What generative AI is, and your first successful exchange
This lesson is still being tested: it improves with learners' feedback. If something is unclear, it is the lesson that needs fixing, not you.
By the end of this lesson, you will be able to
- explain in two sentences what a generative AI is;
- open a free account on an AI assistant;
- get a useful answer to a real small question from your week;
- notice when the AI is wrong with great confidence.
Why it matters to you
You have to write a delicate message, make sense of a complicated letter, organize a dinner, find ideas, prepare a list… An AI can help you do all of that faster, as long as you know how to talk to it and check what it says. That is exactly what you are going to learn, step by step.
Before you start
| You need | A computer (or a phone), a web browser, an email address; for Claude, a phone that can receive text messages |
| Cost | €0: a free account is enough for this lesson |
| Duration | 30 to 45 minutes |
| What you never give to an AI | Your passwords and codes, your numbers (social security, bank card, IBAN), health information, private information about other people |
Understand it in 3 minutes
A generative AI produces text by predicting, word after word, what is most likely to fit (in reality, piece of a word after piece of a word). Some also create images or sounds, with similar methods. It learned to do this by analyzing an enormous amount of text. It does not “know” things the way an encyclopedia does: it produces a plausible answer.
The picture to keep in mind: an intern who has read everything, who answers very fast and with a lot of confidence, but who checks nothing on their own. Very useful if you explain clearly what you want, and if you proofread their work.
A chatbot is the conversation window that gives you access to an AI. Claude (from the company Anthropic) and ChatGPT (from the company OpenAI) are two well-known chatbots. This course mostly uses Claude, but the method works with both.
| What it does well | What it does badly |
|---|---|
| Writing, rephrasing, summarizing, translating | Giving precise facts without mistakes (dates, figures, prices, names) |
| Suggesting ideas, plans, lists | Knowing what happened very recently (some chatbots search the web, like Claude on the free plan as of 24/10/2026, but the pages they find may be old or wrong) |
| Explaining a complicated subject simply | Saying “I don’t know”: it often prefers to make up an answer that sounds right |
| Adapting to the tone and format you ask for | Remembering you: some tools have a memory from one conversation to the next (Claude too on the free plan, as of 24/10/2026), but it stays partial. Don’t count on it for now: a later lesson covers it (module 2, lesson 5) |
When it makes something up with confidence, this is called a hallucination. It is not a rare bug: it is its nature. Hence the number one rule of this course: check what matters.
Do it together
1Step 1: Open a free account
- In your browser, go to claude.ai (or to chatgpt.com if you prefer ChatGPT).
- Create an account with your email address, following the instructions on the screen. For Claude, the help page says you also need a phone number that can receive text messages: a verification code is sent to it. Pick a number you will keep: it cannot be changed once verified. Claude is only for people aged 18 and over (Anthropic’s terms of service, read on 28/10/2026). ChatGPT requires you to be at least 13, or the minimum age required in your country, with permission from a parent or legal guardian if you are under 18 (OpenAI’s terms of use, read on 28/10/2026).
Claude’s free plan exists (€0, checked on 24/10/2026 on claude.com/pricing) and it lets you chat with Claude. ChatGPT also has a free plan (€0, checked on 24/10/2026 on chatgpt.com/pricing).
2Step 2: Look at the screen
Mostly you see a text box, at the bottom or in the center: that is where you write. The answer appears above it. That is all you need to get started.
3Step 3: Your first request, guided
Copy this request, paste it into the text box, then press Enter:
I'm organizing a dinner for 8 people on Saturday evening, including one vegetarian.
Suggest a simple menu I can prepare in 2 hours, with the shopping list.
Present the menu in 3 lines, then the shopping list organized by aisle.
Look at how it is built. It says 4 things:
- what you want: a menu and a shopping list;
- what it is for: a dinner for 8, Saturday evening;
- what the AI needs to know: one vegetarian, 2 hours of preparation;
- the form: the menu in 3 lines, then the list by aisle.
4Step 4: Keep the conversation going
The answer is not quite right? Don’t start over: continue the conversation. For example:
Replace the dessert with something that doesn't need an oven.
