Skip to content

1 · Machines learn from examples (not magic)

When a phone unlocks by recognizing your face, or an app guesses what's in a photo, it can feel like magic. It isn't. Here's the big idea behind this whole course, and we'll keep coming back to it:

Machines learn from examples — not magic, and not rules someone typed in.

Think about how you learned to tell a cat from a dog. Nobody handed you a rulebook that said "a cat is exactly 9 inches tall with these exact whiskers." You just saw a bunch of cats and a bunch of dogs, someone told you which was which, and after enough examples your brain figured out the pattern on its own. Now you can spot a cat you've never seen before.

Teaching a computer this way is called machine learning (often shortened to ML). Instead of a programmer writing step-by-step rules, you show the computer lots of examples and let it find the pattern itself. The finished pattern-finder it builds is called a model — that's the thing you "train," and that's what this course is named after.

Here's the difference, side by side:

Old way: rulesML way: examples
A human writes every rule by handA human collects examples; the computer finds the rules
"IF pointy ears AND whiskers THEN cat"Show it 200 cat photos and 200 dog photos, labeled
Breaks on anything the rules didn't coverCan handle new, never-seen examples — by pattern

One thing to keep straight: the model is making a guess, a really educated guess based on patterns it saw — not looking up a fact in a book. That's why a model can be confident and wrong, something we'll dig into later.

The best news: you don't need to be a programmer to do this. Free school programs from Code.org and MIT's Day of AI teach exactly this idea, and by the end of this course you will have trained a real model yourself.

Think about it. Describe something you learned from examples instead of from a rulebook — like recognizing a friend's voice, or knowing when a song is "your kind of music." How is that like machine learning?

Sources