Watch the idea
The concept
Examples
Labeled pictures
Model
Tiny numbers get nudged
Guess
cat or dog?
A computer has never met a cat
A machine cannot smell, pet, or love an animal. When we ask it to tell cats from dogs, we show it many labeled examples: this is a cat, this is a dog. It does not learn the meaning of a cat. It learns patterns — ears, whiskers, shape — that often sit next to the word cat.
Tiny numbers, lots of guesses
Inside the model are millions of tiny numbers. Each example nudges those numbers a little. After enough nudges, the next guess is better. That loop — example, guess, nudge — is called training. It is practice, not understanding.
Narrow examples make a narrow student
If every training cat is orange, a black cat looks strange. The model is not mean. It is unfinished. Teachers who care show many kinds of examples so the guess stays fair when the world changes.
Worked example
The tomato guest
If a kitchen only ever cooks red tomatoes, a yellow tomato looks 'wrong' to a picky guest. The model is that guest. Show red, yellow, small, and large — and the guess gets kinder.
Picture to keep
Data → Model → Guess
- 1Examples in
- 2Pattern machine
- 3Guess out