The AI keeps in mind what was said earlier in this conversation, within the limit of what it can hold. This is the context. If you open a new conversation, it starts almost from zero: only its memory and your settings (for example personal instructions), if they are turned on, give it some information about you.
Your turn
- Pick a small, concrete thing from your week, and write your request with the 4 elements: what you want, what it is for, what the AI needs to know, the form you expect. Ideas depending on what suits you:
- Personal life: a message to politely decline an invitation, in 4 lines.
- Work: turn 3 loose notes into a clear email for a colleague (with notes that contain nothing confidential).
- Creating: 5 title ideas for a post, in the tone you want.
- Building tools: have it explain, line by line, a spreadsheet formula you don’t understand.
- Send your request, then read the answer: can you use it as it is, or after a single correction?
- The truth test: ask it a question about a subject you know very well (your city, your job, your sport, your favorite series). Read the answer carefully, and look for at least one mistake, one approximation, or one overconfident claim.
Resources to download
- My assistant’s ID sheet: to fill in during module 1.
- My requests: the template for your request library, with the example from this lesson.
Check your work
Your request worked if:
- the answer matches what you asked for (the content and the form);
- you can use it right away, or after a single correction;
- you proofread what matters, without taking it on trust.
The truth test worked if you found at least one mistake or imprecision. If you found none, ask a more precise question (a date, a figure, a name): that is where the AI goes wrong most easily.
Common pitfalls:
- taking the first answer without rereading it;
- believing a figure, a date, or a price without checking the source;
- giving personal information “so that it understands better”;
- thinking it will remember everything next time.
Keep it and pick it up later
- Copy your successful request into a note (on your phone or your computer), titled “My requests”. It is the start of your library, which you will build up in module 2.
- Also write down the mistake you found in the truth test, in one sentence. It will remind you why we check.
Going further
optional
- Objectif IA (OpenClassrooms and Institut Montaigne): a course in French, about 6 hours long, to understand AI in more depth (checked on 24/10/2026).
- AI capabilities and limitations (Anthropic Academy, anthropic.skilljar.com/ai-capabilities-and-limitations): a free course, in English, about 15 minutes of video (the Claude Academy page lists about 3.5 hours in total for its 13 lessons), on what models can and cannot do. The page says the course is now on Claude Academy (academy.claude.com): if the old address no longer works, look for it there (checked on 24/10/2026).
New words
6 words explained
| Word | What it means |
|---|---|
| Generative AI | An AI that produces text, images, or sounds by predicting what is most likely to fit |
| Model | The AI’s “brain”, trained on a lot of text. Claude is the name of Anthropic’s models |
| Chatbot | The conversation window that gives access to the model |
| Request (or “prompt”) | What you write to the AI |
| Context | What the AI knows about your situation: roughly, what has been written in the conversation |
| Hallucination | Information made up by the AI, presented with confidence |
Check your understanding
3 questions, no grade. An explanation appears after each answer.
What's next?
Lesson 2 · What AI does badly: you will learn to spot the most common mistakes, and how to avoid them.
Module 1:
✏️ Do the exercisesMy space: your progress, your exercises and your last lesson.
Still to verify: the exact steps and labels for creating an account, on Claude and on ChatGPT.
Follow this lesson with Claude Code
With the course companion installed, open Claude Code in your working folder and type:
/formation-ia:suivre m01-l01 enIt reads the up-to-date lesson on this site and guides you step by step, without changing anything without your consent. Install the companion. Claude Code is not included in the free plan: an eligible Claude subscription, or a Console account billed by usage for the terminal.
Sources and checks
5 sources
Date, version, sources Written on 24/10/2026. Checks:
- free plans of Claude and ChatGPT, web search and memory on Claude: claude.com/pricing and chatgpt.com/pricing (pages shown in France), on 24/10/2026;
- chat available on the free plan: claude.com/download, on 24/10/2026;
- additional courses: anthropic.skilljar.com and openclassrooms.com, on 24/10/2026.
Sources checked: