AI Learns from Examples
Have you ever taught a younger brother, sister, or friend how to do something? Maybe you showed them how to tie their shoes, step by step. Maybe you showed them which berries in the garden are safe to touch by pointing and saying "this one yes, this one no." When you teach by showing examples, the other person watches and figures out the pattern. AI does something very similar. It learns by looking at thousands — sometimes millions — of examples that people show it.
How AI Learns: The Example Recipe
Here is how AI learning works, step by step. First, people collect a huge pile of examples. If someone wants to teach an AI to recognize photos of cats, they collect thousands of photos labeled "cat" and thousands labeled "not a cat." Second, the AI looks at every example and makes a guess. At first, its guesses are terrible — it might think a dog is a cat or a chair is a cat. Every time it guesses wrong, something inside it adjusts, just a tiny bit. Third, after seeing millions of examples and adjusting millions of times, the AI gets quite good at recognizing cats — even in photos it has never seen before. The examples are the teacher. Without examples, the AI cannot learn.
AI learns by studying many examples. The examples teach the AI what to look for. If the examples are good, the AI learns well. If the examples are bad or missing, the AI learns badly — or not at all.
Let us look at a story. Luna wanted to teach her AI helper to sort her family's recycling photos into two groups: paper and plastic. She showed the AI 500 photos of paper items and 500 photos of plastic items. Each photo was labeled. At first the AI made lots of mistakes. But after seeing all 1,000 photos, it got really good. Now Luna takes a photo of any item and the AI sorts it correctly almost every time. But then something interesting happened. Luna's neighbor Priya tried the same AI with items from her home. Some of the AI's answers were wrong! Why? Because Priya's plastic containers looked different from Luna's. The AI had only learned from Luna's examples, so it did not know everything yet. This shows something very important: the examples the AI learns from shape everything it knows.
Match each step of AI learning to what happens at that step.
Terms
Definitions
Drag terms onto their definitions, or click a term then click a definition to match.
Here is something really important to remember: the AI does not decide which examples to learn from. People do. People choose what goes into the example pile. People decide what gets labeled and how. That means people have a huge amount of power over what the AI learns. If the people choosing examples are careful and thoughtful, the AI can learn wonderful things. If the people choosing examples make mistakes — or if they leave important things out — the AI learns those mistakes too. This is why the examples matter so much. They are the foundation of everything the AI knows.
Think of examples as seeds. Plant good seeds and you grow a helpful, fair AI. Plant bad seeds — or forget to plant some — and the AI that grows can have problems. The people who choose the examples are like gardeners.
How does AI learn?
Luna's recycling AI worked great for Luna but made mistakes for Priya. What most likely caused this?
Be the Example Collector
- You are going to teach a friend or family member using only examples — no explaining allowed!
- Choose a simple sorting rule in your head. For example: things with straight edges go in group A, things with curved edges go in group B.
- Collect 10 household objects and sort them into two piles according to your secret rule.
- Point to each item one at a time. Let your helper guess which group it belongs to.
- After each guess, say yes or no. Do not explain the rule yet.
- Keep going until your helper figures out the rule from the examples alone.
- Then talk about it: was it easy or hard to learn from just examples? What would have helped? How is this like how AI learns?