AI glossary
Short, plain-language definitions of the words you will meet when learning about AI. Each term has its own page with an example and related terms.
72 terms
A
Agent memory
Information an AI assistant or agent keeps between conversations, such as your preferences or facts about a project, so it does not start from zero each time.
Agent skills
Packaged instructions, scripts and resources that an AI assistant loads only when a task needs them, to do that task in a specific, repeatable way.
AGI (artificial general intelligence)
A hypothetical AI able to match or exceed humans at most cognitive tasks, not just specific ones.
AI agent
A system in which a language model uses tools in a loop — deciding, acting, checking the result — to reach a goal with little step-by-step guidance.
Alignment
The work of making AI systems pursue the goals and values their developers and users intend, and not harmful or unintended ones.
API
An interface that lets one program talk to another. AI providers offer APIs so developers can use their models inside their own apps.
API key
A secret string that identifies you to an AI provider’s API, so it can authorise your requests and bill your account.
Artificial intelligence (AI)
The field of building computer systems that perform tasks we associate with human intelligence, such as understanding language or recognising images.
B
C
Chain of thought
Having a model work through a problem step by step before giving the final answer, which improves accuracy on reasoning tasks.
Chatbot
A program you talk to in natural language, by text or voice. Modern chatbots such as ChatGPT, Claude or Gemini run on large language models.
Coding agent
An AI agent that works on a codebase: it reads files, edits code, runs commands and tests, and iterates until a task is done.
Computer use
The ability of an AI agent to operate software through its graphical interface: looking at the screen, moving the pointer, clicking and typing.
Context compaction
Summarising the older part of a long conversation or agent run so it fits in the context window and the work can continue.
Context engineering
Choosing what goes into a model’s context at each step (instructions, documents, tool results, memory) so it has what it needs and nothing more.
Context window
The maximum amount of text, measured in tokens, that a model can take into account at once: your messages, its replies, files and instructions.
D
Deep learning
Machine learning with neural networks that have many layers, able to learn complex patterns from raw data like text, images and sound.
Deepfake
An image, video or audio made or altered with AI to show a real person saying or doing something they never did.
Diffusion model
A type of generative model that creates images (or video and audio) by starting from random noise and removing it step by step.
Distillation
Training a smaller “student” model to imitate a larger “teacher” model, keeping much of its quality at a fraction of the cost.
E
Effort (reasoning effort)
A setting that tells a reasoning model how hard to think before answering: more effort means better answers on hard problems, but slower and more expensive.
Embedding
A list of numbers that represents the meaning of a piece of text (or an image), so that similar meanings end up close together.
EU AI Act
The European Union’s law on artificial intelligence, which sets obligations according to risk, from banned practices to transparency rules.
Evals (evaluations)
A set of test cases with expected results used to measure how well an AI system performs on a specific task.
F
Few-shot prompting
Including a few worked examples of input and desired output in the prompt so the model copies the pattern.
Fine-tuning
Further training an existing model on your own examples so it adapts to a specific task, style or format.
Frontier model
One of the most capable AI models available at a given moment, the kind that pushes the state of the art.
Function calling (tool use)
A model capability to request that an app run a specific function with specific arguments, such as “search_flights(from, to, date)”.
G
Generative AI
AI that creates new content — text, images, audio, video or code — instead of only classifying or predicting from existing data.
Grounding
Tying a model’s answer to specific sources — search results, documents, a database — so it is based on checkable facts rather than memory alone.
Guardrails
Checks and limits placed around an AI system to keep its inputs and outputs safe, on-topic and within policy.
H
I
J
K
L
M
Machine learning
A way of building AI where the system learns patterns from examples instead of following rules written by hand.
MCP server
A program that exposes tools, data or prompts to AI applications through the Model Context Protocol.
Mixture of experts (MoE)
A model design that splits the network into many “experts” and activates only a few of them for each token, so a huge model runs at the cost of a smaller one.
Model Context Protocol (MCP)
An open standard for connecting AI applications to external tools and data, so one connector works across many AI apps.
Multi-agent system
A setup where several AI agents, often with different roles or tools, work together on a task.
Multimodal
Able to work with more than one type of data — for example text, images, audio, video or PDFs — in the same model.
N
O
P
Parameters
The internal numbers (mostly weights) a model learns during training. Their count is a rough measure of its size.
Pre-training
The first and most expensive stage of building a language model, in which it learns from a very large body of text by predicting the next token.
Prompt
The input you give an AI model — a question, instruction, examples or documents — that it responds to.
Prompt caching
Reusing the processed beginning of a prompt across requests, which makes repeated long prompts cheaper and faster.
Prompt engineering
The practice of designing and testing prompts so a model reliably produces the output you need.
Prompt injection
An attack in which text hidden in content the model reads — a web page, email or document — tries to override its instructions.
Q
R
Rate limit
The maximum number of requests or tokens you may send to an API in a period of time, such as per minute.
Reasoning model
A language model trained to think through a problem internally before answering, trading extra time and tokens for better results on hard tasks.
Red teaming
Deliberately attacking an AI system, the way an adversary would, to find dangerous behaviour and security holes before release.
Reinforcement learning (RL)
A way of training where a model tries things, receives a reward or penalty, and gradually learns which actions pay off.
Retrieval-augmented generation (RAG)
A technique where relevant documents are retrieved first and given to the model, so it answers from your sources instead of from memory.
RLHF
Training a model with reinforcement learning where the reward comes from human judgements of which answers are better.
S
Structured outputs
Making a model answer in an exact machine-readable format, usually JSON that matches a schema you provide.
Subagent
An agent launched by another agent to handle one part of a task, usually with its own fresh context, and that reports back a summary.
Synthetic data
Training data created by AI models or programs rather than collected from people, used to teach skills where real examples are scarce.
System prompt
Instructions set by the developer or app that frame the whole conversation: the model’s role, rules, tone and tools.
T
Temperature
A setting that controls how random a model’s word choices are: low values give predictable text, high values more varied text.
Token
The unit of text a language model reads and writes: a whole word, part of a word, a number or a punctuation mark.
Transformer
The neural network architecture behind modern language models. Its “attention” mechanism lets every token look at every other token in the input.
V
W
Z
No term matches that search.