Artificial intelligence is now used across business, technology, data and everyday productivity tools. This A-Z glossary explains common AI, generative AI, machine learning, large language model and responsible AI terminology in clear, practical language.
A
AI Agent
An AI system designed to pursue a goal by planning steps, using tools or data, taking actions and responding to results with some degree of autonomy.
Artificial Intelligence (AI)
A broad field concerned with computer systems that perform tasks associated with human intelligence, such as reasoning, language understanding, perception, prediction and decision-making.
Artificial General Intelligence (AGI)
A hypothetical form of AI able to perform a very wide range of intellectual tasks at a general human-like level rather than being specialised for particular tasks.
B
Bias
Systematic patterns in data, model behaviour or decisions that can produce unfair, inaccurate or unbalanced results for particular groups or situations.
Benchmark
A standard test or dataset used to compare the performance of AI models on defined tasks such as reasoning, coding, language or image recognition.
C
Chatbot
Software that communicates with users through conversational text or speech. Modern chatbots often use large language models to generate responses.
Computer Vision
A branch of AI that enables systems to analyse and interpret images, video and other visual information.
Context Window
The amount of text or other information a model can consider at one time when generating a response. It may include the prompt, conversation history and retrieved information.
D
Deep Learning
A form of machine learning that uses neural networks with many layers to learn complex patterns from large amounts of data.
Diffusion Model
A generative model commonly used for image creation that learns to reverse a process of adding noise, gradually producing a coherent output from random noise.
E
Embedding
A numerical representation of text, images or other data that places similar meanings or concepts closer together in a mathematical space.
Explainable AI (XAI)
Methods intended to make AI outputs or decisions easier for people to understand, inspect and justify.
F
Fine-tuning
Further training a pre-trained model on a smaller, targeted dataset so it performs better for a particular task, style or domain.
Foundation Model
A large model trained on broad data that can be adapted or prompted for many downstream tasks rather than being built for one narrow purpose.
G
Generative AI
AI that creates new content such as text, images, audio, video, code or structured data based on patterns learned during training.
Grounding
Connecting an AI response to trusted, relevant information such as company documents, databases or search results to improve accuracy and relevance.
H
Hallucination
An AI-generated statement that sounds plausible but is inaccurate, unsupported or invented.
Human in the Loop
An approach in which people review, guide or approve AI outputs or decisions rather than allowing the system to operate entirely without human oversight.
I
Inference
The process of using a trained AI model to generate a prediction, classification, recommendation or other output from new input.
Instruction Tuning
Training a model on examples of instructions and desired responses so it becomes better at following user requests.
J
Jailbreak
A prompt or technique intended to bypass an AI system's normal safety rules, restrictions or behavioural controls.
K
Knowledge Graph
A structured representation of entities and the relationships between them, often used to organise information that AI systems can query or reason over.
L
Large Language Model (LLM)
A neural network trained on very large amounts of text and other data to understand and generate language, answer questions, summarise, reason and perform related tasks.
Latency
The delay between sending a request to an AI system and receiving its response. Lower latency generally produces a more responsive user experience.
M
Machine Learning (ML)
A branch of AI in which systems learn patterns from data and use those patterns to make predictions or decisions without being explicitly programmed for every case.
Model
The trained mathematical system that transforms inputs into outputs. Different models may specialise in language, vision, audio, prediction or other tasks.
Multimodal AI
AI able to work with more than one type of input or output, such as text, images, audio and video.
N
Natural Language Processing (NLP)
The area of AI focused on enabling computers to analyse, understand and generate human language.
Neural Network
A machine learning architecture made from connected processing units arranged in layers and trained to learn patterns from data.
O
Open-source Model
A model made available under terms that permit some level of public access, reuse or modification. The exact freedoms depend on the model's licence and release conditions.
Overfitting
When a model learns its training data too closely and performs less effectively on new, unseen data.
P
Parameter
A value learned during model training that influences how the model processes information and generates outputs. Large models may contain billions of parameters.
Prompt
The instruction, question, context or other input supplied to an AI system to guide the response it generates.
Prompt Engineering
The practice of designing and refining prompts, context and examples to obtain more useful, reliable or consistent AI outputs.
Q
Quantisation
A technique that reduces the numerical precision used by a model so it can require less memory and compute, often with a small trade-off in accuracy.
R
Retrieval-Augmented Generation (RAG)
An approach that retrieves relevant information from external sources and supplies it to a generative model as context before the model creates its response.
Reinforcement Learning
A machine learning method in which a system learns behaviour through rewards or penalties associated with actions and outcomes.
Responsible AI
Principles and practices for designing, deploying and governing AI in ways that consider fairness, safety, privacy, transparency, accountability and human impact.
S
Small Language Model (SLM)
A language model with fewer parameters and lower computing requirements than a typical large language model, often designed for efficiency or focused use cases.
Supervised Learning
Machine learning in which a model is trained using labelled examples that show the correct output for each training input.
Synthetic Data
Artificially generated data designed to resemble real-world data and used for testing, training, simulation or privacy-sensitive scenarios.
T
Temperature
A generation setting that influences how predictable or varied a model's output is. Lower values usually produce more consistent responses and higher values more variation.
Token
A unit of text processed by a language model. A token may represent a whole word, part of a word, punctuation or another text fragment.
Transformer
A neural network architecture based on attention mechanisms and widely used in modern language and multimodal AI models.
U
Unsupervised Learning
Machine learning in which a model identifies structure or patterns in data without being given labelled correct answers.
V
Vector Database
A database designed to store and search embeddings efficiently, often used in semantic search and retrieval-augmented generation systems.
Vision-Language Model (VLM)
A multimodal model able to interpret both images and language, for example by describing images or answering questions about visual content.
W
Weights
The learned numerical values inside a neural network that determine how strongly different inputs and internal signals influence the model's output.
X
XAI (Explainable AI)
A common abbreviation for Explainable AI, referring to methods that help people understand how an AI system reached a result.
Y
Yottabyte
A very large unit of digital storage equal to 10²⁴ bytes. The term is sometimes used when discussing the enormous scale of data associated with future computing and AI systems.
Z
Zero-shot Learning
The ability of a model to perform a task or recognise a category without having been given labelled training examples specifically for that task or category.
Zero-shot Prompting
Giving a model an instruction without providing worked examples of the desired answer format or task.
Build your AI knowledge
Explore our instructor-led Artificial Intelligence training and certification pathways covering generative AI, machine learning, cloud AI, prompt engineering and responsible AI.
View AI training coursesView AI certifications
.png){kind=small}
