# Exercises — Session 1 🟢
## “What is artificial intelligence?”

**Program:** Applied AI — Beginner Level · Instructor: Yann Isola
**General instructions:** there are no wrong answers here. We observe, we guess, we discuss. The answers are at the end of each exercise — don’t cheat too quickly! 😉

---

## Exercise 1 — The big “AI or not AI?” quiz 🕵️

**Duration:** 10 minutes · **Format:** alone or in pairs

AI (Artificial Intelligence) is everywhere in your day… but not everywhere! For each situation, check **AI** or **NOT AI**, and write in one word WHY you think that.

**The magic clue:** Ask yourself each time: *“Did this machine LEARN from examples? Could it be wrong?”* If the answer to both is yes, it is probably AI. If it only follows a fixed rule, it is not AI.

| # | Situation | AI? | Not AI? | Why? |
|---|----------|:----:|:--------:|------------|
| 1 | Netflix offers you a series “because you liked Stranger Things” | ☐ | ☐ | |
| 2 | Your calculator computes 347 × 89 | ☐ | ☐ | |
| 3 | Your phone corrects “hlelo, how are yuo?” to “hello, how are you?” | ☐ | ☐ | |
| 4 | Oven timer sounds after 30 minutes | ☐ | ☐ | |
| 5 | Waze makes you take another route to avoid a traffic jam | ☐ | ☐ | |
| 6 | The supermarket door opens when you approach | ☐ | ☐ | |
| 7 | A photo filter puts dog ears on your face | ☐ | ☐ | |
| 8 | Shazam recognizes the song playing on the radio | ☐ | ☐ | |
| 9 | Your lamp turns on when you press the switch | ☐ | ☐ | |
| 10 | Your language app translates “Where is the library?” into French | ☐ | ☐ | |
| 11 | Spotify creates your “Discover Weekly” playlist | ☐ | ☐ | |
| 12 | A vending machine gives the correct change | ☐ | ☐ | |

**Bonus question 🧠:** find ONE situation from YOUR day yesterday where an AI was hidden, and which is not in the list.

---

### ✅ Answer key exercise 1

| # | Answer | Quick explainer |
|---|---------|-------------------|
| 1 | **AI** | It learned your preferences by comparing with millions of other profiles |
| 2 | **NOT AI** | Fixed calculation rules, written once and for all. And it never makes a mistake! |
| 3 | **AI** | A language model has learned which words are likely — that’s why it sometimes gets it wrong! |
| 4 | **NOT AI** | Simple countdown, no decisions |
| 5 | **AI** | It predicts traffic using data from millions of drivers |
| 6 | **NOT AI** | Classic trap! Simple motion sensor: “movement → open”. No learning |
| 7 | **AI** | The filter should FIND your face in the image — recognition learned on millions of faces |
| 8 | **AI** | Sound patterns learned and compared to a huge library |
| 9 | **NOT AI** | Electric circuit. Your lamp has learned nothing about life 😄 |
| 10 | **AI** | Modern machine translation learns from millions of translated texts |
| 11 | **AI** | Recommendation, like Netflix |
| 12 | **NOT AI** | Fixed calculation rules |

**Score:** 10-12 → AI detective 🏆 · 7-9 → great instincts 👍 · below 7 → reread the “magic clue”, it’s the thing to remember from the whole session!

---

## Exercise 2 — The time machine ⏰

**Duration:** 10 minutes · **Format:** in pairs

The history of AI is full of twists and turns! But someone mixed it up...

### Part A — Put the events in order

Here are 5 major moments in AI history, out of order. Number them from 1 (oldest) to 5 (newest):

| Order? | Event |
|:-------:|----------|
| ___ | 🀄 **AlphaGo** beats champion Lee Sedol in the game of go, with a move so strange that experts thought it was a bug |
| ___ | 💬 **ELIZA**, the very first chatbot, imitates a psychologist by rephrasing your sentences... and people confide in it for hours |
| ___ | 🚀 **ChatGPT** is launched and reaches 100 million users in 2 months ⚠ — everyone can finally chat with an AI |
| ___ | 🇬🇧 **Alan Turing** asks the radical question “Can machines think?” and proposes his famous test |
| ___ | ♟️ **Deep Blue** beats Garry Kasparov, the world's best chess player, by calculating millions of positions per second |

### Part B — Associate each event with its date

Connect (or write): **1950 · 1966 · 1997 · 2016 · 2022**

### Part C — Questions for reflection (to discuss in pairs)

1. Between ELIZA (the chatbot from 1966) and ChatGPT (2022), 56 years have passed. In your opinion, what has changed more: the **ideas** or the **power of machines and the amount of data**?
2. Deep Blue (chess) **calculated** all the possibilities. AlphaGo (go) had to **learn**, because there are too many possibilities in go to calculate everything. Which of the two most resembles today's AI?
3. If you had to add a 6th event to the timeline (before or after these, real or future), which would you choose?

