Menu
Concepts

Artificial intelligence

5 min readbeginnerUpdated 28 Sept 2026
1 · In one line

An AI system infers from received inputs how to generate predictions, content, recommendations or decisions for explicit or implicit objectives.

1 · What it is

Artificial intelligence covers machine-based systems that receive inputs and infer how to generate outputs for explicit or implicit objectives. Those outputs can be predictions, content, recommendations or decisions.

The field has been broad from its beginning; the 1955 Dartmouth proposal named language, abstraction, problem solving and learning as AI problems. An AI system computes an output by processing input through one or more models and underlying algorithms. Machine learning trains a model from data to make predictions or generate content. Generative AI is a class of models that creates content from user input.

AI risks and benefits can emerge from the interplay of technical aspects and societal factors. Measurements can differ from risks that emerge in operational, real-world settings.

2 · Why it exists

Machine learning trains a model with data to make predictions or generate content.

Too many casesA model can learn a mathematical relationship from a large collection of examples.
Uncertain inputsGenerative AI can take text, images, audio and video as inputs.
Patterns at scaleMachine learning trains a model with data to make predictions or generate content.
3 · How it works

An AI system turns a goal and inputs into an output.

  1. 1 · objectiveThe system has explicit or implicit objectives.
  2. 2 · inputThe system receives input from people or machines. It can include text, images, audio or video.
  3. 3 · inferOne or more models and underlying algorithms process the input to work out an output.
  4. 4 · outputThe system returns a prediction, piece of content, recommendation or decision.
  5. 5 · evaluateEvaluation asks whether the system is fit for purpose.

During inference, the system computes an output from its inputs.

4 · Where it's used
WhoWhat they askWhat it works with
Hypothetical imaging team“Which scans need urgent review?”Medical images and clinical context
Translator“How should this sentence read in another language?”Source text and learned language patterns
Driver-assistance system“What is around the vehicle right now?”Camera, radar and map signals
Knowledge worker“Can you draft and revise this report?”Instructions, documents and model context
5 · What it solves, and what it doesn't
solves
  • Classifying an input into a category.
  • Generating text, images, audio or video from user input.
  • Producing recommendations from supplied inputs.
  • Making predictions from relationships learned in data.
doesn't solve
  • Generative AI can confidently present false content as if it were correct.
  • AI design, development, deployment and use can reflect systemic and human cognitive biases.
  • Risk measurements in a laboratory may differ from risks in operational, real-world settings.
  • Trustworthy AI should respect human rights and democratic values.
6 · Go deeper
Machine learningDeep learningGenerative AILLMAI agents

Sources used

This explainer is written in original language. The links below support its factual claims.

  1. officialUpdates to the OECD’s definition of an AI system explained, OECD.AI · read 28 Sept 2026
  2. officialArtificial Intelligence Risk Management Framework (AI RMF 1.0), National Institute of Standards and Technology · read 28 Sept 2026
  3. officialWhat is Machine Learning?, Google for Developers · read 28 Sept 2026
  4. officialArtificial Intelligence Risk Management Framework — Generative Artificial Intelligence Profile, National Institute of Standards and Technology · read 28 Sept 2026
  5. paperA Proposal for the Dartmouth Summer Research Project on Artificial Intelligence, John McCarthy et al. · read 28 Sept 2026
  6. officialOECD AI Principles, OECD.AI · read 28 Sept 2026
Suggest a correction