AI Glossary: Key Terms Worth Knowing
You’re reading an article about AI and you hit “LLM,” “RAG,” or “fine-tuning.” You open another tab to look it up — and there’s another new term waiting for you.
This glossary is your lifeline. Every term is explained plainly, with no technical background assumed, and where it makes sense, it links to an article where you can see it in action.
Terms are ordered from basic to more advanced — not alphabetically.
Core concepts
How to write prompts in practice — with templates and examples, step by step — is covered in How to write prompts.
How AI works
AI behavior and limitations
More advanced concepts
Adapting an existing model on specific data. The result behaves as if it already knows the style and terminology of a particular organization.
Before answering, the AI searches a document database and builds its response on concrete sources — not just learned patterns.
A model that works with multiple input types at once — text, images, audio, video. GPT-4o and Gemini are multimodal.
AI capable of autonomously carrying out sequences of steps — searching the web, running code, interacting with tools — without asking for confirmation at every step.
Inpainting
An AI image-editing technique where the model “fills in” a missing or removed part of an image based on the surrounding context. Used for example when removing a watermark, erasing objects from a photo, or editing part of an image. Covered in detail in the AI image editing hub.
System prompt
An instruction inserted into a conversation before your first message, which defines the AI’s behavior — its role, tone, constraints, and response format. Users typically never see it; it is set by the tool’s operator. This is why ChatGPT behaves differently from AI Chat in other applications, even when both run on the same underlying model.
Quick acronym reference
| Acronym | Full name | In brief |
|---|---|---|
| AI | Artificial Intelligence | Machines that learn and reason |
| LLM | Large Language Model | AI trained on massive text datasets |
| RAG | Retrieval-Augmented Generation | AI + document search combined |
| GPT | Generative Pre-trained Transformer | The architecture behind ChatGPT |
| API | Application Programming Interface | How to connect AI to your own software |
| SaaS | Software as a Service | Cloud-based subscription tool |
Missing a term? The best way to understand it is to ask directly — AI will explain any concept in plain language.
Ask AI directly
Copy any term from this glossary and send it to AI: "Explain [term] to me as if I knew absolutely nothing about it. Include a real-life example." You'll see exactly how it works in practice.