AI Glossary

An AI system that can take actions, use tools, and make decisions autonomously to complete a goal. Read about Agent →

Artificial General Intelligence — a hypothetical AI that matches or exceeds human ability across virtually any intellectual task. Read about AGI →

Artificial Intelligence — software that performs tasks normally requiring human intelligence, like understanding language, recognising images, or generating content. Read about AI →

A conversational AI tool that helps users with tasks like writing, scheduling, research, or answering questions in natural language. Read about AI Assistant →

The challenge of making AI systems behave in ways that match human values and intentions — not just their literal instructions. Read about Alignment →

An interface that lets developers send requests to an AI model and get responses programmatically — the way most AI tools talk to LLMs. Read about API →

Deliberately crafted inputs that trick an AI model into producing wrong or harmful outputs — a key category of AI security threat. Read about Adversarial Attack →

The scaffolding around an LLM — tools, memory, loops, and orchestration — that turns a model into an agent. Read about Agent Harness →

The coordination layer that decides which agent or tool runs next, manages state across steps, and handles failures in multi-step AI workflows. Read about Agent Orchestration →

A group of AI agents that work together — often with different roles — to solve a problem one agent could not handle alone. Read about Agent Swarm →

AI systems designed to act, not just respond — they plan, use tools, and make decisions across multiple steps to complete a goal. Read about Agentic AI →

A multi-step task where an AI agent autonomously decides the steps, uses tools as needed, and works toward a goal with minimal human steering. Read about Agentic Workflow →

Software that uses an LLM to plan and act — picking tools, taking actions, and adapting based on results to complete a user’s goal. Read about AI Agent →

A photorealistic or stylised digital character driven by AI — used for video presenters, customer service, training, and marketing. Read about AI Avatar →

The structured process of measuring how well an AI model performs — accuracy, safety, cost, latency — usually with a fixed test set called an eval. Read about AI Evaluation →

A product designed from the ground up around AI capabilities — as opposed to bolting AI features onto an existing app. Read about AI Native →

Google’s AI-generated answer summary at the top of search results — synthesised from multiple sources, replacing some traditional blue-link traffic. Read about AI Overview →

A search experience that returns a synthesised AI answer with citations — instead of a ranked list of links — powered by retrieval plus an LLM. Read about AI Search →

A product whose value is mostly a thin UI over someone else’s foundation model — often used as a critique, sometimes as a description. Read about AI Wrapper →

A standardised test used to compare AI models on specific tasks — like coding, maths, reasoning, or following instructions. Read about Benchmark →

When an AI model's outputs systematically reflect unfair patterns from its training data — about gender, race, age, or other groups. Read about Bias →

A prompting technique where you ask the AI to "think step by step" before giving an answer — usually leading to better reasoning. Read about Chain of Thought →

A program that simulates conversation with users — increasingly powered by LLMs to handle natural-language questions and tasks. Read about Chatbot →

AI that can interpret images and video — recognising objects, reading text, detecting faces, or describing scenes. Read about Computer Vision →

The maximum amount of text (tokens) an AI model can read and remember at once during a single conversation. Read about Context Window →

AI systems designed for natural back-and-forth dialogue with users — covering chatbots, voice assistants, and AI agents. Read about Conversational AI →

An AI assistant embedded directly into a workflow — like coding, writing, or design — that suggests, completes, or generates work alongside the user. Read about Copilot →

Using an AI model to write source code from a natural-language description, a partial snippet, or a test. Read about Code Generation →

A sandboxed code-execution tool an AI agent can call to run scripts, do math, analyse files, or generate charts on the fly. Read about Code Interpreter →

An AI agent specialised for writing, editing, and debugging code — usually with the ability to read your repo, run tests, and open pull requests. Read about Coding Agent →

The capability for an AI model to control a computer the way a human does — moving the mouse, clicking, typing, reading the screen. Read about Computer Use →

A type of machine learning that uses layered neural networks to learn complex patterns — the foundation of modern AI. Read about Deep Learning →

