Artificial Intelligence (Years 9–11) — Computer Science

How machine learning lets computers learn from examples, why all of today's AI is narrow, and the real uses and risks of AI in everyday life.

Ask your phone for the weather and it understands you. Open a video app and it queues up exactly the clip that keeps you watching. Point a camera at a friend and a box snaps around their face. None of that was hand-coded rule by rule — it was learned. This is Artificial Intelligence: machines doing tasks that would normally need human intelligence, and it is already woven through the day you are living right now.

Why this matters: AI now decides what you see, filters your messages, screens job applications and even helps doctors read scans. Knowing how it works — and where it goes wrong — is the difference between using it wisely and being quietly steered by it.

What AI actually is

Artificial Intelligence (AI) means getting a computer to do things we think of as needing a human brain: recognising a face, understanding a sentence, spotting a tumour, choosing a chess move. It is not a robot with feelings. It is clever software that produces sensible-looking outputs for problems that are far too messy to solve with a fixed list of rules.

Machine learning: learning from examples

Old-fashioned programs are told exactly what to do, step by step. But nobody can write down every rule for "what a cat looks like". So machine learning flips it around: instead of rules, you feed the computer thousands of labelled examples — this picture is a cat, this one is a dog — and it works out the pattern for itself. The examples it learns from are called the training data.

Traditional program	Machine learning
Human writes the rules	Computer finds the rules
Follows fixed steps	Learns from many examples
Great for clear, exact tasks	Great for messy, fuzzy tasks

A neural network: inputs pass through weighted connections and hidden nodes to an output.

The idea of a neural network

Many modern AIs use a neural network, loosely inspired by brain cells. Numbers go in at the input nodes, travel along connections that each carry a weight (an importance value), get combined in hidden nodes, and produce an output — say, "87% cat". Training simply nudges those weights up and down until the outputs match the correct answers.

The training loop: predict, compare with the correct answer, adjust, repeat.

Narrow AI vs General AI

Every AI in use today is narrow: brilliant at one job and helpless at anything else. A chess AI cannot recognise your face; a face app cannot play chess. General AI — one system as flexible as a human across any task — does not exist. It lives only in films and research dreams.

	Narrow AI	General AI
Range	One task only	Any task, like a human
Exists?	Yes — everywhere	No — not yet
Example	Spam filter, voice assistant	Science-fiction robot mind

All of today's AI sits on the left: powerful, but narrow.

Where AI turns up

You meet narrow AI constantly: recommendations on video and shopping apps; face and voice recognition; chatbots and large language models that write and answer in plain English; medical image diagnosis that flags disease on scans; and self-driving cars reading the road. Different jobs, same core trick — learn from data, then predict.

Worked example — how a spam filter learns: You give it thousands of emails, each labelled spam or not spam. It measures patterns — words like "free prize", odd links, strange senders. It builds a model linking those patterns to "spam". When a new email arrives, it scores the patterns and predicts. Guessed wrong? You mark it, that becomes a fresh example, and next time it does better. It was never told a single rule — it learned them all from your labels.

Bias in the training data becomes bias in the decisions.

Risks and ethics

AI learns from us, so it inherits our flaws. Biased training data produces unfair decisions — a hiring AI trained on mostly one group may reject others. Deepfakes fake voices and faces to deceive. AI can replace jobs, and over-reliance lets skills fade. Above all, remember: AI does not truly understand or feel anything. It matches patterns — which is why it can be confidently, fluently wrong.

Common mistake: Thinking AI is conscious or magic. It has no beliefs and no common sense — it copies patterns from its data. Garbage in, garbage out: feed it biased or wrong examples and it will repeat them, sounding certain the whole time.

Tip: remember AI in one line — learn from examples, then predict. Today it is always narrow, only as fair as its data, and never actually understands what it is doing.