AI UNIT 2 • STAGE 1 OF 5
Learning by Example
Teach the machine a pattern, and watch it copy you
The whole idea, in one move
You're about to do the thing this entire unit is about: teach a machine by showing it examples. Not by explaining. Not by giving rules. Just by showing a pattern a few times and letting it copy.
đĒļ You already know this
You learned some things by being shown, watching someone do it, then trying yourself. A machine can only learn that way. It can't be told; it can only be shown.
Invent a pattern and teach it
Make up a simple pattern of your own, give the AI two or three examples, then a new one and let it finish. Try one of these, or invent your own.
Try one
Finish the pattern: sun -> sunny, rain -> rainy, wind -> ? I say a word, you add "-ish". big -> biggish, slow -> slowish, warm -> ? Reverse each word: cat -> tac, dog -> god, bird -> ?You never told it the rule. You only showed examples, and it figured out the pattern and applied it. That is machine learning in miniature.
Push it to a brand-new case
Now give it a new example it hasn't seen, using your same pattern. Does it get it right? When a machine applies a learned pattern to something new, that's called generalizing.
đĄ Key idea
The machine isn't "understanding" your rule. It's matching the shape of your examples. That's powerful, and it's also why it breaks in surprising ways when the new case is different from what it was shown.
Write down what happened
In your Field Notes, record the pattern you taught, whether it generalized correctly, and whether it felt like the machine understood or just copied.
đ Remember this feeling
You just taught a machine with a handful of examples. Real AI is taught the same way, with billions of them. Hold that in mind, because Stage 2 shows what happens when the examples are bad.
Record your experiment:
- 1. What pattern did you teach the AI?
- 2. Did it apply your pattern to a new example correctly?
- 3. In your own words, what does "learning from examples" mean?
Saved automatically.