What is AI?

Two ways to make a computer clever

A computer only does what it's told. For most of history, people told it with rules: "If the water reaches 100 degrees, switch the kettle off." "If the password is right, open the account." A programmer thinks of every rule and writes it down. That's a normal program.

But some jobs are almost impossible to write rules for. Try writing rules that recognise a cat in a photo. Pointy ears? So do foxes. Whiskers? So do seals. Four legs? So does a table. You'd never finish.

Artificial intelligence (AI) is a different way of making a computer clever. Instead of giving it rules, you give it lots of examples — thousands of photos labelled "cat" and "not a cat" — and let it work out the patterns for itself. That's called machine learning, and it's how most AI you hear about today works.

Where you already meet AI

• A phone that unlocks when it sees your face.
• A video app that suggests what to watch next.
• A map app that predicts how long the journey will take in traffic.
• A voice assistant that understands what you say.
• A translation app that turns Hindi into English.
• A chatbot that answers questions in full sentences.

None of these follow a simple list of rules written by a person. They all learned from huge numbers of examples.

What AI is not

AI isn't a person or a brain, and it doesn't *think* or *feel* the way you do. It's a computer program that has become very good at spotting patterns. That makes it brilliant at some things and surprisingly silly at others — you'll see both in this course.

Try it

One of these follows a simple rule written by a person. The others learned from examples. Tap the one that is not AI.

Away from the screen

Ask someone at home to describe a cat using only rules a computer could check. Then think of an animal that breaks every rule. You've just discovered why AI learns from examples instead.