Exercises — Session 7 🟢
“AI in real life”
Program : Applied AI — Beginner Level · Instructor: Yann Isola Material : paper + markers. Internet access useful for exercise 1 (“screenless” version provided).
Exercise 1 — The sector investigation 🔎
Duration: 20 min · In small groups (2-3) · Feedback: 1 min per group
The idea
You become specialized journalists . Each group receives (or chooses) a sector and conducts the investigation: how is AI used there TODAY, concretely? No science fiction – only reality.
The sectors to be distributed
🏥 Health · 🚗 Transport · 🎓 Education · 🌾 Agriculture · 🎬 Entertainment · 🔬 Science · 💰 Finance · 🌍 Environment
Unrolled
Step 1 (12 min) — The investigation. Complete the survey form below. With the internet: look for real examples (cross at least 2 sources — reflex from session 6!). Without internet: use the interactive map on the session web page + your memories of the course.
Step 2 (3 min) — The synthesis. Choose the most surprising example from your investigation: the one you would tell your family this evening.
Step 3 (1 min per group) — The news flash. Each group presents its sector in 1 minute, TV news style: “In the healthcare sector, AI now knows…”
Survey form (to be completed)
| Question | Your answer |
|---|---|
| Our sector: | |
| 3 real uses of AI in this sector: | 1. / 2. / 3. |
| The most surprising example: | |
| Who the AI help in this sector? (job, person) | |
| What does the AI do, and what does the human keep? | AI: / Human: |
| A limit or risk in this sector: |
What to remember
In each sector, the same pattern recurs: the AI processes the massive and the repetitive, the human retains the judgment and the relationship . If your investigation doesn't reveal this duo, dig some more!
Exercise 2 — “Invent your AI”: the pitch workshop 🚀
Duration: 25 min (10 min preparation + 2 min pitches) · Alone or in pairs
The idea
The inventors of the applications we saw today all started the same: a real problem + an idea. Your turn. You will design a useful AI application and present it in 2 minutes flat — like a start-up creator in front of investors.
Golden rule
Your AI must to assist a human, not replace him. A pitch where AI does everything and humans disappear will be gently but firmly challenged. 😉
Unrolled
Step 1 (2 min) — Choose a playing field. One of the 8 sectors of the course, or another: sport, cooking, music, games, animals, justice, fashion, travel…
Step 2 (8 min) — Complete the pitch canvas.
┌─────────────────────────────────────────────────┐
│ CANEVAS DE PITCH — MON IA │
├─────────────────────────────────────────────────┤
│ 1. LE PROBLÈME │
│ Quel problème réel et concret ? │
│ Qui en souffre aujourd'hui ? │
│ │
│ 2. MA SOLUTION IA │
│ Que fait mon application, en une phrase ? │
│ │
│ 3. LES DONNÉES │
│ De quoi mon IA a-t-elle besoin pour │
│ apprendre ? (images ? sons ? historiques ?) │
│ Où les trouver ? Y a-t-il un souci de vie │
│ privée ? │
│ │
│ 4. CE QUE L'HUMAIN GARDE │
│ Quelle décision reste humaine ? │
│ Qui vérifie l'IA ? │
│ │
│ 5. LE NOM + LE SLOGAN │
│ Un nom qui claque + une phrase choc │
└─────────────────────────────────────────────────┘
Step 3 (2 min per pitch) — Present. Recommended structure: 20 s of problem → 40 s of solution → 30 s of data → 20 s on the role of humans → 10 s of name + slogan. The instructor keeps a timer. At 2 minutes sharp: applause, even in full sentence!
Step 4 — Voting. The public votes for their favorite pitch (not the most technical — the most useful and clear ).
Examples to unlock inspiration breakdowns
- 🐕 “WoofCare”: the app that analyzes your dog’s gait on video and alerts if lameness appears. (The veterinarian diagnoses.)
- ⚽ “CoachEye”: the AI that analyzes training videos and suggests personalized exercises. (The coach decides the program.)
- 🍎 “FrigoZéro”: photo of the inside of the fridge → anti-waste recipes with what will soon expire. (You cook… or not!)
Self-assessment grid before pitching
- My problem is real and I can tell WHO is suffering from it
- My solution is contained in a sentence understandable by my grandmother
- I know what data my AI needs to learn
- I thought about the privacy of this data
- A human keeps a real decision in my system
- My pitch takes 2 minutes (I rehearsed it once!)
Exercise 3 — “Is it AI or not?” »: real cases 🕵
Duration: 15 min · In whole group or small groups · Commented correction together
The idea
AI is everywhere… but not EVERYTHING is AI! For each real-world situation below, decide: AI (the system learns from data) or NO AI (fixed rule, classic program, or simply a human). Then justify in one sentence.
The 10 cases
| # | Situation | AI or not? | For what ? |
|---|---|---|---|
| 1 | Netflix offers you a series “because you watched…” | ||
| 2 | Your calculator displays 7 × 8 = 56 | ||
| 3 | Your bank blocks your card after a bizarre payment at 3 a.m. in a faraway country | ||
| 4 | The pedestrian crossing turns green every 90 seconds, right on time | ||
| 5 | An app identifies a plant’s disease from a photo of a leaf | ||
| 6 | The oven stops when the timer reaches zero | ||
| 7 | Subtitles automatically appear on a YouTube video | ||
| 8 | A vending machine gives exact change | ||
| 9 | Waze announces “traffic jam in 2 km” and suggests a detour | ||
| 10 | Your phone unlocks by recognizing your face |
Answer key (for the instructor — comment out loud)
- AI — recommendation learned from millions of viewing histories.
- No AI — fixed calculation rule, no learning. The result is always the same.
- AI — fraud detection: the system has learned YOUR habits and identifies the anomaly.
- No AI — fixed timer. (Bonus nuance: some modern “adaptive” lights use AI — the question to ask is: does the system adapt by learning?)
- AI — computer vision trained on thousands of photos of diseased leaves.
- No AI — simple countdown.
- AI — voice recognition (speech → text), learned on huge audio corpora.
- No AI — calculation and mechanics. Returning €3.40 requires no learning.
- AI — traffic prediction based on data sent in real time by millions of phones.
- AI — facial recognition: the phone has learned YOUR facial features.
The magic question to remember
To decide, always ask yourself: “Did this system LEARN from data, or does it apply a rule written once and for all? » Learning → AI. Fixed rule → classic program. This is THE border (rewatch session 2 if necessary!).
To go further (at home)
For 24 hours, note 3 moments where you come across AI in your day without having thought about it before this course. Bring your list to the next session — the most unexpected finds earn the group's undying respect. 🏆