Page language: Italiano English Français Deutsch

What is artificial intelligence and how does it work? A simple explanation

By the FAIND newsroom · Updated on 17 October 2026

Abstract network of glowing connected dots on a dark blue background, denser in the center

In brief: artificial intelligence (AI) is a set of programs that carry out tasks that normally require human intelligence, such as understanding a sentence, recognising a face or writing a text. It does not follow hand-written rules: it learns from huge numbers of examples.

In brief

AI learns from examples rather than from hand-written rules.
ChatGPT, Gemini and Claude are language models: they predict the most likely next word.
It is powerful but does not understand like a person: it can be confidently wrong.

What artificial intelligence is, in simple words

It is a computer's ability to do things we consider intelligent: recognise images, understand and produce language, make decisions, make predictions. The name dates back to 1956, to a conference at Dartmouth College in the United States. For decades it stayed in the lab; since 2022, with ChatGPT, it has entered everyone's life.

How it works: it learns from examples

A traditional program follows instructions written by a programmer. An AI system is trained instead: it is shown millions of examples and adjusts billions of internal numbers, its parameters, until its answers get close to the right ones. To recognise a cat, nobody explains what a cat is: it is shown a great many photos of cats. This method is called machine learning; its most powerful form, based on many-layered neural networks, is deep learning.

How ChatGPT, Gemini and Claude work

They are large language models: they have read a huge share of the available text and learned to predict, word by word, the most likely continuation. After that they are refined with example conversations and with the judgement of people who rate the answers. The result writes fluently about almost anything, but it does not consult an archive of facts, which is why it sometimes makes things up. More in is AI reliable?

Where you already use it without knowing

In film and music recommendations, spam filters, face unlock on your phone, automatic translation, maps that estimate traffic and cameras that improve photos. Generative AI, which creates text, images and video, is only the newest and most visible part.

Narrow AI and general AI

Every system that exists today is specialised: excellent at some tasks, useless outside them. The idea of an AI able to do everything a person can is called artificial general intelligence (AGI): some companies say they are aiming for it, many researchers doubt it is close. Either way, today's AI has no consciousness or intentions.

What it can and cannot do

It can summarise, translate, write, program, analyse data and recognise images and voices. It cannot check by itself whether what it says is true, it has no experience of the world and it reflects the flaws of the data it learned from. To get started there is the guide how to use artificial intelligence; for pros and cons, risks and benefits.

Frequently asked questions

What is artificial intelligence in simple words?

Programs that learn from many examples to carry out tasks typical of human intelligence, such as understanding language, recognising images or making decisions.

How does artificial intelligence work?

It is trained on large amounts of data, adjusting billions of parameters until its answers get close to the correct ones. Chatbots such as ChatGPT predict the most likely next word.

Who invented artificial intelligence?

There is no single inventor. The name was proposed by John McCarthy for a conference at Dartmouth College in 1956; Alan Turing had laid the theoretical foundations in 1950.

Sources

J. McCarthy et al., Dartmouth Summer Research Project on Artificial Intelligence proposal (1955)
A. M. Turing, “Computing Machinery and Intelligence” (1950)