Rules vs Pattern-Finding
Imagine two students learning to add. One student reads the rule: 'To add two numbers, count them all together.' The other student is shown 100 addition problems with answers, and figures out the rule on their own. Both students can add — but they got there very differently. Today we will learn why that difference matters a lot, especially for AI.
Two Ways to Learn
Way 1: Someone tells you the rule. A grown-up says: 'Green means go. Red means stop.' You memorize the rule and follow it. This is great when the rule is clear and does not change. Way 2: You figure out the rule from examples. You watch 50 street crossings. You notice that every time the light is green, cars go. Every time the light is red, cars stop. You figure out the rule yourself from the pattern in the examples. Both ways get you to the rule. But way 2 is more powerful for complicated situations — because sometimes rules are too hard to write down, and it is easier to just show lots of examples. That is how almost all modern AI learns. Nobody writes down a rule for 'what does a dog look like.' Instead, the AI is shown millions of dog photos and figures out the pattern on its own.
A rule is told to you. A pattern is figured out from examples. AI mostly learns by finding patterns in examples — not by being handed a list of rules.
Here is a fun way to see this. Imagine you have never heard the word 'mammal.' But I show you these facts: Dogs have fur and feed their babies with milk. Cats have fur and feed their babies with milk. Whales have no fur but do feed their babies with milk. Robins do not feed their babies with milk. From those examples, can you spot a pattern for what makes an animal a mammal? You might guess: animals that feed their babies with milk are mammals. You figured out the pattern — and you are right! You did not need me to tell you the rule. The examples showed you.
Complete this sentence about how AI learns.
Pattern-finding from examples is powerful, but it has a weakness. If the examples are wrong or unfair, the pattern will be wrong or unfair too — just like we talked about in Lesson 7. That is why the people who train AI work hard to give it good examples. Good examples in, good patterns out. Bad examples in, bad patterns out. The examples shape everything.
Think of something you know how to do really well — like catching a ball, reading faces, or recognizing your friend's laugh. Did someone give you a list of rules, or did you learn from practice and examples over time? Most skills come from examples, just like AI!
A teacher writes on the board: 'Verbs are action words.' How did the students learn about verbs?
An AI is shown 10,000 photos labeled 'hot dog' or 'not hot dog.' After training, it can recognize hot dogs in new photos. How did it learn?
Be an Example-Based Learner
- Sit with a family member. One person picks a secret rule (for example: 'words that start with a vowel').
- The other person asks for examples one at a time: 'Does apple fit your rule?' 'Does banana fit your rule?'
- Keep asking until you can guess the rule.
- Then switch — you pick the secret rule and they ask.
- Afterward, talk about this: How many examples did you need before you were confident? That is how AI feels when it is learning!