With answers — labelled ripe/unripe photos.
Recap
5 minMachines learn in three main ways: with answers (supervised), by finding groups (unsupervised), and by reward (reinforcement). The right way depends on the job and the data you have.
Vocabulary Card
- supervised learning
- Learning from examples that come with the right answers (labels).
- unsupervised learning
- Finding groups or patterns in data that has no labels.
- reinforcement learning
- Learning by trying actions and getting rewards or penalties.
Homework · Guess the Way
≤ 20 minPick two AIs you know. For each, guess which of the three ways it most likely learned, and write a one-line reason.
Sample · Guess the Way
- Spam filter — with answers; it trained on emails marked spam or not.
- Chess AI that improves by playing itself — by reward; it learns from winning and losing.
Yours will be different — sensible reasons matter more than being exactly right.