What is AI?

Rules written by people, or rules learned from data

Artificial intelligence is an umbrella term for software that does things we would call intelligent if a person did them: recognizing a face, translating a sentence, choosing a move in a game. The term describes what a system does, not how it does it.

There are two broad ways to build one. You can write the rules yourself — if the subject line says free money, mark it as spam — or you can collect examples and let the system find the rules. The first is predictable and brittle: it fails on the first case nobody thought to write down. The second copes with cases nobody anticipated, and cannot fully explain itself.

When people say AI today they almost always mean the second kind, and usually one family of it: large models trained on huge amounts of data. Everything that follows is about that family. Keep the older meaning in mind anyway, because plenty of systems sold as AI are still mostly hand-written rules.