B1 · Introduction to Artificial Intelligence

What Is Artificial Intelligence?

Artificial Intelligence (AI) is the field of computer science that builds systems able to do tasks which normally need human intelligence, such as understanding language, recognising images, making decisions and learning from experience.

12 min Beginner Jump to the activity

What you need: Nothing. This page starts from the beginning, with no programming and no formulas.

Definition

Your phone unlocks when it sees your face. Is that intelligence?

Artificial Intelligence (AI) is the field of computer science that builds systems able to do tasks which normally need human intelligence, such as understanding language, recognising images, making decisions and learning from experience.

In simple words

A normal program follows exact steps a person wrote. An AI system handles situations nobody wrote steps for. A calculator does exactly what its rules say. A spam filter looks at an email it has never seen before and still decides whether it's spam. That ability to deal with something new is the heart of AI.

Why it matters

AI now decides which posts you see, flags fraud on UPI payments, types what you speak, and helps doctors read scans. To use it well, build it, or just not be fooled by it, you first need to know what it is and what it isn't. Almost every topic in this course depends on this one.

Key terms

Artificial intelligence (AI)
The field of computer science that builds systems able to do tasks that normally need human intelligence.
AI system
A machine or program that takes in information, works out what to do, and produces an output such as a prediction, a decision or new content.
Rule-based AI
AI built from rules written by experts, such as "if the fever is high and the rash is red, suggest a test".
Machine learning
The part of AI in which systems learn patterns from data instead of being given every rule.
Generalisation
Doing well on new cases that were not in the examples a system learned from.
The AI effect
The habit of no longer calling something AI once it works well and becomes ordinary.

Explained in detail

Where the name came from

The term "artificial intelligence" was first used in 1955 by the computer scientist John McCarthy, in a proposal for a summer workshop at Dartmouth College in the United States. The proposal stated a bold idea: that every part of learning and intelligence could, in principle, be described so precisely that a machine could be made to imitate it. That question, whether and how machines can do what minds do, has driven the field ever since.

AI is a field, not a single machine

AI is a branch of computer science, in the same way that medicine is a field with many specialities. Inside it are several areas: machine learning (learning from data), computer vision (understanding images), natural language processing (working with human language), speech (listening and talking), robotics (sensing and moving), and planning and reasoning (working out steps towards a goal). When people say "an AI", they usually mean one system built using one or more of these areas.

Two ways to build AI

In the first decades, most AI was rule-based. Experts sat with programmers and wrote down their knowledge as rules. A medical system might hold hundreds of rules about symptoms and diseases. Such systems were clear and easy to check, but they broke as soon as a case did not fit the rules, and they could not learn. Today most AI uses machine learning instead: the system is shown many examples and finds the patterns itself. This handles messy real-world cases far better, but it needs a lot of data and its reasoning is harder to inspect. Many real products combine the two.

The test of a new case

A useful way to recognise AI is to ask: can the system produce a sensible output for a case nobody wrote instructions for? A calculator cannot; it only does what its rules say. A translation app can translate a sentence it has never seen before. A spam filter can catch a brand-new scam. This ability to generalise from past examples to new situations is what most modern AI is built for, and it is also where most of its mistakes happen.

Why the meaning of "AI" keeps moving

When computers first played chess, recognised handwriting or found routes on a map, people called it AI. Once these things became ordinary, many stopped calling them AI and saved the word for whatever was new and surprising. This is known as the AI effect. It is worth remembering, because much of the AI that affects your life is the quiet, ordinary kind: spam filters, map routes, fraud checks and recommendations, not only chatbots.

How it works

What makes a system count as AI?

There is no single test, but most AI systems share three abilities. They take in information from the world, they work out what it means, and they act on it to reach a goal. Many also improve with experience.

AbilityWhat it meansExample
PerceiveTake in images, sound or textA camera spots a pedestrian
ReasonWork out what follows from what it knowsMaps finds the fastest route
LearnGet better from data or experienceA spam filter improves as users mark emails
ActDo something that affects the worldA robot picks up a box

What is not AI?

Fixed automation isn't AI. An alarm clock, a washing machine programme and a plain if-else rule all do only what was written in advance. Software isn't AI just because a company's advert says so, either. Ask one question: can it handle a case that nobody wrote a rule for? If not, it's probably ordinary software.

Two ways to build AI

  • Write the knowledge by hand: experts turn what they know into rules. This is symbolic AI, and it powered the expert systems of the 1980s.
  • Learn from data: show the system many examples and let it find the patterns. This is machine learning, and it powers nearly all of today's AI, including ChatGPT.

Try it yourself

A calculator and a spam filter meet something new

Watch what each one does with a case nobody planned for.

AI or not AI?

For each product, decide whether it uses AI. Remember the test: can it handle a case nobody wrote a rule for?

  • Face unlock on a phone

  • A maps app that reroutes around a traffic jam

  • A spam filter that catches new kinds of spam

  • A chatbot that answers your questions

  • A digital clock

  • A ceiling fan with a regulator

  • A washing machine on a fixed programme

  • A calculator

Common mistakes

  • Thinking AI means robots. Most AI is software with no body at all.
  • Thinking AI understands like a person. Today's systems find patterns. They have no proven understanding or feelings.
  • Using AI and machine learning as if they mean the same thing. ML is one way of doing AI.
  • Trusting an answer because it sounds confident. AI can be confidently wrong.

Check yourself

Question 1

1. Which of these is the best example of AI?

Question 2

2. How is nearly all of today's AI built?

Question 3

3. A chess engine beats every human. What does that show?

Summary

  • AI builds systems that do tasks normally needing human intelligence.
  • Key abilities: perceive, reason, learn and act.
  • Fixed rules aren't AI. The test is whether it can handle cases nobody wrote a rule for.
  • Today's AI is narrow, and mostly learns from data.

Finished reading?

Sources: Russell & Norvig, Artificial Intelligence: A Modern Approach