Rocket fixes the tree. He adds a rule: if it is shiny, then tool.
The metal ruler lands in tool. "Fixed!" he cheers. Then a shiny wet rock lands in tool too.
Raven holds it up. "What do you notice? Rocks can be shiny too!"
Nova circles the cards. "Rules are one way," she hints. "Is there another way to teach a helper?"
There are two big ways to make a computer helper.
Way 1: A programmer writes exact rules. The computer follows them.
Way 2: The computer learns from examples. It finds patterns in data.
Finding patterns in data is called machine learning.
| Way 1: Rules | Way 2: Learning from examples | |
|---|---|---|
| Who does the main work? | A programmer writes the rules | The computer finds patterns |
| What does it need? | Clear rules for every case | Lots of examples with answers |
| Sorting photos | If it has a handle, then tool | Look at many labeled tool photos |
Rocket's new rule only looked at one feature: shiny.
The metal ruler is shiny. But a wet rock can be shiny too.
One simple rule could not tell them apart.
Writing rules for every tricky case is hard.
| Statement | True or false? |
|---|---|
| A programmer can write exact rules for a helper. | ? |
| A computer can learn patterns from examples. | ? |
| Rocket's shiny rule worked for every photo. | ? |
| Machine learning means finding patterns in data. | ? |
Great thinking! Tomorrow we write and test if-then rules like a programmer.