Rocket is teaching his young neighbor to sort recycling at the station. First he reads the rules from the sign.
The neighbor still mixes up foil and plastic. So Rocket shows her twenty labeled examples instead.
Raven watches. "You switched from rules to examples halfway through."
"People learn both ways," Nova says. "But I should be honest about one thing."
"I am a story character. A real AI is a tool people build. It finds patterns, but it does not think the way you do."
Rocket nods and holds up another example.
People often learn from rules, like "look both ways before you cross."
People also learn from examples, like recognizing a friend's handwriting.
Computers can follow rules or find patterns in data. But AI does not think the way a person does, and it can be wrong.
| Week review | True or false? |
|---|---|
| Machine learning is a kind of statistical inference. | ? |
| Hand-written rules are easy to trace and explain. | ? |
| Learning from data needs only a handful of examples. | ? |
| A real AI thinks and feels like a person. | ? |
| To compare methods fairly, test them on the same cards. | ? |
Next week the crew does it properly: a big labeled training set, and a separate test set held back.
Wonderful week, explorer! Tomorrow your family finds a cut-off from data.