THE POINT / WHAT CHANGES FOR YOU

From news to practice.

Understanding the terminology helps ask better questions and assess promises. Start from the glossary and try the examples, one term at a time.

Prompt, model, token, context

Prompt: the instruction or material provided to a system. Model: the trained component that produces the output. Token: a small unit into which textual information is divided; it does not always coincide with a word. Context: the material the system can consider during processing. These are operational definitions, useful for understanding functions without confusing them with the commercial name of an application.

Generative, agent, hallucination, RAG

Generative AI: produces synthetic content, for example texts or images. Agent: a system that can perform actions towards a goal with a certain degree of autonomy. Hallucination: an inaccurate or invented output presented in a plausible way. RAG: an approach that retrieves documents and makes them available to the model to help it respond. Source retrieval alone does not guarantee correctness.

An example of a useful request

"Summarise this document in five points, separate facts from assumptions, and indicate the steps supporting each conclusion." This request is more verifiable than a generic "explain everything." Before allowing an agent to send a message or modify data, it is advisable to clarify the destination, limits, and approvals. The entries collected here are explanations from the editorial staff, with references to institutional glossaries.

The complete glossary, from A to Z

84 entries, definitions, examples and search by word or topic.

Explore the AI glossary →

Sources and documents

Document date or reference period: NIST and UTHSC glossaries · consulted on 23 October 2026.

Text prepared with AI assistance based on the indicated sources. Interpretive evaluations are distinct from data and statements from sources.

Report an update or correction →