Definitions of Artificial Intelligence: The Four Approaches
Four ways to define a clever machine — and the famous test that tries to settle the argument.
You open your bank app one sleepy morning. It looks at your face, decides it really is you, flags a strange ₦80,000 transfer from last night as "suspicious", and a chatbot cheerfully offers to block your card. Three decisions in ten seconds, and no human was involved in any of them.
Now ask yourself: is that app intelligent? Is your calculator? Is a ceiling fan with a remote? By the end of this chapter you'll have a precise answer, and a clear reason for it.
The phrase "Artificial Intelligence" was coined by John McCarthy for a summer workshop at Dartmouth College in 1956. The organisers wrote, with lovely optimism, that "every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it." Seventy years later we are still working on the "precisely described" part.
A good working definition: Artificial Intelligence is the study and design of computer systems that perform tasks which normally require human intelligence — perceiving, reasoning, learning, planning and acting to achieve goals.
But researchers never fully agreed on what the target should be. Should a machine think like us, or just behave well? Should it copy humans (who are brilliant but make mistakes), or follow an ideal standard of rationality? Cross those two questions and you get the famous four approaches.
| Approach | The question it asks | Famous idea | The catch |
|---|---|---|---|
| Thinking humanly | Does the machine think the way a human mind does? | Cognitive modelling — compare program traces with human thought (cognitive science) | We don't fully know how humans think, so it's hard to check |
| Acting humanly | Can it behave so well that you can't tell it from a person? | The Turing Test (1950) | Fooling people is not the same as understanding |
| Thinking rationally | Does it reason by the "laws of thought"? | Logic — Aristotle's syllogisms, formal proofs | Much real knowledge is uncertain and can't be written as pure logic |
| Acting rationally | Does it do the right thing given what it knows? | The rational agent — the approach most modern AI uses | Perfect rationality is often too expensive to compute |
Pause & think
A chatbot passes the Turing Test. Has it proven that it understands what it is saying?
Spotted a mistake?