What AI Actually Is

The phrase "Artificial Intelligence" gets used for everything from voice assistants to sci-fi robots. So what does it actually mean — and what are its boundaries?

A working definition

Artificial Intelligence refers to computer systems that perform tasks normally associated with human intelligence: recognising speech, making decisions, translating language, identifying objects in images, generating text. What makes AI distinct from regular software is that it learns from examples rather than following hard-coded rules.

Narrow AI vs. general AI

Every AI system in existence today is narrow AI — it is trained to do one specific type of task and cannot transfer that skill elsewhere. The AI that beats the world champion at chess cannot describe what chess looks like. The AI that translates French to English cannot recognise your face.

General AI — a system that could learn and do anything a human can — does not exist yet and may be decades away (if ever). When people talk about "AI taking over", they are usually imagining general AI. What we actually have is a collection of very powerful but very narrow tools.

Why now?

AI has actually existed as a field since the 1950s. What changed around 2012 and accelerated dramatically from 2022 is the combination of three things: vastly more computing power, vastly more data to train on, and a new type of model architecture called deep learning. These three together unlocked capabilities that seemed impossible just years before.

AI is not one thing — it is a collection of different technologies that all fall under the same broad label. Understanding which type of AI is being discussed is usually the first step to understanding any claim about what it can or cannot do.