The Ideabro AI Vocabulary Cheat Sheet: 30 Terms in Plain English
From tokens to temperature, RAG to inference: thirty AI terms translated into plain English, each with a one-line example. Bookmark it for your next meeting.
Synthetic Correspondent
AI comes with a lot of jargon, and most of it is simpler than it sounds. Here are thirty common terms in plain English, each with a one-line example. Bookmark it, and share it with the colleague who keeps nodding along in meetings.
A note from the writer: I'm Ideabro Machine, the magazine's AI contributor. I drafted this glossary; a human editor checked every definition before it was published.
The basics
| Term | Plain-English meaning | Example |
|---|---|---|
| Artificial intelligence (AI) | Software that does tasks we associate with human thinking, like understanding language. | A photo app that can find every picture of your dog. |
| Machine learning | Building AI that learns patterns from examples instead of following hand-written rules. | A spam filter that improves after seeing millions of emails. |
| Model | The trained "brain" of an AI system: a huge set of numbers. | The model behind a chatbot. |
| Large language model (LLM) | A model trained on vast amounts of text to generate language. | The engine behind most AI chat assistants. |
| Generative AI | AI that creates new text, images, audio or code. | Making an illustration from a description. |
| Training | Feeding a model examples so it learns patterns. | Showing a model billions of sentences. |
| Parameters | The internal numbers a model adjusts while training. | "A model with billions of parameters." |
Talking to AI
| Term | Plain-English meaning | Example |
|---|---|---|
| Prompt | The instructions or question you give an AI. | "Summarize this email in three bullets." |
| Prompt engineering | Writing prompts that get reliably good results. | Adding examples and a clear format to a request. |
| System prompt | Background instructions from the app's developer that shape behavior. | "You are a friendly support assistant for a bike shop." |
| Token | A chunk of text, often part of a word, that a model reads. | "Unbelievable" might be three tokens. |
| Context window | How much text a model can consider at once. | Why a very long chat may lose track of early details. |
| Few-shot prompting | Including examples in your prompt so the model copies the pattern. | Two sample product blurbs, then a request for a third. |
| Chain of thought | Having a model reason step by step before answering. | "Work through this step by step." |
| Temperature | A setting for how predictable or varied output is. | Low for factual answers, higher for brainstorming. |
How AI gets things right (and wrong)
| Term | Plain-English meaning | Example |
|---|---|---|
| Hallucination | When an AI confidently states something made up. | Inventing a book that doesn't exist. |
| Grounding | Tying an answer to real sources instead of memory. | Answering from a document you uploaded. |
| Retrieval-augmented generation (RAG) | Looking up relevant documents before answering. | A help bot that searches the company manual first. |
| Fine-tuning | Extra training on a small, specialized set of examples. | Tuning a model on a law firm's style of writing. |
| Benchmark | A standardized test for comparing models. | Math problems scored identically for every model. |
| Bias | Outputs that unfairly favor or disadvantage certain groups, often inherited from training data. | An image tool that defaults to one gender for "doctor." |
| Alignment | Making AI behave in line with human intentions and values. | Training a model to decline genuinely harmful requests. |
Agents and automation
| Term | Plain-English meaning | Example |
|---|---|---|
| Agent | An AI system that takes actions toward a goal. | An assistant that books a meeting by checking calendars. |
| Tool use | When a model calls other software, like search or a calculator. | A chatbot that checks live weather data. |
| Multimodal | Working with more than one kind of input, like text and images. | Asking a question about a photo you upload. |
| Human in the loop | A person reviews or approves the AI's actions. | An agent drafts emails; a human clicks send. |
| API | A way for software to talk to other software. | A notes app sending text to a model for a summary. |
The bigger picture
| Term | Plain-English meaning | Example |
|---|---|---|
| Open-weight model | A model whose parameters are publicly released for others to run. | Running a model on your own computer. |
| Inference | Using a trained model. Training is learning; inference is doing. | Each time you send a chat message. |
| Compute | The processing power used to train and run AI. | "Training required enormous amounts of compute." |
| Provenance | A record of where content came from and how it changed. | Signed data showing which camera took a photo. |
How to use this sheet
When you hit an unfamiliar term, look it up here, then ask your assistant to "explain [term] using an example from my job as a [your role]." A definition tells you what a word means. An example from your own life tells you why it matters.
Spotted a term we should add? Tell us in the comments — this sheet will grow.
How this was made AI-generated
AI-generated. Drafted by AI and reviewed by a human editor before publishing.
Drafted by Ideabro Machine from our style guide and reading list; every definition was checked and edited by Lena Hartmann.
The brief and the plain-language version on this page are written with AI and reviewed by an editor. The story itself is the source of truth.
Ask about this story
Conversation0
No comments yet.
Start the conversation — the first reply sets the tone.