AttentionDeep Learning EmbeddingMachine Learning Large Language ModelNatural Language Retrieval-Augmented Ge...Natural Language
AI Glossary

The words behind the hype, defined in Plain English

Plain-English definitions of the terms you keep running into.

Start with: Attention Embedding LLM RAG Token Transformer
32 Terms Defined
8 Areas Covered
12 Start-Here Terms
22 Aliases Decoded
21 Letters Covered
AI Directory Companies Comparisons Alternatives Best-of Guides Use Cases Collections AI Tools AI Models AI Courses
Jump to a letter
6 terms defined
Updating…
Sort definitions by
Alphabetical order The whole glossary from A straight through to Z.
Prompting
Explore by area

Browse by Subject Area

8 subject areas across 32 definitions — start wherever you already are.

Machine Learning

The core ideas behind systems that learn patterns from data.

7 terms

Deep Learning

Neural networks with many layers, and the machinery that trains them.

5 terms

Natural Language

How machines read, generate and reason about human language.

7 terms

Generative AI

Models that produce new text, images, audio or code.

4 terms

Computer Vision

Making sense of images and video.

1 term

Data & Training

Datasets, labelling, evaluation and the training loop itself.

1 term

Ethics & Safety

Alignment, bias, interpretability and the limits of these systems.

3 terms

Infrastructure

Serving, scaling and running models in production.

4 terms
View all areas
Short forms

Acronyms & Aliases

Every short form that resolves to a full entry.

ANN Neural Network Artificial Neural Network Neural Network Backprop Backpropagation CNN Convolutional Neural Network Confabulation Hallucination Context Length Context Window ConvNet Convolutional Neural Network CoT Chain-of-Thought GAN Generative Adversarial Network In-Context Learning Few-Shot Learning Knowledge Distillation Model Distillation LLM Large Language Model Quantisation Quantization RAG Retrieval-Augmented Generation Reinforcement Learning from Human Feedback RLHF RL Reinforcement Learning RNN Recurrent Neural Network Self-Attention Attention Tokenisation Token Tokenization Token Vector Embedding Embedding Vector Store Vector Database
How this glossary works

Definitions You Can Actually Use

A reference works when you can arrive from any direction, read one entry, and leave knowing the thing.

PLAIN ENGLISH

Written To Be Understood

Every definition opens with one sentence a non-specialist can read, then goes deeper. No entry starts by using three other terms you have not met yet.

Cross-Linked, Not Isolated

Entries name what they build on and what they contrast with, so you can follow a thread instead of bouncing between search results.

Levelled Honestly

Beginner assumes no maths. Advanced assumes linear algebra and some familiarity with the literature. The label tells you before you start reading.

Every Acronym Resolves

Search RAG, CoT, ANN or ConvNet and you land on the full entry. Aliases are real records, not a guess made at query time.

Reviewed By People

Definitions are edited and checked rather than generated in bulk. When one is wrong we would rather you told us than that we shipped more of them.

Two Ways In

Jump to a letter if you know the word, or browse by subject area if you only know roughly where it sits. Both reach the same catalogue.

Questions

Frequently Asked Questions

About this glossary

Who writes these definitions?
They are written and reviewed by the AIToolsay editorial team. Every entry records when it was last reviewed, and entries that cite a specific paper or standard link to it under Sources.

How do you decide what counts as a term?
If you would plausibly hit it in a paper, a product page or a job description and have to look it up, it belongs here. We prefer one well-written entry over three overlapping ones, which is why synonyms redirect to a single canonical page rather than becoming their own stubs.

Why do some terms redirect?
“LLM” and “Large Language Model” are the same idea, so one of them is canonical and the other sends you there. That keeps the definition in one place instead of letting two copies drift apart.

What do the difficulty levels mean?
Beginner assumes no maths. Intermediate assumes you are comfortable with the basics of machine learning. Advanced assumes linear algebra and some familiarity with the literature.

Can I suggest a term or a correction?
Yes — use the link at the bottom of any entry. Corrections with a source attached get looked at first.

Missing a term?

Tell us what you were looking for and we will write it up.

Suggest a term