Skip to content
Last checked: October 2026.
AI comes with a lot of jargon. This glossary explains 63 of the most common terms in plain English, grouped by topic, including the agent terms you see everywhere now.
Jump to a section
- Basics
- Language models
- Training and data
- AI agents
- Images, audio and video
- Safety, ethics and business
Basics
| Term | What it means |
|---|---|
| Artificial intelligence (AI) | Computer systems that do tasks that normally need human thinking, like understanding language or recognizing images |
| Machine learning (ML) | A way of building AI where systems learn patterns from data instead of following hand-written rules |
| Deep learning | Machine learning that uses neural networks with many layers |
| Neural network | A model made of connected layers of simple units, loosely inspired by the brain |
| Algorithm | A set of steps a computer follows to solve a problem |
| Model | The trained system that makes predictions or generates output |
| Training data | The examples a model learns from |
| Parameters | The internal numbers a model adjusts during training; large models have billions |
| Generative AI | AI that creates new text, images, audio, video or code |
| Narrow AI | AI built for one kind of task, like spam filtering |
| AGI (artificial general intelligence) | A hypothetical AI that can match people across almost any task |
Language models
| Term | What it means |
|---|---|
| Large language model (LLM) | A model trained on huge amounts of text to understand and generate language, such as GPT, Claude or Gemini |
| Transformer | The neural network design behind most modern language models |
| Token | A small chunk of text, often part of a word, that models read and write; usage is often priced per token |
| Context window | How much text a model can consider at once, measured in tokens |
| Prompt | The instruction or question you give an AI |
| System prompt | Hidden instructions that set how an AI behaves |
| Prompt engineering | Writing prompts carefully to get better results |
| Temperature | A setting that controls how predictable or creative output is |
| Hallucination | When AI states something false with confidence |
| Reasoning model | A model that works through a problem step by step before answering |
| Multimodal | Able to handle more than one type of input or output, like text and images |
| Foundation model | A large general model that other tools are built on |
| Open-weight model | A model whose trained parameters are published so anyone can run it |
| Small language model (SLM) | A compact model that can run on a phone or laptop |
Training and data
| Term | What it means |
|---|---|
| Pre-training | The first, large-scale training of a model on general data |
| Fine-tuning | Extra training on a smaller, focused dataset to specialize a model |
| RLHF | Reinforcement learning from human feedback: tuning a model using people’s ratings of its answers |
| Embedding | A list of numbers that captures the meaning of text or images so they can be compared |
| Vector database | A database built to search embeddings by meaning |
| RAG (retrieval-augmented generation) | Letting a model look up relevant documents before answering, to improve accuracy |
| Synthetic data | Artificial data created to train or test models |
| Benchmark | A standard test used to compare models |
| Inference | Running a trained model to get an answer |
| Overfitting | When a model memorizes training data and does poorly on new data |
AI agents
| Term | What it means |
|---|---|
| AI agent | An AI system that can plan and take actions to complete a goal, not just answer questions |
| Agentic AI | AI designed to act on its own across several steps |
| Tool use (function calling) | When a model calls outside tools, such as search, a calendar or a database |
| MCP (Model Context Protocol) | An open standard for connecting AI models to apps and data sources |
| Multi-agent system | Several AI agents working together, each with its own role |
| Orchestration | Coordinating models, agents and tools in a workflow |
| Human in the loop | A person reviews or approves an AI’s actions |
| Memory | An agent’s ability to remember earlier conversations or facts |
| Computer use | An AI controlling a computer screen with clicks and typing |
| Guardrails | Rules and checks that keep AI from unsafe or unwanted actions |
| Prompt injection | An attack that hides instructions in content to hijack an AI |
Images, audio and video
| Term | What it means |
|---|---|
| Computer vision | AI that understands images and video |
| Diffusion model | A type of model that creates images by gradually removing noise |
| Text-to-image | Creating images from a written description |
| Text-to-video | Creating video clips from a written description |
| Text-to-speech (TTS) | Turning text into spoken audio |
| Speech recognition | Turning spoken audio into text |
| Voice cloning | Copying a person’s voice with AI |
| Deepfake | Realistic fake media of a real person made with AI |
Safety, ethics and business
| Term | What it means |
|---|---|
| Bias | Unfair results caused by skewed data or design |
| Alignment | Making AI pursue the goals people actually intend |
| Explainability | How well people can understand why an AI made a decision |
| Red teaming | Deliberately attacking an AI to find weaknesses |
| Watermarking | Hidden marks that identify AI-generated content |
| AI governance | The policies and controls an organization uses to manage AI |
| GEO (generative engine optimization) | Optimizing content so it appears in AI answers, like ChatGPT or Google AI Overviews |
| Copilot | An AI assistant built into an app to help you work |
| API | A way for software to talk to an AI model or service |
FAQs
What is the difference between AI and machine learning?
AI is the broad goal of making computers do smart tasks. Machine learning is the most common way to get there, by learning from data.
What is the difference between a chatbot and an AI agent?
A chatbot answers questions. An agent can also take actions, such as booking a meeting, updating a spreadsheet or replying to a customer.
What does LLM stand for?
Large language model, the type of AI behind tools like ChatGPT, Claude and Gemini.
Ready to see AI agents in action? Browse the Klyvoo directory of AI agents, sorted by the job you need done.