Français
Applied AI · Beginner 🟢 · Session 1
📝 Teacher's Guide
← Return to program 📄 Source .md

Teacher’s Guide — Session 1 🟢

“What is artificial intelligence?”

Program: Applied AI — Beginner Level (ages 12 and up; no specialist background required) Instructor: Yann Isola Duration: 2 hours (with breaks and practical activities) Prerequisites: None. Zero. Nada. That’s the goal!

🎯 Learning objectives

At the end of this session, each participant should be able to:

  1. Define AI in their own words — “a program that learns from examples instead of following rules written one by one”
  2. Name 3 key moments in the history of AI (among: Turing, ELIZA, Deep Blue, AlphaGo, ChatGPT)
  3. Distinguish the 3 types of AI: narrow (today), general (science fiction), superintelligence (debate)
  4. Explain the cycle: data → learning → prediction
  5. Spot AI in daily life (at least 5 examples)
  6. Debunk 3 common myths about AI
  7. Have their first real conversation with a chatbot, with a critical eye

💡 Philosophy of the session: we don’t train engineers today. We ignite a spark of curiosity. If a participant leaves saying “ah, actually it’s not magic, and it’s even more interesting than I thought”, it’s a win.

📋 Materials required

Plan B without internet: the entire session works offline EXCEPT the chatbot activity (part 7). In this case, use the “simulated ELIZA” demonstration: the teacher plays the chatbot at the board and follows simple rules, such as turning a student’s statement into a question. It is an excellent learning exercise in its own right.

⏱️ Minute-by-minute plan (120 min)

Time Duration Sequence
0:00 10 mins 1. Icebreaker: “What does AI mean to you?”
0:10 15 mins 2. Simple definition + the big idea
0:25 15 mins 3. A brief history of AI (interactive timeline)
0:40 10 mins 4. The 3 types of AI
0:50 10 mins ☕ PAUSE
1:00 15 mins 5. How it works: data → learning → prediction
1:15 10 mins 6. Game: “AI or not AI?” (interactive web page)
1:25 20 mins 7. Practical activity: first conversation with a chatbot
1:45 10 mins 8. Myths vs reality + AI jobs
1:55 5 mins 9. Quick quiz + conclusion

1. Icebreaker (10 min) — “What does AI mean to you?”

Format: quick round-robin OR post-its on the board.

Ask the question: “When I say ‘artificial intelligence,’ what is the FIRST image that comes to mind?”

Typical responses to expect (and note on the board, we’ll come back to that):

Facilitation tip: DO NOT correct ANYTHING at this point. Write everything down. At the end of the session, return to this board and ask, “What would we keep?” What would we cross out?” This is the best indicator of learning of the day.

Second quick question: “Who has used AI today? Raise your hand.” → Usually 2–3 hands go up. Then ask: “Who used GPS, watched a recommended video, or used autocorrect?” → all hands go up. First “wow” moment: you have all been using AI for years without realizing it.

2. Simple definition (15 min)

The big idea to get across

Write these two approaches on the board:

Traditional program:

A human writes ALL the rules. “IF the email contains “won 1 million” THEN it’s spam.” Problem: impossible to write a rule for each case.

Artificial intelligence:

We show the program THOUSANDS of examples (“this is spam, this is not, this is…”) and it finds the patterns on its own.

Definition to remember (to copy/repeat)

🎯 AI (Artificial Intelligence) is a program that learns from data (examples), instead of being programmed rule by rule.

Key analogy: learning to ride a bike 🚲

“Has anyone learned to ride a bike by reading a 500-page rule book? No. You got on it, you fell, you did it again, and your brain adjusted. AI learns the same: by example and error, not by manual.”

Vocabulary to be introduced naturally (without jargon):

Follow-up question for lively groups: “If AI learns from examples, what happens when the examples are poor?” → Plant the seed of critical thinking (we return to it in section 8).

3. Short history of AI (15 min)

Use the animated timeline of the web page (“The great adventure of AI” section) or the slides.

The 5 dates to tell like stories (not like dates!)

1950 — Alan Turing asks THE question 🇬🇧 Tell the story: Turing was an English mathematician and wartime codebreaker who helped decipher Enigma messages. In 1950, he published a paper that asked a radical question for its time: “Can machines think?” He proposed a game: if you exchange written messages without knowing whether the other party is human or machine, and you cannot tell which it is, does that count as “thinking”? This became known as the Turing test. Important anecdote: at the time, a “computer” was less powerful than a school calculator.

