What Is Artificial Intelligence? A Plain-Language Explanation

How AI works, in three steps

A traditional program does exactly what a programmer wrote, line by line. A machine learning system, today's most common form of AI, works the other way round: you start from many examples and the system works out the rules by itself.

• Data: you collect many examples, such as emails already labelled as spam or not spam, photos of cats and dogs, translated sentences.
• Training: a mathematical model compares the examples, finds regularities and corrects itself until it makes few mistakes.
• Use: faced with new data it has never seen, the model applies what it learned and produces a prediction, a label or a piece of text.

That is why AI is, to a large extent, statistics at scale: it does not "understand" the way a person does, but it is very good at recognising patterns.

Where you already use AI every day

• Your email spam filter.
• Film, music and shopping recommendations on the platforms you use.
• Machine translation and automatic subtitles.
• Unlocking your phone with your face.
• Navigation apps that estimate traffic.
• Assistants such as ChatGPT, Gemini or Claude, which write, summarise and answer questions.

Narrow AI and general AI: what actually exists

Almost all the AI in use today is "narrow": very good at one specific task and useless outside it. A chess engine that beats champions cannot translate a sentence. Large language models are more versatile and handle many different tasks, but they are still far from reliable general intelligence. "General" AI, able to tackle any intellectual task the way a human can, is for now a research goal rather than an available product.

What AI can get wrong

• Errors and "hallucinations": generative models can produce answers that sound plausible but are false.
• Bias: if the training data is unbalanced, so are the answers.
• Lack of context: the system recognises patterns in data; it does not know whether something is true.
• Opacity: it is often hard to explain why it gave a particular answer.

That is why, when AI enters a company's processes, human oversight and clear rules about who decides what matter.

What European law says

The EU AI Act, the European regulation on artificial intelligence, defines an "AI system", in short, as a machine-based system that operates with varying levels of autonomy and, from the input it receives, infers how to generate outputs (predictions, content, recommendations or decisions) that can influence physical or virtual environments. The definition is deliberately broad: the rules change with the risk level of the use, not with the technology.

Why it is worth understanding

You do not need to code to use AI well: you need to know what it does well, where it fails and who answers for its decisions. For organisations the question is where to use it, who oversees it and under what rules. That is the subject of my book, The Modern AI Organisation.