AI

The agentic era is the phase of AI in which systems take actions as well as answering questions, planning steps and using tools with limited human direction. Tech companies began using the phrase in late 2024, as AI agents moved

An AI classifier is a machine learning model that sorts inputs into set categories, such as spam or not spam, or positive or negative. It learns from labelled examples, then applies the patterns it found to new items, from emails

The AI development lifecycle is the series of stages from defining the problem and gathering data to training, testing and deploying a model. It is a loop rather than a line: once in use, a model keeps being monitored, retrained

An AI model is a system trained on data to recognise patterns and use them to make predictions, decisions or new content. It is what sits inside an AI tool: the chatbot, image generator or recommendation feed is the product,

AlphaGo is an AI system built by DeepMind that beat Lee Sedol, one of the strongest Go players in the world, four games to one in 2016. Go has more possible positions than there are atoms in the known universe,

Andrew Ng is a British-American computer scientist who co-founded Google Brain and the online learning platform Coursera. He is one of the best-known educators in AI, and his courses have taught machine learning to millions of people who never studied

Context engineering is deciding what information an AI model receives for a task, and in what form, so it can do the job well. It covers the instructions, examples, documents, conversation history, memory and tools the model can draw on,

A context window is the amount of text, measured in tokens, that an AI model can take into account at once. It includes the conversation, instructions and any documents, and works like short-term memory: anything outside the window is not

Deep learning is a type of machine learning that uses artificial neural networks with many layers to learn from large amounts of data. It drives most of the recent progress in AI, from speech and image recognition to translation and

General AI, or artificial general intelligence (AGI), is a hypothetical AI that could learn and perform any intellectual task a person can. It would carry knowledge from one field to another, and while it is the stated goal of several

Generative AI is artificial intelligence that creates new content, such as text, images, audio, video or code, in response to a prompt. It learns patterns from huge amounts of existing material and uses them to produce something new, which is

Google DeepMind is the AI research lab of Alphabet, Google’s parent company, and the team behind AlphaGo, AlphaFold and the Gemini models. It was formed in 2023 by merging DeepMind, a London start-up Google bought in 2014, with Google Brain,

Grounding is linking an AI model’s answers to specific, reliable sources, such as documents or search results, rather than only what it learned in training. A grounded answer can be traced back to its source, which makes it easier to

An AI hallucination is an answer that sounds confident and plausible but is wrong or made up, such as an invented statistic, quote, source or event. It happens because language models generate the most likely-sounding text, not text checked against

A large language model (LLM) is an AI model trained on vast amounts of text to predict the next word in a sequence. That one skill lets it write, summarise, translate and answer questions, and LLMs such as GPT, Claude,

Machine learning (ML) is the branch of AI in which computers learn from data instead of following rules written by hand. Shown enough examples, a model finds the patterns itself and uses them to make predictions about new cases, from

The Model Context Protocol (MCP) is an open standard for connecting AI applications to tools and data, such as files, calendars and business software. Instead of a custom integration for every pair, any assistant that supports MCP can use any

A multi-agent system is a setup in which several AI agents, each with its own role and tools, work together on a task too complex for one. One agent might research, another write and a third check the result, while

Narrow AI is artificial intelligence built for one task or a limited range of tasks, such as recommending videos, recognising faces or translating text. It can match or beat people within its domain but cannot transfer that skill to problems

Natural language generation (NLG) is the part of natural language processing that produces human-like text or speech from data or instructions. It writes the sentences in a chatbot reply, an automated sports or financial report, a product description or an

Natural language processing (NLP) is the field of AI that helps computers understand, interpret and produce human language, written or spoken. It powers search engines, voice assistants, translation tools, spam filters and chatbots, and today’s large language models grew out

A prompt is the instruction or question a person gives an AI tool to get a response, from a one-line question to a detailed brief with examples and context. The quality of the answer depends heavily on it: a vague

Prompt engineering is designing, testing and refining the instructions given to AI models so they produce reliable, useful results. It matters most when a prompt is reused, inside a product, an automated workflow or a team template, where small changes

A prompt library is a collection of tested, reusable prompts kept for recurring tasks, such as writing captions, summarising reports or drafting replies. It saves time and keeps AI-assisted work consistent across a team, because everyone starts from a prompt

Prompt literacy is the everyday ability to work with AI tools well: asking clearly, giving context, judging the answers and knowing when not to trust them. Where prompt engineering is a specialist skill, prompt literacy is the basic fluency anyone

Retrieval-augmented generation (RAG) is a technique in which an AI first looks up relevant passages in a set of documents, then writes its answer from them. It lets a model work with current or private information it was never trained

The Turing Test is a test of machine intelligence that the mathematician Alan Turing proposed in 1950. A judge holds text conversations with a person and a machine, and if the judge cannot reliably tell which is which, the machine

Vibe coding is building software by describing what you want to an AI tool in plain language and letting it write the code, often without reading it closely. The computer scientist Andrej Karpathy coined the term in February 2025, and