Lesson 2 of 4 / 7 minute read

Rules versus learning

Traditional software follows instructions a person wrote: if the password matches, open the door. That works well when we can describe the rules exactly.

Some tasks resist rules. Try writing instructions that identify every cat in every photo. Machine learning takes another route: show the computer many labelled examples and let it adjust itself until its guesses match the labels.

The result is called a model. Nobody typed in what a cat looks like. The model worked out useful patterns from the examples it was given, which is powerful and also the source of its mistakes.

Check your understanding

In machine learning, where do the patterns come from?

Choose the answer you think is right.