Showing, Not Telling
In the last lesson we learned that machines — just like you — can learn from examples. Now let's dig into WHY we teach that way. Here is a puzzle: if you had to write down every single rule for recognizing a dog, how many rules would you need? Let's find out just how tricky that question is.
Why Rules Are So Hard to Write
Imagine you want to teach a robot to recognize dogs. You decide to write rules. Rule 1: Dogs have four legs. But wait — cats have four legs too. And tables. Hmm. Rule 2: Dogs bark. But puppies sometimes squeak. And a barking seal — is that a dog? Rule 3: Dogs have fur. But some dog breeds have very little fur. And polar bears have fur too. You keep adding rules, but every rule has exceptions. You could write a thousand rules and STILL miss some dogs or accidentally include some non-dogs. It gets very complicated very fast. This is the problem researchers ran into when they first tried to build AI programs. Writing all the rules by hand just did not work well enough.
Instead of writing rules yourself, you SHOW the machine many labeled examples, and the machine figures out the rules on its own. Showing works far better than telling.
Here is a story that shows the difference. Two friends, Mira and Leo, both want to teach their little brother what a "comfortable chair" looks like. Mira writes him a list: "Must have a soft seat. Must have a back. Must not wobble." Her brother reads the list, then looks at a rocking chair and says "it wobbles, so it is NOT comfortable" — but actually rocking chairs are very cozy! Leo takes his brother around the house and says "comfortable chair" every time they find one — the armchair, the rocking chair, the beanbag, the reading nook cushion. His brother sees the pattern and gets it right on the first try with a new chair he has never seen before. Leo used showing. Mira used telling. Leo's brother learned better.
Match each approach to what it means.
Terms
Definitions
Drag terms onto their definitions, or click a term then click a definition to match.
When you show a machine labeled examples, the machine does something amazing: it looks for patterns across ALL the examples at once. It might notice things you never even thought to write down as a rule. That is why showing is so much more powerful than telling. Your brain — and the machine's pattern-finder — is better at spotting patterns than any list of rules you could write.
Pick any animal and try to write THREE rules that describe ONLY that animal and nothing else. It is harder than it looks! This shows you exactly why we teach machines with examples instead.
Why is it hard to write rules to recognize a dog?
When you show a machine many labeled examples, what does it do?
The Exception Hunt
- Think of a simple rule about an animal you like — for example, 'all birds can fly.'
- Now try to find ONE exception — an animal that breaks your rule (like a penguin).
- Keep going: can you find a rule with NO exceptions at all? (Hint: it is very hard!)
- Talk about what you discovered with a family member or friend.
- Explain why this shows that SHOWING examples works better than WRITING rules.