Skip to content
AI Lakay Nou

Les mots de l'IA

Tous les mots pour parler de l'IA, expliqués simplement.

Aucun mot ici n'est expliqué à l'aide d'un mot plus difficile. Si une définition vous trouble, l'erreur est la nôtre, pas la vôtre. Your Brain Is the Boss™

95 mots au total

A

Grandes idées

Access

Pas encore traduit

Being able to actually use a tool — because it exists in your language, on your device, at a price you can pay.

Rester honnête

Accountability

Pas encore traduit

Being answerable for the results, even when a machine helped produce them. A person is always responsible, never the tool.

À retenir

The AI does not sign the letter — you do.

Types d'IA

Agent

Pas encore traduit

An AI given a goal and permission to take steps by itself — search, click, send — to reach it.

Les bases de l'IA

Algorithm

Pas encore traduit

A set of clear steps a computer follows to solve a problem or make a choice.

Exemple

A recipe is an algorithm for cooking.

Sécurité et vie privée

Anonymous

Pas encore traduit

Without a name attached. True anonymity online is rare; assume you can be found.

Les bases de l'IA

Artificial Intelligence

Pas encore traduit

Computer systems that do tasks that usually need human thinking, like writing, reading, or recognizing pictures.

À retenir

AI is a tool, not a person.

Rester honnête

Attribution

Pas encore traduit

Saying clearly who or what made a piece of work — including when AI helped.

Les bases de l'IA

Automation

Pas encore traduit

When a machine does a task on its own that a person used to do by hand.

B

Quand l'IA se trompe

Bias

Pas encore traduit

When the AI answers unfairly because its training data leaned one way — one language, one country, one point of view.

C

Types d'IA

Chatbot

Pas encore traduit

A program you talk to in writing, that answers you in writing.

Rester honnête

Citation

Pas encore traduit

Naming the exact source of a fact, quote, or idea so a reader can check it.

Les bases de l'IA

Computer

Pas encore traduit

A machine that stores information and follows instructions very fast.

Quand l'IA se trompe

Confabulation

Pas encore traduit

Another name for hallucination: filling a gap in memory with something invented.

Quand l'IA se trompe

Confidence Trap

Pas encore traduit

The habit of trusting an AI more because it wrote its wrong answer in smooth, polite sentences.

Sécurité et vie privée

Consent

Pas encore traduit

Saying yes, in a way that is free and informed, before someone uses your data or your image.

Rester honnête

Consent to Use

Pas encore traduit

Getting permission before you feed someone's writing, voice, or face into an AI.

Parler à l'IA

Context

Pas encore traduit

Everything the AI can see right now — your prompt, the earlier messages, any file you shared.

Parler à l'IA

Conversation

Pas encore traduit

A back-and-forth session where the AI remembers what was just said.

Sécurité et vie privée

Cookie

Pas encore traduit

A small file a website leaves on your device to remember you. Some cookies are useful, some follow you.

D

Les bases de l'IA

Data

Pas encore traduit

Facts, numbers, words, or pictures that a computer can read and count.

Sécurité et vie privée

Data Breach

Pas encore traduit

When information a company was supposed to protect ends up in the wrong hands.

Comment l'IA apprend

Deep Learning

Pas encore traduit

Machine learning that uses many layers of small math steps stacked on top of each other.

Sécurité et vie privée

Deepfake

Pas encore traduit

A fake video, image, or voice made by AI to look or sound like a real person.

À retenir

If it feels wrong, slow down and verify before you share.

Les bases de l'IA

Digital

Pas encore traduit

Anything stored as numbers a computer can read: files, photos, voice notes.

Grandes idées

Digital Divide

Pas encore traduit

The gap between people who have fast internet, good devices, and training, and those who do not. AI can widen this gap or close it, depending on who builds it.

Rester honnête

Disclosure

Pas encore traduit

Telling the reader or listener that AI was used to make what they are seeing or hearing.

E

Sécurité et vie privée

Encryption

Pas encore traduit

Scrambling information so only the right person can read it.

Quand l'IA se trompe

Error

Pas encore traduit

A wrong answer. AI makes errors even when it sounds sure.

Grandes idées

Ethics

Pas encore traduit

The rules we set for ourselves about right and wrong — including how we build and use AI.

F

Types d'IA

Facial Recognition

Pas encore traduit

An AI that tries to match a face in a photo or video to a name.

