Level : Beginner
Artificial Intelligence Glossary
This glossary explains 120 AI and agent terms in plain English, from everyday words to technical jargon. Type a word in the search box or filter by level to find the definition you need in seconds.
Level
120 of 120 terms
A
- Agent evaluation Advanced
- Methods for measuring whether an AI agent really completes a task end to end, reliably, safely and at an acceptable cost.
- Agent memory Advanced
- The mechanisms that let an AI agent store, retrieve and update information beyond a single request or conversation.
- Agent orchestration Advanced
- Coordinating several model calls, tools or AI agents: who does what, in which order, and how the results are brought together.
- Agent2Agent protocol (A2A) Advanced
- An open protocol that lets AI agents from different vendors discover each other, communicate and hand tasks to one another.
- Agentic coding Advanced
- Software development in which an AI agent reads the codebase, edits files, runs tests and fixes its own mistakes over several steps, under supervision.
- Agentic workflow Intermediate
- A sequence of steps in which an AI model plans, uses tools, checks and revises its work instead of answering in a single pass.
- AI agent Beginner
- An AI that does more than answer: it plans and carries out a series of actions with tools to reach a goal you set.
- AI alignment Advanced
- The research and methods aimed at making an AI system actually pursue the intentions and values of the people who build and use it.
- AI assistant Beginner
- A consumer app such as ChatGPT, Claude or Gemini that answers, writes, summarizes and helps with everyday tasks using AI.
- AI bias Beginner
- A systematic error that makes an AI treat some groups unfairly or repeat stereotypes, often inherited from its training data.
- AI browser (agentic browser) Beginner
- A web browser with a built-in AI assistant that summarizes pages and can sometimes act for you: clicking, filling in forms, booking.
- AI detector Beginner
- A tool that estimates whether a text was written by AI; its verdict is a probability that can be wrong, never proof.
- AI ethics Beginner
- The principles for designing and using AI in a way that is fair, transparent and respectful of people and their rights.
- AI memory Beginner
- A feature that lets an AI assistant remember facts about you from one conversation to the next.
- AI model Beginner
- The trained “brain” behind an AI app: a large file of learned numbers that turns an input into an output, such as a reply or an image.
- AI scams Beginner
- Frauds that use AI, such as cloned voices, fake videos or very convincing messages, to steal money or personal data.
- AI search Beginner
- A search where an AI reads web pages for you and writes a summarized answer with its sources, instead of a list of links.
- AI watermarking Beginner
- A mark, often invisible, embedded in an image, audio, video or text to signal that it was generated by AI.
- Algorithm Beginner
- A finite list of precise steps, followed in a set order, to solve a problem or complete a task, like a recipe for a computer.
- API Intermediate
- An interface that lets one program send requests to a service, such as an AI model, and get its responses back automatically.
- API key Intermediate
- A secret code that identifies your account to an API and is used to bill your usage: never share it or publish it anywhere.
- Artificial general intelligence (AGI) Beginner
- A hypothetical AI able to match humans across most intellectual tasks; there is no single agreed definition or timeline.
- Artificial intelligence Beginner
- Techniques that let machines perform tasks we associate with human intelligence, such as understanding, predicting or creating.
- Attention mechanism Intermediate
- The technique that lets a model weigh, for each word it processes, how relevant every other word in the text is.
- Automation Beginner
- Handing a repetitive task to software that runs it on its own, following fixed rules or with the help of AI.
B
- Benchmark Intermediate
- A standardised test, a fixed set of scored questions or tasks, used to compare how different AI models perform.
C
- Chain of thought Intermediate
- Getting a model to write out intermediate reasoning steps before its final answer, which helps on multi-step problems.
- Chatbot Beginner
- A program you talk to in writing or by voice, which replies in everyday language like a conversation partner.
- Chunking Advanced
- Splitting long documents into small, coherent passages so they can be indexed and only the useful pieces are sent to the model.
- Computer use Advanced
- The ability of an AI model to operate a computer the way a person does: reading screenshots, moving the mouse, clicking and typing.
- Computer vision Beginner
- The branch of AI that lets machines analyze images and video to recognize objects, text, faces or scenes.
- Constitutional AI Advanced
- An Anthropic training method in which an AI critiques and revises its own answers against a written list of principles called a constitution.
- Context engineering Advanced
- The practice of deciding what goes into a model's context window at each step: instructions, tools, documents, memory and history.
- Context window Beginner
- The maximum amount of text (in tokens) a model can read and produce at once, including its own answer.
