A simple definition
Artificial intelligence (AI) is software that does tasks we used to think needed a person's judgement: recognising a face, understanding speech, translating a sentence, writing a reply.
The important word is learned. Most AI is not told the answer step by step. It is shown many examples and finds the pattern itself.
Rules versus learning
Think of two ways to sort mobile-money messages into "real" and "scam".
A rule-based program. A person writes every rule: if the message asks for your PIN, mark it as a scam. It does exactly what it was told, nothing more. When scammers change their wording, the rules miss them.
An AI model. You show it 50,000 messages that people have already labelled "real" or "scam". It learns which words, links and patterns tend to come with scams, including ones nobody thought to write down. Shown a new message, it gives its best guess.
Ordinary software follows rules a person wrote. AI learns its own rules from examples.
AI you already use
- The keyboard on your phone guessing your next word.
- Maps finding a faster road because of traffic.
- A bank blocking a strange payment on your card.
- Photos sorting pictures by face.
- Chat assistants such as Flash's own AI chat answering questions in plain language.
What AI is not
AI is not a mind. It does not know things the way you do, it has no feelings, and it does not want anything. It is a very good pattern-finder. That is why it can be brilliant at one task and make a silly mistake at the next.
Guided practice
A calculator app adds 245 + 380. Is that AI?
No. A person wrote the exact steps for adding, and the app follows them the same way every time. Nothing was learned from examples. A camera app that recognises your friend's face is AI: nobody wrote a rule for every face; it learned from many examples.