What's Your IQ

Introduction

Artificial intelligence is the field of building computer systems that perform tasks we'd normally say need intelligence: recognising images and speech, understanding language, making decisions and learning from experience.

Most AI in use today is "narrow": very good at one task, like translating text or spotting tumours in scans, and unable to do anything else. "General" AI that matches humans across tasks remains a goal and a debate, not a product. Much of modern AI is machine learning, where systems learn patterns from data instead of following rules written by hand.

ARTIFICIAL INTELLIGENCE, IN SHORT

AI is software that performs tasks we associate with intelligence. The term dates from 1956. Today almost all practical AI is narrow and built with machine learning, mostly deep neural networks, while human-level "general" AI does not yet exist.

By What's Your IQ · Reviewed 4 Oct 2026 · 6 sources

Term coined: 1955–56, John McCarthy
Founding event: Dartmouth workshop, 1956
Main method today: Deep learning
Major winters: 1974–80 and 1987–93
General AI: Not yet achieved

AI vs machine learning vs deep learning vs generative AI

They're nested, not rivals. Machine learning is one way of doing AI; deep learning is one kind of machine learning; generative AI is one use of deep learning. A chess program from 1997 is AI but not machine learning; ChatGPT is all four.

Artificial intelligenceMachine learningDeep learningGen
AI

What it isExampleAIAny technique making machines act intelligentlyDeep Blue chess, route plannersMachine learningLearning patterns from data, not hand-written rulesSpam filters, credit scoringDeep learningMachine learning with many-layered neural networksFace recognition, speech-to-textGenerative AIDeep models that produce new text, images or audioChatbots, image generators

History: summers and winters

AI has moved in cycles: bursts of optimism and funding ("summers") followed by disappointment and cuts ("winters"). Pick a period.

The birth of AI1956–74At the 1956 Dartmouth workshop, John McCarthy coined 'artificial intelligence'. Early programs proved theorems, played draughts and solved algebra problems, and researchers predicted human-level AI within a generation.

Core concepts: one problem, five approaches

Take one task, filtering spam email, and see how each major approach to AI would tackle it.

1950S–80SSymbolic AI (rules)People write explicit rules and logic. The system follows them exactly.

Good at:transparent, predictable, needs no dataStruggles with:rules you didn't think of; real-world messiness

SPAM FILTER, THIS WAY

IF subject contains 'WIN' AND sender unknown THEN spamIF body contains 'unsubscribe' AND link count > 10 THEN spam→ Spammers write 'W1N' and slip through.

SUBFIELDS AND APPLICATIONS

A day with AI you didn't notice

Most AI is invisible. Step through an ordinary day to see where it shows up and which branch of the field makes it work.

07:00Your phone unlocks when it sees your faceA neural network compares the camera image with a mathematical map of your face, in the dark and with glasses on.

BRANCH OF AI

Computer visionInterpreting images and video: faces, objects, medical scans.

CHALLENGE 1 OF 5 · HALLUCINATION · ROUND 1 OF 3

Spot the confident mistake

Generative AI writes fluent, confident text whether or not it's true. Each round, two statements are correct and one is the kind of thing a model might invent. Find it.

2Bias and fairnessSystems trained on historical data can reproduce and amplify discrimination, for example in hiring or lending.

3ExplainabilityDeep networks are hard to interpret, which matters when decisions affect people's lives.

4Alignment and safetyEnsuring increasingly capable systems pursue the goals we intend, not just the ones we specify.

5Energy and costTraining and running large models takes enormous computing power, money and electricity.

Key terms, defined

Artificial intelligence: The field of building computer systems that perform tasks normally requiring human intelligence.
Narrow AI: A system built for one task or a small range of tasks. Almost all AI today.
Artificial general intelligence (AGI): A hypothetical system matching humans across most cognitive tasks.
Model: The learned set of numbers (parameters) that turns inputs into outputs.
Training data: The examples a machine-learning system learns from.
Neural network: Layers of simple connected units whose connection weights are adjusted during training.
Large language model (LLM): A very large neural network trained to predict text, which can then generate and summarise language.
Hallucination: When a generative model states something false as if it were true.

Sources

McCarthy, J., Minsky, M. L., Rochester, N., & Shannon, C. E. (1955). A proposal for the Dartmouth summer research project on artificial intelligence.
Turing, A. M. (1950). Computing machinery and intelligence. Mind, 59(236), 433–460.
Lighthill, J. (1973). Artificial intelligence: A general survey. Science Research Council.
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). ImageNet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems, 25.
Silver, D., et al. (2016). Mastering the game of Go with deep neural networks and tree search. Nature, 529, 484–489.
Vaswani, A., et al. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30.

Written and fact-checked by the What's Your IQ editorial team against the sources above. Last reviewed 4 October 2026. See our editorial standards.

Artificial intelligence