- Custom instructions Beginner
- Preferences you save once in an AI assistant (who you are, tone, format) that it then applies to every conversation.
D
- Deep learning Beginner
- Machine learning that uses neural networks with many layers; it is the technique behind today's chatbots, image and voice AI.
- Deepfake Beginner
- An AI-made or AI-altered image, video or voice recording that makes a real person appear to say or do something they never did.
- Diffusion model Intermediate
- A type of generative model that creates an image (or audio, or video) by starting from random noise and cleaning it up step by step.
- DPO (direct preference optimization) Advanced
- An alignment method that trains a model directly on pairs of preferred and rejected answers, with no separate reward model.
E
F
- Few-shot prompting Intermediate
- A technique where you include a few worked examples of input and expected output in the prompt to show the model what you want.
- Fine-tuning Intermediate
- Continuing the training of an already trained model on a small, targeted set of examples to adapt it to a task, style or output format.
- Free vs paid AI Beginner
- How free and paid AI assistant plans differ: usage limits, available models, features and privacy options.
G
- GDPR and AI Beginner
- The EU's General Data Protection Regulation, which also applies to AI tools that process personal data.
- Generative AI Beginner
- AI that creates new content such as text, images, audio, video or code from a simple written or spoken request.
- GGUF (model file format) Advanced
- A single-file format bundling a model's weights, often quantized, with its metadata, so it can run locally with llama.cpp and related tools.
- GPQA Advanced
- A benchmark of PhD-level multiple-choice questions in biology, physics and chemistry, designed so a web search does not reveal the answer.
- GPU (graphics processing unit) Advanced
- A chip built to run thousands of calculations in parallel, which became the engine for training and running AI models.
- Grounding Intermediate
- Making an AI base its answer on supplied sources (web results, documents, databases) rather than only on what it memorized in training.
- Guardrails Intermediate
- Checks placed around an AI model to filter what goes in and what comes out, and to block unwanted answers or actions.
H
- Hallucination Beginner
- An AI answer that sounds confident and plausible but is false or made up: a figure, a quote, a source or a reference.
- Human-in-the-loop Intermediate
- A design where a person reviews, corrects or stops the important steps of an AI system or agent before they take effect.
- Hybrid search Advanced
- Search that combines exact keyword matching with meaning-based (embedding) search, then merges both result lists into a single ranking.
I
- Image generation Beginner
- When an AI creates a brand-new picture from a written description (a prompt) or from another image.
- Inference Intermediate
- The moment a trained model is actually used to produce an output: every question you send to an AI triggers an inference run.
- Inference server Advanced
- Software that loads an AI model into memory and answers requests through an API, while optimizing throughput and GPU memory.
- Interpretability Advanced
- Research that tries to understand what happens inside an AI model: which concepts it represents and how it actually reaches its answer.
K
- Knowledge cutoff Beginner
- The date after which an AI model received no more training data, so it knows nothing about later events.
- Knowledge distillation Intermediate
- A technique that trains a small student model to imitate the outputs of a large teacher model, so it becomes lighter and cheaper to run.
- KV cache Advanced
- Memory where a model keeps the work already done on previous tokens so it does not have to redo it for every new word.
L
- Large language model (LLM) Beginner
- A program trained on huge amounts of text to predict what comes next in a text; it is the engine behind AI assistants.
- Latency Intermediate
- The waiting time between sending a request to an AI model and receiving its answer, from the first word through to the last.
- LLM observability Advanced
- The traces, metrics and logs that show what an AI application is doing: prompts, responses, tools called, cost and latency.
- LLM-as-a-judge Advanced
- An evaluation method in which a language model scores or compares another model's answers against a rubric, standing in for a human grader.
- Long context Advanced
- A model's ability to process very long inputs in one go, such as books or code repositories, and the techniques that make it possible.
- LoRA (Low-Rank Adaptation) Advanced
- A fine-tuning method that freezes the model and trains only small added matrices, making it cheap to adapt a large language model.
M
- Machine learning Beginner
- A branch of AI where software learns patterns from examples instead of following rules written one by one by a person.
- Mixture of experts (MoE) Intermediate
- An architecture in which a model holds many expert sub-networks but only switches on a few of them for each token it processes.
- Model collapse Advanced
- The gradual degradation of models trained, generation after generation, on content produced by other models rather than by people.
- Model Context Protocol (MCP) Intermediate
- An open protocol that standardises how AI applications connect to external tools, data sources and services.