1966 — ELIZA, the first chatbot 💬 A researcher at MIT (Massachusetts Institute of Technology), Joseph Weizenbaum, created ELIZA: a program that imitated a psychotherapist by rephrasing users’ sentences: “I’m sad” → “Why do you say you’re sad?” It used a simple technique, yet people still confided in ELIZA for hours. Weizenbaum himself was troubled by their reactions. Lesson: humans readily attribute intelligence to something that talks.

1997 — Deep Blue beats Kasparov in chess ♟️ Garry Kasparov, then the world chess champion, lost a match to IBM’s Deep Blue computer. It was a global shock. Deep Blue was not “thinking”: it evaluated around 200 million chess positions per second ⚠. This was search and computation, not machine learning.

2016 — AlphaGo beats Lee Sedol in the game of go 🀄 Go is an ancient board game with more possible board positions than estimated atoms in the observable universe ⚠. Exhaustive search is impossible. AlphaGo, created by DeepMind, learned from human games and self-play. In the second game against Lee Sedol, it played the famous “move 37,” which commentators initially considered highly unusual before recognizing its strength. Wow moment: AI surprised its own creators.

2022 — ChatGPT: AI reaches everyone’s pocket 🚀 OpenAI launches ChatGPT (GPT = Generative Pre-trained Transformer - just remember: a program trained on huge amounts of text to predict the next words). 100 million users in 2 months ⚠ — the fastest adoption in technology history at the time. For the first time, anyone can chat with an AI.

Mini-activity (in the timeline game)

Deal out 5 cards (or use the reordering game on the web page / exercise 2): participants put the events in order. 3 minutes, in pairs.

4. The 3 types of AI (10 min)

Draw 3 boxes on the board:

Narrow AI 🎯 General AI 🧠 Superintelligence 🌌
What is it? Very strong on ONE task As versatile as a human Surpasses humans in everything
Does it exist? Yes, everywhere ❌ No — science fiction (for now) ❌ No — philosophical debate
Example Translate, recognize a face, play go A movie robot that can do everything Science-fiction stories

Analogy that works: “Today’s AI is like an Olympic swimming champion… who doesn’t know how to walk. AlphaGo crushes humans at Go, but it can’t play tic-tac-toe, add up, or say hello. Each AI = a single talent.”

Teaching note: some participants will ask, “But ChatGPT can do lots of things, right?” Good question. ChatGPT is highly capable at generating and transforming text across many domains, but that does not make it general AI. Researchers also debate where to draw the boundary.

☕ BREAK (10 min)

Leave the web page open for free exploration: curious participants can use the timeline during the break. That is intentional.

5. How it works: data → learning → prediction (15 min)

The central analogy: learning to recognize cats 🐱

Tell this story (it structures the ENTIRE rest of the course):

“How did you learn to recognize a cat as a baby? No one gave you a formal definition. You saw cats again and again: ‘Look, a cat!’ Then one day you saw a new kind of cat—ginger, tailless, and huge—and still recognized it. Your brain had found the pattern.”

AI does exactly the same, in 3 steps:

  1. 📦 DATA — We give it thousands of photos labeled: “cat” / “not cat”
  2. 🧠 LEARNING — The program looks for patterns: ear shapes, whiskers, eyes… It makes a mistake, we correct it, it adjusts. Millions of times. Very quickly.
  3. 🔮 PREDICTION — We show it a photo it has never seen → it says “97% cat.” This is a prediction, not a certainty!

Participatory demonstration (5 min) — “You are the AI”

Board game: draw (or project) simple scribbles. Announce: “I will train you.” Show 4 drawings by giving the label: “this is a blorp”, “this is not a blorp”… (for example: blorps all have points, non-blorps are round — don’t say it!). Then show a 5th drawing: “Blorp or not blorp?” → The group guesses. Ask: “How did you know?” Nobody gave you the rule!” You have just experienced machine learning from the inside.

The 3 key messages to hammer home

6. Game: “AI or not AI?” (10 min)

Launch the interactive game from the web page (webpage/index.html, “AI or not AI?” section). As a whole class or in teams: we display an everyday situation, we vote (raised hands, or “left = AI / right = not AI” while moving), then we reveal.

