← Back to LetsDoItAi
LetsDoItAiHome › AI Glossary

The AI glossary

AI terms in plain English. Short, clear definitions of the words you'll see again and again, without the unnecessary jargon.

Large language model (LLM)

The technology behind chat tools like ChatGPT and Claude. It has learned patterns from huge amounts of text and predicts the most likely next word to build its answers.

Prompt

The instruction you give an AI tool. A clear, specific prompt gets a far better result than a vague one.

Token

The small chunks of text an AI reads and writes, roughly a word or part of a word. Pricing and length limits are often measured in tokens.

Hallucination

When an AI states something false with total confidence. Because it predicts what sounds right rather than looking up facts, it can invent names, dates and sources, so always verify.

Context window

How much text an AI can 'hold in mind' at once. A larger context window means it can work with longer documents or conversations without forgetting the start.

Agent

An AI that doesn't just answer but carries out multi-step tasks for you, such as browsing, using apps or completing a workflow with less hand-holding.

Multimodal

An AI that can handle more than one type of input or output, such as text, images, audio and files together in a single conversation.

Generative AI

AI that creates new content, text, images, audio, code, rather than just analysing or sorting existing data. It's the category most consumer AI tools fall into.

Model

The specific AI 'brain' a tool uses. Companies release new, more capable models over time, which is why the same tool can get noticeably better.

Prompt engineering

The skill of writing prompts that get good results, giving context, being specific, adding examples and refining. It's the single habit that most improves your output.

Fine-tuning

Training an existing model a bit further on specialised data so it performs better at a particular task or in a particular style.

Retrieval-augmented generation (RAG)

A technique where the AI looks up relevant information (from the web or your documents) before answering, so its reply is grounded in real sources.

Open-source model

A model whose code and weights are freely available, so anyone can run or adapt it, sometimes even on their own computer. DeepSeek and Mistral are examples.

Inference

The act of the AI actually producing an answer from your prompt. When people talk about the 'cost of inference', they mean the computing cost of running the model.

Training data

The huge collection of text or images a model learned from. Its strengths, blind spots and knowledge cut-off all come from this data.

Knowledge cut-off

The point after which a model wasn't trained on any new information. Unless it can search the web, it won't know events after this date.

Chatbot

A tool you interact with by typing or speaking in a conversation. ChatGPT, Claude and Gemini are the best-known AI chatbots.

API

A way for developers to plug an AI model into their own apps and products, rather than using it through a chat window.

← Back to LetsDoItAi
Cookie settings