- Model parameters Intermediate
- The internal numbers a model tunes during training; how many it has indicates its size, not directly how good it is.
- Multi-agent system Intermediate
- A setup where several AI agents, each with its own role, split a task between them and share their results to get it done.
- Multimodal AI Beginner
- AI that can understand or produce several kinds of content, such as text, images, audio and video, often within the same conversation.
N
- Natural language processing (NLP) Beginner
- The branch of AI that lets computers read, understand, translate and generate human text and speech.
- Neural network Beginner
- Software made of many small connected computing units that learns by adjusting the strength of the links between them.
O
- Open-weights model Intermediate
- An AI model whose trained parameters (the weights) can be downloaded, so you can run it on your own hardware.
P
- Pre-training Intermediate
- The first training phase of a language model: it learns to predict the next piece of text across a huge corpus, with no specific task.
- Price per million tokens Intermediate
- How AI APIs bill usage: one rate per million tokens the model reads as input, and another, usually higher, per million it generates.
- Prompt Beginner
- The text, question or instruction you give an AI to get a response; how clear it is directly shapes the result.
- Prompt caching Intermediate
- Reusing the start of a prompt the model has already processed, so repeated requests run faster and the reused part is billed at a lower rate.
- Prompt engineering Intermediate
- The practice of writing, testing and refining instructions to an AI model so its answers become more reliable and useful.
- Prompt injection Intermediate
- An attack that slips hidden instructions into text an AI reads, to make it ignore its original instructions or act against its user.
Q
- Quantization Advanced
- A technique that stores a model's weights with fewer bits (8, 4 or even fewer) so it needs far less memory to run.
R
- RAG (retrieval-augmented generation) Intermediate
- A technique that first retrieves relevant passages from your documents, then hands them to the model so its answer is based on them.
- Rate limit Intermediate
- A cap set by an AI provider on how many requests or tokens you may send within a given time window, often per minute.
- ReAct pattern Advanced
- A loop in which an AI agent alternates written reasoning with actions such as searches or tool calls, reading each result before going on.
- Reasoning model Intermediate
- A language model trained to think step by step before answering: slower and pricier, but stronger at maths, code and logic.
- Red teaming Advanced
- Adversarial testing in which people or other AIs deliberately attack a model to uncover its weaknesses before real attackers do.
- Reranking Advanced
- A second search stage in which a more precise model rereads the top candidates and puts them back in the right order of relevance.
- Reward model Advanced
- An auxiliary model that scores an answer by predicting human preference, used as the training signal for another model.
- RLHF (reinforcement learning from human feedback) Intermediate
- A training method where people compare model answers, and the model learns to favour the responses humans judge helpful and safe.
- RLVR (reinforcement learning with verifiable rewards) Advanced
- Reinforcement learning in which the reward comes from an automatic check, such as a correct answer or passing code tests, not from a judgement.
- RoPE (rotary position embedding) Advanced
- A technique that tells a Transformer where each token sits by rotating its vectors through an angle that depends on that position.
S
- Sandboxing Advanced
- Running an AI agent inside an isolated environment with restricted files, network and commands, so any damage stays contained.
- Scaling laws Advanced
- Measured relationships that predict how a model's performance improves with its size, its training data and its compute.
- Self-reflection Advanced
- A technique in which a model or agent reviews its own output, critiques the mistakes and uses that critique to try again.
- Semantic search Intermediate
- Search that finds content by meaning rather than exact words, by comparing numerical representations of the texts.
- Small language model (SLM) Intermediate
- A compact language model, light enough to run on a laptop or phone, that performs well on focused tasks rather than everything at once.
- Speculative decoding Advanced
- A speed-up technique where a small model drafts several tokens ahead and the large model checks them all in a single pass.
- Speech recognition Beginner
- Technology that turns spoken words into written text: dictation, captions, voice commands and meeting notes.
- Structured output Intermediate
- An AI model's answer produced in an enforced format, usually JSON that matches a schema, so that a program can read it reliably.
- SWE-bench Advanced
- A benchmark that asks an AI to resolve real GitHub issues from Python projects, with each fix checked against the project's own tests.
- Synthetic data Intermediate
- Artificially created data, often generated by another AI model, used to train, fine-tune or test a model when real data is scarce.
- System prompt Intermediate
- Background instructions given to a model before the conversation starts (role, tone, rules) that shape every answer that follows.
