AI Basics· 9 min read

AI Glossary for Beginners

AI comes with a lot of jargon, and most of it sounds harder than it is. This is a plain-English glossary of the terms you keep running into. Each one is a short, clear definition you can actually understand. Where a term deserves a fuller explanation, there is a link to a deeper guide.

The big picture

Artificial intelligence (AI)

Software that learns patterns from examples and makes its own guesses, instead of following fixed rules a person wrote. The broad umbrella term. Read more

Machine learning

The main method behind modern AI: software that learns from examples instead of being given rules. It sits inside AI. Read more

Deep learning

A powerful kind of machine learning, loosely inspired by how brain cells connect. It made recent leaps in language and images possible.

Generative AI

Any AI that creates new things, like text, images, or music, rather than just sorting or labeling. Chatbots are generative AI.

Large language model (LLM)

An AI built for text. It learned from huge amounts of writing and predicts words. ChatGPT, Claude, and Gemini are LLMs.

How it works

Token

A small chunk of text a model reads, usually a short word or part of a word. Length and price are measured in tokens. Read more

Prompt

The message or instruction you give an AI. A clearer prompt gets a better answer. Read more

Next-word prediction

The core trick behind chatbots: they build an answer by guessing the next word, one at a time. Read more

Training

The process where a model learns, by guessing, checking against real answers, and adjusting, across huge amounts of data. Read more

Training data

The examples a model learns from. A model is only as good, and as fair, as the data it was trained on.

Parameters

The internal settings a model adjusts during training. More parameters roughly means more capacity to learn, though bigger is not always better.

Inference

The moment a trained model is actually used to produce an answer. Training is learning; inference is doing.

Embedding

A way of turning words into numbers that capture meaning, so the AI can measure how related two things are. Read more

Context window

How much text a model can look at once, measured in tokens. Go over it and the oldest text drops out of view.

Neural network

The layered structure at the heart of deep learning, loosely modeled on connected brain cells.

Prompting terms

Prompt engineering

The skill of writing clear instructions that get better, more reliable answers from AI. Read more

System prompt

A behind-the-scenes instruction that sets how the AI should behave for a whole conversation. Read more

Zero-shot

Asking the AI to do a task with no examples, just the instruction. Read more

Few-shot

Giving the AI a few examples of what you want before your real request, which improves consistency. Read more

Chain-of-thought

Asking the model to reason step by step before answering, which improves accuracy on harder problems. Read more

Role prompting

Telling the AI who to act as, like "act as an editor," to shape its tone and focus. Read more

Context

The background you give a prompt, like the audience and goal, so the model does not have to guess. Read more

Quality and safety

Hallucination

When an AI confidently states something false or made up. It happens because the model writes what sounds right, not what it checked. Read more

AI detector

A tool that guesses whether text was written by AI, by measuring how smooth and predictable it is. Read more

False positive

When an AI detector wrongly flags human writing as AI. Plain, even writing gets flagged most. Read more

Bias

When a model reflects unfair patterns from its training data, treating some groups or ideas differently.

Fine-tuning

Extra training that adapts a general model to a specific job or style, using a focused set of examples.

RAG (retrieval-augmented generation)

Giving a model real documents to pull from before it answers, which reduces made-up facts.

Temperature

A setting that controls how random a model’s word choices are. Lower is more predictable; higher is more varied.

Multimodal

An AI that handles more than just text, such as images, audio, or video, alongside words.

How to keep learning

You do not need to memorize any of this. The terms make far more sense once you see how the pieces fit together. If you want the fuller picture, start with what artificial intelligence is, then read how AI predicts words, and the rest of the glossary will click into place.

Put the words into practice

Deepclario helps you write clearer prompts and check whether text looks AI-written. Free, no account needed.

Try the prompt improver →