Situations covered (with subtleties to comment):

Rule of thumb to give: “Did it learn from examples?” Could it be wrong? If yes to both → probably AI.”

7. Practical activity: first conversation with a chatbot (20 min)

This is THE highlight of the session. Objective: to move from “wow, it’s magical” to “interesting, I see what it can do and where it falls short”.

Organization: pairs, 1 device per pair. Use the Guided Sheet for Exercise 3 (exercises/exercises.md).

The 4 missions (approximately 5 min each)

Mission 1 — Get acquainted 👋 Ask the chatbot: “Who are you? How do you work? Do you think?” → Observe: it explains that it is a program and does not think like a human.

Mission 2 — Ask it to amaze you 🌟 Ask for something creative: “Write a poem about my pirate dog”, “Explain black holes to a 6 year old”… → Objective: see the power. Guaranteed “wow” moment.

Mission 3 — Try to catch it out 🕵️ Try to make it make a mistake: ask about very recent events, tricky calculations, absurd questions (“How many ‘r’s are in ‘strawberry’?”), or invented people. → Objective: see the limits. It can invent with total aplomb — we call it a hallucination.

Mission 4 — The Reverse Turing Test 🎭 Ask: “Pretend you are human and try to convince me that you are not an AI.” → Discuss: is it convincing? Why? Would Turing be impressed?

Group debriefing (5 min — NEVER skip it)

Three questions on the board:

  1. “What impressed you the most?”
  2. “When did you see it make a mistake or invent something?”
  3. “Does the chatbot think?” → a natural transition to the myths.

Safety note (to be stated explicitly): never give personal information (full name, address, school, etc.) to a chatbot. Conversations can be saved.

8. Myths vs reality + jobs (10 min)

Use the flip cards on the web page, or run the activity aloud as a “True or false?” discussion.

Myth 1: “AI thinks like us” ❌

Reality: it calculates probabilities and identifies patterns. ChatGPT predicts the next most likely word — brilliantly — but has no awareness, desires, or understanding of the world. The parrot that says “hello” does not wish you a good day.

Myth 2: “AI will replace everyone” 🤔 (nuanced)

Reality: AI is transforming jobs more than eliminating them altogether. It replaces tasks (repetitive, predictable), rarely entire jobs. Historically, big technologies have displaced work and created new jobs (who had heard of “community manager” in 1990?). But yes, certain jobs change quickly, and it is a real social issue. The best defense: understand the tool — exactly what you’re doing today.

Myth 3: “AI is infallible” ❌

Fact: AI is often wrong — and the worst part is that it is confidently wrong. Chatbot hallucinations, facial recognition errors, biases inherited from data. Golden rule to note: an AI response must be checked; never take it at face value.

AI jobs (3 min, opening)

Message: AI is not just for math people. We need creative people, writers and humanities specialists, the curious, the skeptical.

9. Quick quiz + conclusion (5 min)

Homework (optional and fun): “By the next session, write down 5 moments when you come across an AI in your day. The most surprising wins.”

🧯 Teacher FAQ — common trick questions

“Can AI become evil like in Terminator?” → Today’s AI has neither will nor objectives of its own. The real current risks are more boring but very real: errors, bias, misinformation, malicious use by humans. The long-term risks are debated by the researchers themselves — an open question, honesty required.

“How does ChatGPT know so much?” → It was trained on enormous quantities of text (websites, books, articles). It does not “know” facts in the human sense: it has encoded statistical patterns from its training data. That is why it can invent information—it produces plausible text, not guaranteed truth.

“Does AI have feelings?” → No. It can write “I’m happy for you” because that’s what a human would write in this context. Simulation, not emotion. (ELIZA, 1966: already the same misunderstanding!)

“What is the difference between AI and robot?” → The robot is the (mechanical) body. AI is possibly the brain (software). There are robots without AI (a factory arm that repeats a fixed motion) and AI without robots (ChatGPT has no body).

“Why do we say that AI has biases?” → It learns from data produced by humans. If the data contains biases or gaps, AI reproduces them — sometimes by amplifying them. Simple example: an AI trained only on photos of ginger cats will say that a black cat is not a cat.

End of session checklist