AI is an umbrella term. It includes older rule-based systems written by hand and, far more commonly today, machine learning, where systems learn patterns from data. Generative AI, which creates text, images, audio and video, is the most visible branch, but much AI works quietly inside search, maps, banking and phones.

How it works

The term was coined for a 1956 workshop at Dartmouth College. Early AI relied on hand-written rules, which worked in narrow settings but broke on messy real-world input. Modern AI mostly uses machine learning: instead of writing rules, engineers give a model many examples and a learning algorithm adjusts the model's internal numbers, called parameters, until its predictions improve. Deep learning uses neural networks with many layers and became dominant in the 2010s, helped by large datasets and GPUs. The transformer architecture, introduced in 2017, underpins today's large language models, which learn by predicting the next token across huge amounts of text. A finished model is then used for inference: it takes new input and produces an output, such as a label, a forecast or a paragraph. Most AI systems are narrow, built or tuned for particular kinds of tasks, and their quality depends heavily on the data they learned from and how they are tested.

Why it matters

AI now sits inside everyday tools, from spam filters and photo search to writing assistants and video editors. For creators, it speeds up editing, captioning and ideation. For professionals, it drafts, summarises and analyses. For businesses, it can cut routine work and personalise service. Knowing what AI is, and what it is not, helps you pick tools that fit, check their output and explain your choices to clients, students or regulators.

Where you will see it

  • Filtering spam and phishing emails
  • Generating captions and transcripts for videos
  • Detecting suspicious bank transactions in real time
  • Recommending products, songs or videos based on past behaviour

Limits and common misconceptions

  • AI does not understand the world as people do; it can produce fluent but wrong answers.
  • Systems inherit biases and gaps from their training data.
  • 'AI' is used loosely in marketing, so a product labelled AI may use simple automation.

How it differs from related terms

Tools that use it

Frequently asked questions

What are the main types of AI?

A common split is by capability: narrow AI, which handles specific tasks and covers everything in use today, and the hypothetical general AI. Another split is by method: rule-based systems, machine learning, deep learning and generative AI, each a more specific layer within the one before, apart from rule-based systems.

Is ChatGPT an example of AI?

Yes. ChatGPT is a chatbot powered by large language models, a type of generative AI built with deep learning. It is one of many forms of AI; others include recommendation systems, speech recognition and computer vision tools.

Will AI replace jobs?

AI is more likely to change jobs than remove them outright in most fields, by automating particular tasks such as drafting, sorting or transcribing. Some roles shrink, others grow and new ones appear. People who learn to use AI tools well tend to gain an advantage in their work.

Related terms

All AI terms