Guess, test, learn

This is how both scientists and computers learn:

1. Guess — «I think the big ball bounces highest.»
2. Test — drop the balls from the same height and watch.
3. Learn — was it right? If not: a new guess.

There is nothing embarrassing about guessing wrong. Wrong guesses are how you learn. A machine being trained guesses wrong thousands of times — and gets better with every one.

Try it with a sheet of paper: guess how many times you can fold an A4 sheet in half. Write the guess down. Test it. Most people manage 6 or 7, not 10.

Now: why? A new guess: «it gets too thick». How would you test that? (Try a thinner sheet, or a bigger one.)

Write it down: guess → test → what happened → new guess.