What Is AI? A Beginner's Guide to Artificial Intelligence

In short: artificial intelligence (AI) is software that does tasks which normally need human thinking - recognising faces, understanding speech, translating languages, answering questions, writing text or code. Most AI today is machine learning: instead of being given rules, the software learns patterns from large numbers of examples. Chatbots such as ChatGPT, Claude and Gemini are one kind of AI, called large language models.

What is AI, in simple words?

AI is a computer program that has learned to do something that used to need a person. Your phone unlocking when it sees your face, an email service moving spam out of your inbox, a map app predicting traffic, a chatbot answering a question - all of these are AI.

A more formal definition comes from the OECD, an intergovernmental economic organisation: an AI system is "a machine-based system that … infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions". The key word is infers: the system works out its answer from what it learned, rather than looking it up in a fixed list of rules.

Where did AI come from?

The term "artificial intelligence" was coined in 1955, in a proposal for a summer research workshop at Dartmouth College. John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon proposed the study on the basis 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". Five years earlier, Alan Turing had asked "Can machines think?" in his 1950 paper Computing Machinery and Intelligence, which proposed what is now called the Turing test.

A few milestones since then:

How does AI work?

Most modern AI learns from examples rather than following rules a programmer wrote. There are two broad approaches:

1. Rule-based AI (the older approach). A person writes instructions: "if the email contains these words, mark it as spam." This works for simple, predictable problems, but breaks down when the rules are too many or too fuzzy to write down - nobody can write rules for recognising every cat photo.
2. Machine learning (the approach behind almost all AI today). The program is shown thousands or millions of examples - emails already marked spam or not spam, photos labelled "cat" or "dog" - and adjusts itself until it gets them right. It can then handle new examples it has never seen.

Inside, many machine-learning systems use neural networks: layers of simple mathematical units, loosely inspired by brain cells, that pass numbers to each other. A network with many layers is called deep learning. The 2015 Nature review describes deep learning as allowing models "composed of multiple processing layers to learn representations of data with multiple levels of abstraction" - in plain terms, early layers notice simple patterns such as edges, and later layers combine them into complex ones such as faces.

For the full picture of how learning works, see machine learning basics.

AI vs machine learning vs deep learning

These terms nest inside each other, like circles within circles:

• Artificial intelligence - the whole field: making computers do tasks that need intelligence.
• Machine learning - one way of doing AI, by learning from data.
• Deep learning - one kind of machine learning, using neural networks with many layers.
• Large language models (LLMs) - one kind of deep learning, trained on text. ChatGPT, Claude and Gemini are built on LLMs; see how LLMs work.

So every LLM is deep learning, all deep learning is machine learning, and all machine learning is AI - but not the other way round.

What are the types of AI?

AI is usually grouped in two ways: by how broad it is, and by what it does.

By how broad it is:

• Narrow AI - built for one kind of task, such as recognising speech or recommending films. All AI in use today is narrow AI, even very capable chatbots.
• Artificial general intelligence (AGI) - a hypothetical system as flexible as a person across almost any task. It does not exist; it is a research goal, and researchers disagree about whether, when and how it might be reached.

By what it does:

• Predictive AI - estimates or classifies: "is this transaction fraud?", "what will demand be next week?"
• Generative AI - creates new content: text (ChatGPT, Claude), images (FLUX, Nano Banana), video, music or code.
• Computer vision - understands images and video; see computer vision explained.
• Speech and audio AI - turns speech into text and back, or holds live voice conversations, as with Gemini 3.8 Live.
• AI agents - systems that carry out multi-step tasks on their own, using tools such as search or code.

Examples of AI in everyday life

• Phone features - face unlock, photo search ("show me pictures of beaches"), voice assistants.
• Email - spam filters and suggested replies.
• Maps and travel - traffic prediction and route planning.
• Shopping and streaming - product and film recommendations.
• Banking - fraud alerts on unusual card payments.
• Translation - instant translation of text, speech and signs through your camera.
• Chatbots and assistants - ChatGPT, Claude, Gemini and others, for writing, questions and code.

What can AI do well, and where does it fall short?

AI is strong at tasks with lots of examples to learn from, and weaker where it must be certain, current or truly original.

It does well at:

• Finding patterns in large amounts of data - more examples than a person could review.
• Language tasks - drafting, summarising, translating and answering questions.
• Recognising images and speech.
• Writing and editing code, with a person checking the result.

Its known limits:

• Making things up. Language models can state false information fluently, a problem called hallucination - see why LLMs make mistakes.
• Knowledge cut-offs. A model only knows what was in its training data, up to a date, unless it can search the web.
• Bias. A system trained on biased data can repeat that bias - covered in our AI ethics guide.
• Confidence without certainty. Models often give an answer even when they should say they do not know.

How to start using AI

The easiest way to learn is to try a free chatbot and ask it about something you know well, so you can judge its answers. Our getting started with AI tools guide walks through the main ones, and the free AI courses for beginners page lists free courses that need no coding, such as the University of Helsinki's Elements of AI.

Words to know

• Artificial intelligence (AI) - software that does tasks needing human-like thinking.
• Machine learning - AI that learns from examples instead of rules.
• Neural network - layers of simple mathematical units that learn from data.
• Deep learning - machine learning with many-layered neural networks.
• Large language model (LLM) - a deep-learning model trained on text, behind chatbots.
• Generative AI - AI that creates text, images, audio, video or code.
• AGI - artificial general intelligence; a hypothetical, human-level general system.
• Hallucination - a confident but false answer from an AI model.