AI terms explained in plain English
No account required. Clear, practical information created for everyday users, freelancers and small teams.
35 concise definitions of essential AI, security, prompting and automation terms.
AI (artificial intelligence)
Software techniques that perform tasks associated with human intelligence.
AI agent
A system designed to plan or perform tasks, sometimes using external tools.
AI assistant
Software that helps users answer questions or complete tasks.
API
An interface that lets software programs communicate.
Automation
Using software to perform steps with limited manual effort.
Benchmark
A defined test used to compare performance, often with important limitations.
Bias
A systematic skew in data, model behavior or outcomes.
Chatbot
A conversational interface that responds to messages.
Context window
The amount of information a model can consider in one interaction.
Dataset
A structured collection of examples or records.
Embedding
A numerical representation used to compare semantic similarity.
Evaluation (eval)
A procedure for measuring behavior against defined criteria.
Fine-tuning
Adapting a model through additional training on selected examples.
Foundation model
A broadly trained model that can support multiple downstream tasks.
Generative AI
AI that generates text, images, audio, code or other content.
Hallucination
A plausible-sounding model output that is inaccurate or unsupported.
Human-in-the-loop
A workflow where humans review or approve parts of an automated process.
Inference
The process of using a trained model to produce an output.
Input token
A segment of input text processed by a model.
Large language model (LLM)
A model trained to predict and generate language-like sequences.
Latency
Time between making a request and receiving a response.
Multimodal
Able to process or generate more than one kind of data.
OCR
Optical character recognition: reading printed or handwritten text from images.
Open source
Software or other material distributed under terms that permit specified forms of use and modification.
Output token
A segment of content produced by a language model.
Prompt
Instructions or context supplied to an AI model.
Prompt injection
Untrusted content attempting to override an AI system’s intended instructions.
RAG
Retrieval-augmented generation: supplying retrieved sources as context for a response.
Rate limit
A restriction on how frequently an API or service can be used.
Red teaming
Structured adversarial testing for weaknesses or harmful behavior.
Retrieval
Finding potentially relevant material from stored sources.
System prompt
Higher-priority instructions that shape the operation of an AI assistant.
Temperature
A generation setting that influences sampling variation in some models.
Tool calling
A model requesting use of a connected function or external system.
Vector database
A system that stores and searches numerical vector representations.
Definitions are simplified for learning. Consult original technical documentation for implementation or compliance decisions.