T
- Task decomposition Advanced
- Breaking a complex goal into smaller subtasks that a model or AI agent can handle one at a time or in parallel.
- Temperature Intermediate
- A setting that controls how much randomness goes into a model's answers: low keeps it predictable, high makes it more varied.
- Terminal-Bench Advanced
- A benchmark that measures whether an AI agent can complete hard, realistic tasks on its own inside a command-line terminal.
- Test-time compute Advanced
- The computing power a model spends while answering, for instance to think for longer or to compare several candidate solutions.
- Text-to-speech Beginner
- Technology that turns written text into spoken audio, now sounding far more natural thanks to AI.
- Token Beginner
- A small chunk of text (a word, part of a word or a symbol) that an AI model reads and writes; the unit for limits and pricing.
- Tokenizer Intermediate
- The program that splits text into tokens (word pieces) and turns them into numbers, the only form a language model can actually read.
- Tool calling (function calling) Intermediate
- The ability of an AI model to request that an external function be run with specific arguments, then use the result in its answer.
- Training compute (FLOP) Advanced
- The total number of arithmetic operations, measured in FLOP, needed to train an AI model.
- Training data Beginner
- The examples (text, images, audio and more) an AI model learns from; they shape what the model knows and where it goes wrong.
- Transformer Intermediate
- A neural network architecture built on attention, introduced in 2017, that underpins almost every large language model today.
- Turing test Beginner
- A test proposed by Alan Turing in 1950: a machine passes if a judge cannot tell it apart from a human in a text conversation.
V
- Vector database Intermediate
- A database that stores embeddings and quickly finds the ones closest to a query, which means the most relevant pieces of content.
- Voice assistant Beginner
- An assistant you control by speaking, on a phone, smart speaker or car: it listens, works out the request and answers out loud.
- VRAM (video memory) Advanced
- The memory built into a graphics card, where an AI model must fit to run fast; it is often what limits what you can run locally.
No term matches. Try a synonym or a shorter word.
How to use the glossary
- Search for a word: the list filters instantly as you type the first few letters.
- Filter by level: beginner, intermediate or advanced. If you are new to AI, start with the beginner level; the other terms will make sense afterwards.
- Read the short definition, then open the term’s page if you want a fuller explanation and examples.
Keep this page handy when you read a comparison or an article: it is the fastest way to clear up a doubt without losing your train of thought. Definitions are deliberately short and jargon-free; when one term leads to another, the term’s page points you to related concepts so you can build your understanding step by step.
The essential terms to get started
If you only remember a few words, make it these.
Artificial intelligence (AI): the set of techniques that let a machine handle tasks we usually associate with human intelligence, such as understanding text, recognizing an image or making a decision.
Model: a program that has learned from large amounts of examples and produces an answer based on what you give it. When people compare AIs, they are usually comparing models.
LLM (large language model): a model specialized in text, trained on huge volumes of writing. It is the technology at the heart of chat assistants such as ChatGPT.
Token: a small chunk of text, often a word or part of a word, that the model reads and writes. Tokens are the unit used to measure text length and to calculate what a model costs to use.
Prompt: the instruction you write to the AI. The more specific it is (context, goal, expected format), the more likely the answer will be useful.
Context window: the amount of text, counted in tokens, that a model can take into account at once, including your request, any documents you provide and its own reply. Beyond that, it “forgets” the beginning.
Hallucination: a false or made-up answer delivered with confidence. It is a known limitation of language models, so always double-check important facts.
AI agent: a system that uses a model to reach a goal over several steps, with the help of tools (web search, spreadsheet, email). Where an assistant answers, an agent acts.
Where to go next
Once these basics click, put them into practice. The Start track guides you step by step, then Go further and Build take you deeper. To see what a token really looks like, try the token estimator. And if you want to know where you stand, take the AI level test.
Frequently asked questions
How are the glossary terms organized?
The 120 terms are split into three levels, beginner, intermediate and advanced, with 40 terms each. The level filter lets you show only the ones that match where you are, then move up at your own pace.
What is the difference between AI, a model and an agent?
AI is the overall field. A model is a specific program, trained on data, that produces answers. An agent uses a model to act over several steps, for example looking something up and then filling in a document.
I can't find a term. What should I do?
Try a different spelling, an abbreviation or a synonym: many AI concepts go by several names. If the word is not there, you can suggest it through the contact page.
Are the definitions beginner-friendly?
Yes. Each term has a one-sentence definition written without jargon. Its dedicated page then goes deeper, with examples.