What is AI? A clear explanation without the jargon

AI in a nutshell: a computer that learned from examples

Artificial intelligence, or AI, is an umbrella term for computer programs that do things that used to need human judgement. They recognise pictures, understand speech, suggest routes or write text.

The key difference from an ordinary program is this: an ordinary program follows rules that a person wrote for it, line by line. An AI has instead learned its task by looking at an enormous number of examples. Nobody wrote it a rule saying "a cat has whiskers and pointy ears". It has seen millions of cat photos and worked out for itself what a cat looks like.

So AI is not a thinking being, and it is not a film robot. It is a clever tool that is good at some things and poor at others.

You already use AI – probably every day

Many people say they have never used AI. In reality, almost everyone with a smartphone uses it daily. Some familiar examples:

• The spam filter. Your email moves scams and junk ads into the spam folder. The filter learned from millions of messages what spam typically looks like.
• Sat nav. When you drive to the countryside and your phone suggests a different route because of traffic, AI is estimating congestion and travel times in the background.
• Face unlock. Your phone unlocks when you look at it, even if you are wearing a hat or new glasses.
• Recommendations. Netflix, Spotify or an online shop suggests your next show, song or product based on what you chose before.
• Speech recognition. When you dictate a message or ask a smart speaker about the weather, your speech is turned into text using AI.
• Bank fraud detection. If your card is suddenly used on the other side of the world, the bank's system may block the payment. That is AI too.

All of these are what is called narrow AI.

Narrow AI: one job, done well

Narrow AI is trained to do a single thing. The spam filter cannot drive a car, and the sat nav cannot recognise faces. Think of it as one tool in a toolbox: a hammer is excellent for nails, but you would not saw a plank with it.

Narrow AI has been around for decades. That is why we rarely talk about it – it has become an invisible part of everyday services.

Generative AI: the newcomer that creates content

In recent years, the news has almost always been about generative AI. "Generative" simply means it generates, or creates: this kind of AI does not just sort or recognise things, it produces something new – text, images, audio or video.

The best known are chat assistants such as ChatGPT, Gemini, Copilot and Claude. You type to them in ordinary language and they reply like a conversation partner. You could ask one to:

• write a polite message to your landlord or building manager
• explain the difficult words in a letter from the tax office in plain English
• plan a week of dinners using what is on offer at your supermarket
• make a checklist for opening up a holiday cottage in spring

Generative AI is far more versatile than narrow AI, but it still does not "know" things the way a person does. Read more in the guide how AI works.

What AI is not

Many misconceptions come from films. So it helps to say a few things out loud:

• AI is not conscious. It does not feel, want or fear anything, even if its words sound warm.
• AI is not always right. It can state something false in a very confident voice. More on this in what AI can't do.
• AI is not one single machine. It is many programs from different companies that work in different ways.
• AI is not just for young people or programmers. If you can write a text message, you can use a chat AI.

Many worries are understandable, and we go through them honestly in AI myths and fears.

Why is everyone talking about AI now?

Narrow AI has been in use for a long time, but a few years ago chat assistants became available to everyone. For the first time, anyone could write to a computer in their own words and get a sensible answer. That turned AI from an engineer's tool into everyone's tool.

Today AI is also built into many familiar programs: phones, email, Word and search engines. That makes it hard to avoid – and worth understanding.

Try it yourself: your first question

The best way to understand AI is to try it. Open a free assistant (instructions in get started with AI in 10 minutes) and copy this question:

You will notice straight away that the answer is written for you. You can carry on the conversation: "Tell me more about the third example".

Do I have to learn it all at once?

Not at all. Learning AI is more like getting used to a new phone than taking a course. Start with one small task that is genuinely useful to you, such as asking for help wording an email or asking for a recipe using leftovers in the fridge. Once you see what AI can do and where it stumbles, new uses will come to you naturally. If you do not like an answer, say so – the AI will not be offended.

Summary

AI is a program that has learned its task from examples. Narrow AI does one thing well, and you already use it. Generative AI creates text and images and acts as a conversation partner you can ask almost anything. It is a useful helper, as long as you remember to check its answers and keep your personal details to yourself.