AI glossary
Definitions written for someone who has to make a decision, not pass an exam. Each entry says what the term means, when it matters and what it does not solve.
A
C
Chain of thought CoT
Prompt the model to write out its intermediate steps before giving the final answer.
Connector
A link between an AI app and an outside service, such as a drive, calendar or chat tool, that lets the model read your data from there.
Context window
The most text, counted in tokens, that a model can consider at once. Your prompt and its answer share that space.
E
Embeddings
Lists of numbers that stand for text, built so that similar meanings end up close together. Semantic search runs on them.
F
Few-shot prompting
Put two or three worked examples of the task in the prompt, on top of the description.
Fine-tuning
More training for an existing model on your own examples. It changes how the model behaves, not what it knows.
H
Hallucination
A fluent, confident answer that is plain wrong: an invented citation, function, statistic or fact.
I
Inference
Running a trained model to get output. It happens every time you send a prompt.
L
Large language model LLM
A model trained on huge amounts of text to predict the next token. That one skill lets it write, summarise, translate and reason over text.
M
Model Context Protocol MCP
An open standard that lets an AI app connect to outside tools and data, such as files, databases or services, through small programs called servers.
Multimodal
A model that takes in or produces more than one kind of data, such as text with images, audio or video.
P
Prompt injection
An attack where instructions hidden in content the model reads get carried out as though you had written them.
Q
Quantisation
Storing a model with lower-precision numbers so it needs less memory and runs on ordinary hardware.
R
Reasoning model
A model trained to spend extra computation working through a problem before it answers.
Retrieval-augmented generation RAG
Fetch the relevant documents first, paste them into the prompt, and the model answers from your data instead of its training.
S
Skill
A packaged set of instructions, and sometimes scripts, that an AI app loads when a task calls for it, so you do not repeat the same guidance every time.
Slash command
A command typed as /word at the start of a message. The app recognises it itself and does not send it to the model as text.
System prompt
Instructions given to the model outside the conversation. They apply to every turn, not to one message.
T
Temperature
A sampling setting that decides how much randomness goes into picking each next token.
Token
The unit a model reads and writes, roughly a word fragment. Price, context limits and speed are all counted in tokens.
V
Vector database
A store built to find the vectors nearest to a query vector fast, across millions of items.