Rester honnête

Fact-Check

Pas encore traduit

Comparing a claim to a trusted source before you believe it or repeat it.

Grandes idées

Fairness

Pas encore traduit

Treating people evenly. An AI is fair only if it works as well for the least-served group as for the most-served.

Comment l'IA apprend

Feature

Pas encore traduit

One thing the AI pays attention to about an example, like the color of a fruit or the length of a sentence.

Parler à l'IA

Follow-up

Pas encore traduit

A shorter next message that builds on the answer you just got.

G

Types d'IA

Generative AI

Pas encore traduit

AI that produces new text, images, sound, or video instead of just picking from a list.

H

Quand l'IA se trompe

Hallucination

Pas encore traduit

When the AI writes something that sounds true and confident but is simply false — a fake quote, a fake source, a fake fact.

Exemple

An AI invents a book title that does not exist and attributes it to a real author.

À retenir

Confidence is not proof. Check before you share.

Grandes idées

Human Oversight

Pas encore traduit

A person watching what the AI does and stepping in when it goes wrong.

I

Types d'IA

Image Generator

Pas encore traduit

An AI that draws a picture from a written description.

Les bases de l'IA

Input

Pas encore traduit

What you give the AI: a question, a photo, a voice clip, a file.

Parler à l'IA

Instruction

Pas encore traduit

A clear command inside your prompt, like "answer in one sentence" or "use simple words".

K

Grandes idées

Kreyòl AI

Pas encore traduit

AI that reads, writes, and reasons in Kreyòl Ayisyen as a first-class language — not as an afterthought translated from English.

L

Comment l'IA apprend

Label

Pas encore traduit

The correct answer attached to a training example, so the AI knows what right looks like.

Types d'IA

Large Language Model

Pas encore traduit

An AI trained on huge amounts of text so it can read and write in many languages.

Exemple

ChatGPT, Claude, and Gemini are large language models.

Grandes idées

Literacy

Pas encore traduit

The ability to read, question, and use a technology on your own terms. AI literacy is the goal of this glossary.

M

Les bases de l'IA

Machine

Pas encore traduit

The physical or online device where the AI runs.

Comment l'IA apprend

Machine Learning

Pas encore traduit

A way of building AI where the computer learns from examples instead of being told every rule.

Quand l'IA se trompe

Missing Context

Pas encore traduit

When the answer is technically true but leaves out the piece that would change your decision.

Quand l'IA se trompe

Mistake

Pas encore traduit

Any answer that does not match reality. A mistake by a machine is still a mistake.

Les bases de l'IA

Model

Pas encore traduit

The finished shape of what an AI has learned. You send it a question and it sends back an answer.

N

Comment l'IA apprend

Neural Network

Pas encore traduit

A math structure loosely inspired by brain cells, used in most modern AI. It is not a brain.

Grandes idées

NI + AI™

Pas encore traduit

Natural Intelligence first, Artificial Intelligence second. Use your own mind to lead, and let the machine help — never the other way around.

O

Rester honnête

Original Work

Pas encore traduit

Something you thought through and wrote yourself, even if you used AI to check spelling or brainstorm.

Quand l'IA se trompe

Outdated Information

Pas encore traduit

An answer based on the world as it was months or years ago, not as it is today.

Les bases de l'IA

Output

Pas encore traduit

What the AI gives back to you: text, an image, a sound, a number.

Quand l'IA se trompe

Overfitting

Pas encore traduit

When the AI has memorized its examples too closely and stumbles on anything new.

P

Comment l'IA apprend

Parameter

Pas encore traduit

One small number inside the model that gets adjusted during training. A big model has billions of them.

Sécurité et vie privée

Password

Pas encore traduit

A secret word or phrase that unlocks your account. Long and unique beats clever.

Les bases de l'IA

Pattern

Pas encore traduit

Something that repeats. AI learns by finding patterns in many examples.

Sécurité et vie privée

Personal Data

Pas encore traduit

Anything that points to you: your name, your face, your voice, your address, your phone.

À retenir

Do not paste personal data into an AI unless you have to.

Sécurité et vie privée

Phishing

Pas encore traduit

A fake message that tries to trick you into giving up a password or clicking a bad link. AI now writes very convincing ones.

Rester honnête

Plagiarism

Pas encore traduit

Passing off someone else's words or ideas as your own. Copying from AI without saying so is plagiarism too.

