Lesson 3: Three Ways AI Learns

What You'll Learn

The three main styles of machine learning and what makes each different
A simple everyday analogy for supervised, unsupervised, and reinforcement learning
How to recognize which style fits a given problem

Not One Way, But Three

In the last two lessons we said AI learns from examples by adjusting a model. But there isn't just one way to do that. There are three main styles of learning, and the big difference between them comes down to a single question: what kind of feedback does the AI get?

Sometimes it gets an answer key. Sometimes it gets no answers at all. And sometimes it just gets rewarded or penalized for what it tries. Let's meet all three.

Supervised Learning: Learning With an Answer Key

Supervised learning is learning from examples that come with the correct answers attached. Each example is labeled, so the AI can check its guess against the truth and improve.

Think of studying with flashcards. The front shows a question; the back shows the answer. You guess, flip the card, and see if you were right. Over many cards, you get better. The answer on the back is the "label."

Everyday examples:

Photos labeled "cat" or "dog," so the AI learns to tell them apart.
Past emails marked "spam" or "not spam," so a filter learns the difference.
Houses with their known sale prices, so the AI learns to estimate prices.

The key feature: every example comes with the right answer, like a teacher's answer key.

Unsupervised Learning: Finding Patterns With No Answer Key

Unsupervised learning is learning from examples that have no labels. There's no answer key. The AI's job is to find structure on its own — usually by grouping similar things together.

Imagine dumping a big box of mixed LEGO bricks on the floor and sorting them into piles without being told the categories. You might end up with a pile of red ones, a pile of long ones, a pile of wheels. Nobody told you those were the "right" groups — you discovered them by noticing what's similar.

Everyday examples:

A shop grouping customers with similar buying habits, without deciding the groups in advance.
Sorting a huge pile of news articles into clusters that turn out to be about sports, politics, or food.
Noticing that a few transactions look very different from all the rest.

The key feature: no answers are given. The AI discovers the groups or patterns itself.

Reinforcement Learning: Learning by Trial and Reward

Reinforcement learning is learning by trying things and getting rewards for good outcomes and penalties for bad ones. There's no answer key telling the AI exactly what to do — only signals about whether things are going well.

The classic analogy is training a pet. You can't explain the rules to a puppy. Instead, when it sits on command, you give it a treat. When it does something you don't want, no treat. Over time, the puppy learns which actions lead to rewards and does more of them.

Everyday examples:

A program learning to play a game, scoring points for good moves and losing them for bad ones.
A robot learning to walk, "rewarded" for staying upright and moving forward.
A system learning to recommend videos, rewarded when people keep watching.

The key feature: the AI learns from the consequences of its actions — reward or penalty — rather than from labeled examples.

How to Tell Them Apart

A quick way to recognize each style:

Supervised — "Here are examples and the right answers." (Flashcards.)
Unsupervised — "Here are examples, no answers; find the patterns." (Sorting LEGO.)
Reinforcement — "Try things; I'll reward what works." (Training a pet.)

Most AI you encounter uses one of these three, and some advanced systems blend them. But the core idea is just about the feedback: an answer key, no answer key, or rewards.

Key Takeaways

Supervised learning uses labeled examples — an answer key — like studying with flashcards.
Unsupervised learning has no labels and finds patterns or groups on its own, like sorting mixed LEGO.
Reinforcement learning learns from rewards and penalties through trial and error, like training a pet.