AI Glossary

Plain-English definitions of key AI concepts — from LLMs to RAG to agents.

A

An LLM-powered system that autonomously plans and executes multi-step tasks by calling tools, browsing the web, writing code, or taking actions in external systems.

C

Context Window

Fundamentals

The maximum number of tokens an LLM can process in a single request, covering both input and output.

Chain of Thought (CoT)

Techniques

A prompting technique that instructs the model to reason step by step before giving a final answer, improving accuracy on complex tasks.

D

Diffusion Model

Models

A class of generative AI models that learn to create images (or audio/video) by reversing a process of adding noise to training data.

E

Embeddings

Fundamentals

Dense numerical vectors that represent the semantic meaning of text, enabling similarity comparisons between pieces of content.

F

Fine-Tuning

Techniques

Continued training of a pretrained model on a smaller, domain-specific dataset to adapt its behavior.

Few-Shot Learning

Techniques

Providing a small number of input-output examples in the prompt to teach the model a pattern without changing its weights.

H

Hallucination

Evaluation

When an LLM generates factually incorrect, fabricated, or nonsensical content presented with apparent confidence.

I

Inference

Infrastructure

The process of running a trained AI model on new inputs to generate predictions or responses — as opposed to training.

L

Large Language Model (LLM)

Models

A deep learning model trained on massive text datasets to understand and generate human language.

M

Model Context Protocol (MCP)

Agents

An open standard by Anthropic that lets AI agents connect to external tools and data sources via a unified protocol.

Multimodal AI

Models

AI models that process and generate multiple types of data — text, images, audio, and video — within a single unified model.

P

Prompt Engineering

Techniques

The practice of crafting inputs to an AI model to reliably elicit desired outputs.

R

Retrieval-Augmented Generation (RAG)

Techniques

A technique that grounds LLM responses in retrieved external documents, reducing hallucination and enabling up-to-date answers.

Reinforcement Learning from Human Feedback (RLHF)

Techniques

A training technique where human raters rank model outputs and those preferences train a reward model that guides further LLM fine-tuning.

S

System Prompt

Fundamentals

A privileged instruction block sent to an LLM before the conversation begins, used to set role, constraints, and behavior.

T

Tokens

Fundamentals

The basic units LLMs process — roughly word-pieces. Text is split into tokens before the model reads it, and token count determines API cost.

Temperature

Fundamentals

A sampling parameter (0–2) that controls the randomness of LLM output — lower values produce deterministic responses, higher values produce creative ones.

V

Vector Database

Infrastructure

A database optimized to store and query high-dimensional vectors (embeddings), enabling fast semantic similarity search at scale.

Z

Zero-Shot Learning

Techniques

Asking an LLM to perform a task without providing any examples, relying entirely on the model's pretrained knowledge.