Les bases de l'IA

Prediction

Pas encore traduit

The AI's best guess about what comes next or what an answer should be. A guess, not proof.

Sécurité et vie privée

Privacy

Pas encore traduit

Your right to decide who sees your information and what they do with it.

Les bases de l'IA

Program

Pas encore traduit

A written set of steps that tells a computer what to do.

Parler à l'IA

Prompt

Pas encore traduit

What you write or say to the AI to tell it what you want.

Exemple

"Write a two-line birthday message for my mother" is a prompt.

R

Types d'IA

Recommendation System

Pas encore traduit

An AI that suggests what to watch, read, buy, or follow next based on what you and others have chosen before.

À retenir

The suggestion is not a friend. It is a guess based on clicks.

Comment l'IA apprend

Reinforcement Learning

Pas encore traduit

Learning by trying things, getting a small reward for good moves, and adjusting.

Parler à l'IA

Response

Pas encore traduit

The AI's reply to your prompt.

Parler à l'IA

Rewrite

Pas encore traduit

Asking the AI to say the same thing in a different way — shorter, warmer, more formal.

Types d'IA

Robot

Pas encore traduit

A physical machine that senses, decides, and moves. Only some robots use AI.

Parler à l'IA

Role

Pas encore traduit

Telling the AI who to act as, like "a patient teacher" or "a careful editor".

S

Sécurité et vie privée

Scam

Pas encore traduit

A trick designed to take your money, your data, or your trust.

Types d'IA

Search Engine

Pas encore traduit

A tool that looks through the web and shows you pages that match your question.

Types d'IA

Self-Driving Car

Pas encore traduit

A car that uses cameras, sensors, and AI to drive without a person steering.

Quand l'IA se trompe

Silent Failure

Pas encore traduit

When the AI gives you an answer instead of saying "I don't know", even though it does not know.

Les bases de l'IA

Software

Pas encore traduit

Instructions written for a computer. An app is software.

Rester honnête

Source

Pas encore traduit

Where a fact comes from. "An AI told me" is not a source.

Grandes idées

Sovereignty

Pas encore traduit

The right of a people to decide, in their own language and on their own land, how a technology will be used among them.

Parler à l'IA

Summarize

Pas encore traduit

Asking the AI to give you the short version of a long text.

Comment l'IA apprend

Supervised Learning

Pas encore traduit

Learning from examples that already have the right answer marked on them.

Parler à l'IA

System Prompt

Pas encore traduit

A hidden instruction from the developer that shapes how the AI answers everyone.

T

Parler à l'IA

Token

Pas encore traduit

A small piece of text — a word or a part of a word — that the AI counts as one unit.

Comment l'IA apprend

Training

Pas encore traduit

The long practice period where an AI studies many examples so it can answer later.

Comment l'IA apprend

Training Data

Pas encore traduit

The examples an AI learns from. If Kreyòl is missing from the training data, the AI will speak Kreyòl badly. What is left out matters as much as what is put in.

À retenir

An AI only knows what it was shown.

Parler à l'IA

Translate

Pas encore traduit

Asking the AI to move a text from one language into another.

Types d'IA

Translation Tool

Pas encore traduit

An AI that turns text or speech from one language into another.

Rester honnête

Transparency

Pas encore traduit

Being open about how a decision was made and what the AI did or did not do.

Sécurité et vie privée

Two-Factor Authentication

Pas encore traduit

A second check besides your password — a code by text, an app tap — that keeps thieves out.

U

Quand l'IA se trompe

Underfitting

Pas encore traduit

When the AI has not learned enough from the examples and misses obvious patterns.

Comment l'IA apprend

Unsupervised Learning

Pas encore traduit

Learning by grouping examples together without being told the right answer.

V

Types d'IA

Voice Assistant

Pas encore traduit

An AI you speak to out loud that speaks back, like Siri or Alexa.

W

Rester honnête

Watermark

Pas encore traduit

A visible or hidden mark that shows a piece of media was made or edited by AI.

Comment l'IA apprend

Weight

Pas encore traduit

A number that says how much one piece of information counts when the AI makes a decision.

Quand l'IA se trompe

Wrong Source

Pas encore traduit

When the AI names a source that does not exist, or a real source that does not say what the AI claims.

Y

Grandes idées

Your Brain Is the Boss™

Pas encore traduit

The final decision, the final word, the final signature belongs to the human. The AI advises; you decide.