AI-generated media — usually video or audio — that convincingly impersonates a real person saying or doing something they didn't. Read about Deepfake →

The type of AI model behind most modern image and video generators — it learns to create content by reversing a noising process. Read about Diffusion Model →

Training a smaller, cheaper AI model to mimic the outputs of a larger, more capable one — preserving most of the quality at a fraction of the cost. Read about Distillation →

An attack that corrupts a model’s training data to make it behave incorrectly — either degrading performance or installing hidden backdoors. Read about Data Poisoning →

An AI agent feature that spends minutes (not seconds) browsing many sources, reasoning across them, and producing a long-form cited report. Read about Deep Research →

A way of converting text (or images) into lists of numbers so an AI can measure how similar two pieces of content are. Read about Embeddings →

AI that operates a physical body — usually a robot — using vision, language, and motor control to act in the real world. Read about Embodied AI →

A prompting technique where you include a handful of examples in the prompt so the AI learns the pattern you want it to follow. Read about Few-shot Learning →

Further training a pre-trained AI model on your own data to specialise it for a specific task or style. Read about Fine-tuning →

A large, general-purpose AI model trained on broad data that can be adapted (via prompting or fine-tuning) to many downstream tasks. Read about Foundation Model →

AI systems that create new content — text, images, audio, video, or code — rather than just classifying or predicting from existing data. Read about Generative AI →

Generative Pre-trained Transformer — the architecture behind OpenAI's models, and now used as shorthand for any LLM-powered chatbot. Read about GPT →

Rules and filters that constrain what an AI model can output — used to block harmful, off-topic, or non-compliant responses. Read about Guardrails →

A search experience that generates an answer in natural language — using retrieved sources as input to an LLM — instead of returning a list of links. Read about Generative Search →

When an AI confidently states something that is factually wrong or completely made up. Read about Hallucination →

A structured collection of documents an AI system can search and quote — the source-of-truth corpus that grounds RAG and many AI agents. Read about Knowledge Base →

Training a small "student" model to imitate a large "teacher" model — capturing most of the teacher’s capability at a fraction of the size and cost. Read about Knowledge Distillation →

A structured representation of entities and the relationships between them — used to give AI systems explicit, queryable facts. Read about Knowledge Graph →

The time it takes an AI model to respond to a request — from when you hit send to when the first or final word appears. Read about Latency →

Large Language Model — the type of AI behind tools like ChatGPT and Claude, trained to understand and generate text. Read about LLM →

A type of AI where systems learn patterns from data rather than being explicitly programmed with rules. Read about Machine Learning →

Model Context Protocol — an open standard that lets AI models connect to external tools and data sources in a consistent way. Read about MCP →

An AI model that can understand and work with multiple types of input — text, images, audio, or video — not just text. Read about Multi-modal →

A model architecture that has many "expert" subnetworks but activates only a few per token — getting big-model quality at small-model inference cost. Read about Mixture of Experts →

The branch of AI focused on understanding, generating, and working with human language — covering everything from spell-check to ChatGPT. Read about Natural Language Processing →

A computing system inspired by the brain, made up of layers of connected "neurons" that learn patterns from data — the building block of modern AI. Read about Neural Network →

AI tools and platforms that let non-developers build AI-powered apps and workflows through visual interfaces instead of writing code. Read about No-Code AI →

Optical Character Recognition — AI that converts text inside images, scanned documents, or PDFs into editable, searchable text. Read about OCR →

A feature that stores parts of a prompt the model has already processed, making repeat or follow-up requests much faster and cheaper. Read about Prompt Caching →

The practice of crafting inputs to an AI model carefully to get better, more reliable outputs. Read about Prompt Engineering →

A security attack where malicious instructions hidden in user input or external content trick an AI model into ignoring its real instructions. Read about Prompt Injection →

Everything done to a model after pretraining — fine-tuning, RLHF, DPO, safety training — to turn a raw base model into a usable product. Read about Post-training →

Retrieval-Augmented Generation — a technique that gives an AI model access to external documents before it answers, so it can cite real, up-to-date sources. Read about RAG →

