Glossary — AI Terms Explained Simply
AI is often discussed in difficult terms that make it sound more complicated than it is. Here we explain the most common terms simply.
A
AI (Artificial Intelligence) A program that learns from data and performs tasks that traditionally require human intelligence — such as writing, translating, or image recognition.
Agent An AI that can act on its own — not just answer questions, but complete multi-step tasks. For example, an AI agent can browse the web, compare prices, and book a flight. Think of it as the next step up from a chatbot: you give it a goal, and it figures out how to get there.
Algorithm A rule or set of instructions that a computer follows to do something. For example, Google Maps uses an algorithm to calculate a route. An algorithm isn’t AI — it’s just a precise recipe.
API An interface that lets different programs talk to each other. For example, when a weather app shows you the forecast, it fetches data from another program’s API.
B
Bias (AI bias) When AI favors or discriminates against a certain group because its training data has been skewed. For example, if AI is trained only on texts written by men, it may write in a “masculine” style.
C
Chatbot A program you can have a text conversation with. ChatGPT, Claude, and similar tools are chatbots. It’s essentially an interface — the AI itself is the model running behind it.
ChatGPT One of the most popular AI chat programs. Made by a company called OpenAI. You can type text to it and it responds like a human.
Claude An AI chat program that’s especially good at reading and analyzing long texts. Made by a company called Anthropic.
Copilot Microsoft’s AI assistant. Available for Windows and Microsoft Office programs, but no longer mandatory — you can install or remove it as you like.
D
Deepfake A fake video or image created with AI that looks real but isn’t. Can be used for both entertainment and scams.
Discriminative AI AI that classifies or predicts (e.g., whether an email is spam or not).
G
Gemini Google’s AI. Similar to ChatGPT, but integrated into Google’s services.
Generative AI AI that creates new content — text, images, audio, or video. ChatGPT, Midjourney, and similar tools are generative AI.
H
Hallucination When AI makes things up and presents them confidently as facts. For example, AI might claim an obscure historical figure lived in 1850. Hallucinations are the most common reason you shouldn’t trust AI answers blindly.
L
LLM (Large Language Model) A large language model. AI that has learned from a massive amount of text and can therefore write and understand language. ChatGPT is an LLM. When someone says “language model,” they usually mean an LLM.
M
Midjourney An AI that creates images from text descriptions. You type “a cat on a space station” and it draws the image.
Model A trained AI that has been taught to do specific things. For example, GPT-4 is a model that can write text.
N
Neural network A program that mimics the structure of the human brain. It consists of thousands or millions of “neurons” — processing units that handle information. All modern AI is based on neural networks.
O
OpenAI The company that made ChatGPT. One of the best-known AI companies in the world.
P
Prompt A message or instruction you give to AI. A good prompt = a good answer. For example: “Write a short email to my boss requesting a day off.”
Prompting The art of writing good instructions for AI. The more precisely you describe what you want, the better AI can help.
R
Reinforcement Learning A way of training AI by giving it rewards when it does something right and penalties when it does something wrong. For example, an AI learning to play a video game gets points when it progresses and loses points when it makes a mistake.
T
Token AI’s “word.” When AI reads or writes text, it breaks it into tokens. A token doesn’t always equal one word — in some languages, one word can be multiple tokens. This has no practical significance for regular users, but it affects pricing and speed.