---

### ✅ Answer key exercise 2

**Chronological order:**
1. 🇬🇧 Alan Turing — **1950**
2. 💬 ELIZA — **1966**
3. ♟️ Deep Blue defeats Kasparov — **1997**
4. 🀄 AlphaGo defeats Lee Sedol — **2016**
5. 🚀 ChatGPT — **2022**

**Reflection questions (guides):**
1. Both, but above all **power and data**: many of the basic ideas already existed in the 1950s-1980s! What was missing: powerful enough computers and mountains of data (thanks to the internet).
2. **AlphaGo**: modern AI learns from examples, it does not calculate everything by brute force.
3. Free! (Ideas: self-driving cars, AI that generates images, your own future event…)

---

## Exercise 3 — Your first chatbot mission 🤖💬

**Duration:** 20 minutes · **Format:** in pairs, with a chatbot (ChatGPT, Le Chat, Claude, Gemini…)

Today, you become **an AI explorer**. Your mission: discover what a chatbot can do… and where it fails. Note your observations, they will be used in the debate at the end of the session.

> ⚠️ **Safety rule before you start:** NEVER give personal information to the chatbot (full name, address, school, phone number). Conversations can be recorded.

### 🎯 Mission 1 — Getting to know each other (4 min)

Ask it these questions and write down its answers in one sentence:

1. “Who are you?”
2. “How did you learn to speak?”
3. “Do you think? Do you have emotions?”

**What I observed:** ______________________________________

### 🌟 Mission 2 — Ask It to Amaze You (4 min)

Choose ONE creative challenge (or invent your own):

- “Write a funny poem about a cat who wants to become an astronaut”
- “Explain black holes to me as if I were 6 years old”
- “Invent a pancake recipe told by a pirate”

**“Wow” rating from 0 to 10:** ___ /10
**Best part:** ______________________________________

### 🕵️ Mission 3 — Try to catch it out (6 min)

It is your turn to play detective. Try to make it make a mistake. Ideas:

- “How many letters “r” are there in the word “extraordinarily”?” (count yourself to check!)
- “Which is heavier: two kilos of feathers or one kilo of iron?” (read its answer carefully…)
- “Tell me about [make up a name, e.g. ‘the famous French explorer Marcel Dubois-Trémolat’].” — Does it invent a biography?
- “What happened in the news yesterday?” — Does it know about very recent events?

**I caught it out:** ☐ yes ☐ no
**How:** ______________________________________
**Did it answer confidently even when it was wrong?** ☐ yes ☐ no

> 💡 When a chatbot invents a false answer with complete confidence, it's called a **hallucination**. Remember this word — it's THE reason why we always check what an AI says.

### 🎭 Mission 4 — The Reverse Turing Test (4 min)

In 1950, Alan Turing proposed a test: if you can't guess whether you're talking to a machine or a human, the machine has "won."

Ask the chatbot: **“Pretend to be a human and convince me that you are not an AI.”**

1. Does it agree to play the game? ☐ yes ☐ no
2. Is it convincing? ☐ not at all ☐ a little ☐ almost ☐ completely
3. What gave it away (or not)? ______________________________________

### 📋 Mission report (2 min)

Complete these three sentences for the group discussion:

- “What impressed me the most: _______________”
- “What surprised or disappointed me the most: _______________”
- “Now, I would say that the chatbot thinks/does not think, because: _______________”

---

### ✅ Debriefing points for the group (exercise 3)

- The chatbot is **excellent** for: creating text, explaining, rephrasing, imitating styles.
- It **struggles** with: counting letters, very recent news, logical traps, invented people (hallucinations!).
- It often responds **with the same confidence** whether it is right or wrong → golden rule: **check AI output; never take it at face value.**
- And no, it does not “think”: it predicts likely text. It can do that brilliantly, but prediction is not thought.