Reinforcement Learning from Human Feedback — the training technique that teaches AI models to be helpful, harmless, and honest. Read about RLHF →

A cap on how many requests or tokens a user can send to an AI API in a given window — used to manage cost, capacity, and abuse. Read about Rate Limit →

An agent pattern that interleaves "reasoning" steps with "acting" steps — letting the model think out loud, take an action, observe, and reason again. Read about ReAct →

A model variant trained or tuned to spend more compute on internal reasoning before answering — better on math, code, and multi-step problems. Read about Reasoning Model →

Deliberately trying to make an AI model misbehave — find jailbreaks, exploits, and failure modes — before adversaries do. Read about Red Teaming →

Search that finds results by meaning rather than exact keyword matches — so "car" finds results about "automobile" too. Read about Semantic Search →

AI that converts spoken audio into written text — the technology behind voice assistants, transcription tools, and meeting recorders. Read about Speech-to-Text →

Sending an AI model's response token-by-token as it's generated, so the user sees text appear immediately instead of waiting for the full reply. Read about Streaming →

The high-level instructions given to an AI model at the start of a conversation that define its role, behaviour, and constraints. Read about System Prompt →

An isolated execution environment where AI-generated code or agent actions can run without affecting the host system. Read about Sandbox →

An NLP task that classifies text as positive, negative, or neutral — used at scale for reviews, support tickets, social media, and survey responses. Read about Sentiment Analysis →

A compact language model — typically 1B to 15B parameters — designed to run cheaply, fast, or on-device while still being useful for focused tasks. Read about Small Language Model →

Forcing an LLM to return data in a specific format — usually JSON matching a schema — so downstream code can parse it reliably. Read about Structured Output →

Compressing a longer text — a meeting transcript, an article, a chat thread — into a shorter version that keeps the key information. Read about Summarization →

A hypothetical AI that dramatically exceeds human cognitive ability across every domain — beyond AGI on the capability scale. Read about Superintelligence →

A setting that controls how random or creative an AI model's responses are — lower values produce focused answers, higher values produce more varied ones. Read about Temperature →

AI that generates new images from a written description — the technology behind tools like Midjourney, DALL-E, and Stable Diffusion. Read about Text-to-Image →

AI that converts written text into natural-sounding spoken audio — used for narration, accessibility, voice assistants, and content creation. Read about Text-to-Speech →

AI that generates video clips from a text description — the next frontier after text-to-image, with rapidly improving quality. Read about Text-to-Video →

The basic units of text that AI models read and write — roughly ¾ of a word each. Models are priced and limited by token count. Read about Tokens →

The ability of an AI model to call external tools — like a calculator, search engine, or API — to help answer a question. Read about Tool Use →

The dataset an AI model learns from — its quality, diversity, and biases directly shape what the model can do and how well it does it. Read about Training Data →

The neural network architecture introduced in 2017 that powers nearly every modern LLM, image generator, and AI breakthrough. Read about Transformer →

The amount of compute spent at inference time on a single response — increased dramatically by reasoning models to improve quality. Read about Test-time Compute →

A decoding strategy that picks the next token only from the top K most likely candidates — trading diversity for focus. Read about Top-k Sampling →

A database optimised for storing and searching embeddings (numerical representations of text or images) by similarity. Read about Vector Database →

AI that learns to mimic a specific person's voice from a short sample, then generates new speech in that voice from any text. Read about Voice Cloning →

A 2025–26 coined term for writing software by chatting with an AI agent — describing the vibe of what you want and accepting whatever code it generates. Read about Vibe Coding →

A multimodal model that processes both images and text — letting you ask questions about an image, generate captions, or reason over visual content. Read about Vision-Language Model →

AI-powered tools that chain together multiple steps — apps, APIs, and AI models — to automate end-to-end business processes. Read about Workflow Automation →

Embedding a hidden, machine-detectable signal in AI-generated content so it can later be identified as AI-made. Read about Watermarking →

An HTTP callback an AI service makes to your endpoint when a long-running event completes — async results, agent updates, batch jobs. Read about Webhook →