Teacher's Guide — Session 7 🟢

“AI in real life”

Program : Applied AI — Beginner Level (from 12 years old, general public) Instructor: Yann Isola Recommended duration: 2 hours (adaptable 1h30 – 2h30) Prerequisites: Sessions 1 to 6 (concepts: AI, machine learning, data, generative AI, critical thinking)

1. Educational objectives

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

Quote at least 5 sectors where AI is used today, with a concrete example for each.
Explain how AI helps doctors (detection on x-rays, drug discovery) without replacing them.
Describe the autonomy levels of cars (from 1 to 5) and locate where we really are. ⚠
Give examples of AI in education, agriculture, entertainment, science, finance and the environment.
Formulate the golden rule of the session: AI assists humans, it does not replace them — and cite 3 situations where AI fails.
Imagine and present a useful AI application in a sector of your choice (2-minute mini pitch).

⚠ Volatile content : business examples, the capabilities of autonomous cars, drone delivery programs and scientific records are evolving quickly. Check the news before each session (15 min standby). The sections marked ⚠ are those to be re-checked as a priority.

2. Overview and common thread

Common theme of the session: “For 6 sessions, we have been learning what AI is. Today, we're going on a journey: we're going to tour the world of professions and discover where AI is already working - in hospitals, in the fields, in space... and in your pocket. »

This is THE “wow” session of the program: concrete, full of real examples, without new theory. The highlight is the mini final pitch : each participant becomes an inventor and presents THEIR own idea for an AI application. It is also the session that prepares the rest of the program: participants see that AI affects all jobs, including their own (or the one they will do later).

Teaching posture: you are a travel guide, not a lecturer. Each sector = a “stopover” of 8-10 minutes maximum. Fast pace, lots of interactions, memorable anecdotes.

Timed course (based on 2 hours)

Time	Sequence	Format
0:00 – 0:08	Tagline: “AI has already touched you today”	Discussion
0:08 – 0:20	🏥 Health: diagnosis, medications, monitoring	Presentation + interactive map
0:20 – 0:32	🚗 Transport: levels 1-5, traffic, drones ⚠	Presentation + express quiz
0:32 – 0:42	🎓 Education & 🌾 Agriculture	Presentation + discussion
0:42 – 0:52	🎬 Entertainment & 🔬 Science	Presentation + map demo
0:52 – 1:02	Break	—
1:02 – 1:12	💰 Finance & 🌍 Environment	Exposed
1:12 – 1:25	⚖ Limits: AI assists, does not replace	Presentation + “impact meter” (web page)
1:25 – 1:50	Activity: “Invent your AI” + 2 min pitches	Workshop
1:50 – 2:00	Summary + quiz (or quiz at home)	Quiz

Adaptation 1h30: reduce each stopover to 6-7 mins, limit pitches to 1 min each, quiz at home. Adaptation 2h30: add exercise 1 (sector survey) at the start of the workshop and extend the time for preparing pitches.

3. Detailed content, sequence by sequence

3.1 Hook (8 min) — “AI has already touched you today”

Start the challenge: “Raise your hand if an AI has ever done you a favor TODAY, before you even got here. »

Few hands are raised? Perfect. Scroll down the list and watch the hands go up:

You looked at the weather report This morning ? → AI (modern forecasts use AI).
You took a route with Google Maps or Waze ? → AI (traffic prediction).
Your mailbox has filtered spam ? → AI.
A video or music has been sent to you recommended ? → AI.
Did you pay by card? → an AI verified in a few milliseconds that it was not a fraud .

Key message: AI is not “the future”. It is already everywhere, discreet, often invisible. Today, we are opening the hood, sector by sector.

Announce the program as a world tour : 8 stops, then everyone becomes an inventor.

3.2 🏥 Stopover 1 — Health (12 min)

This is the most emotionally rewarding sector: start strong.

a) Assisted diagnosis: eyes that never tire

Teaser story: a radiologist looks at hundreds of images per day. On the 300th radio, the eyes get tired. An AI analyzes the 300th image with the same attention as the first.

How it works (simple version): we showed the AI hundreds of thousands of radios already analyzed by doctors: “this one has a tumor, that one doesn’t”. The AI learned to spot patterns – exactly as it learned to recognize cats in session 2. Except that here, they are breast cancers on mammograms, melanomas on skin photos, lesions on scanners.

Significant figure: on certain specific tasks (detection of cancers on x-rays), the best AIs do as well – sometimes better – than experienced radiologists. ⚠ (studies evolve, check the most recent)

THE nuance to hammer home: AI signals suspicious areas; this is the doctor who decides . AI is a second set of eyes, a tireless assistant. No one wants a diagnosis announced by a robot — and legally, the medical decision remains human.

b) Drug discovery

Simple explanation: finding a new drug means searching a key that fits in a lock (a molecule that acts on a protein in the body). Problem: there are billions of billions of possible keys. Testing each in the laboratory would take centuries.

AI simulated millions of molecules on computer and proposes the most promising candidates. Researchers only really test the top of the list. Result: years of research saved.

Star example: AlphaFold (DeepMind/Google). For 50 years, predicting the 3D shape of a protein was a global puzzle. AlphaFold almost solved it in 2020-2021 and predicted the shape of more than 200 million proteins — a gift to all of science. Its creators received the 2024 Nobel Prize in Chemistry. ⚠

c) Patient monitoring

Smartwatches which detect heart rhythm abnormalities and alert.
AI that monitors patient constants in the hospital and warns the team before degradation.
Medication reminders, fall detection in the elderly.

Question to the group: “Would you like an AI to read your radios? And she decides on the treatment on her own? » → Almost everyone says yes to the first, no to the second. This is exactly the right intuition: assist ≠ replace .

3.3 🚗 Stopover 2 — Transport (12 min)

a) Autonomous cars: the 5 levels ⚠

Write the scale on the board — it’s a quiz staple:

Level	Simple name	Who drives?	Example
1	Assistance	Humans, helped with ONE thing	Adaptive cruise control
2	Combined support	The human, helped with speed AND direction, hands ready	Most current “autopilots” ⚠
3	Conditional autonomy	The car, in certain conditions; humans must be able to resume	Traffic jams on motorways, on some models ⚠
4	High autonomy	The car, in a defined area, without humans	Robotaxis in certain cities (Waymo…) ⚠
5	Total autonomy	The car, everywhere, all the time	Does not exist yet

Myth to bust: “Autonomous cars are for tomorrow”. It has been announced “in 5 years”… for more than 10 years. Level 5 is very difficult: the road is full of unforeseen situations (works, actions of an agent, crossing a ball). AI excels in the predictable, much less in the unexpected – remember this sentence, it comes up in the “limits” section.

How a self-driving car “sees”: cameras + radars + lidars (lasers that measure distances) → the AI merges all of this to build a 3D map in real time and predict what other users will do.

b) Traffic optimization

“Smart” red lights that adapt to the actual flow of cars → fewer traffic jams, less pollution.
Waze/Google Maps: millions of phones send their speed in real time; the AI deduces traffic jams and recalculates the routes. Each driver is a sensor!

c) Delivery by drones ⚠

Drones are already delivering medicines to hard-to-reach areas (the emblematic example: Zipline, which has been delivering blood and vaccines to Rwanda and Ghana since 2016). Parcel delivery tests exist in several countries. Honest Status: still limited — air regulations, safety, noise, weather. Check the news before the session.

Express quiz (1 min): “A car that maintains its own speed AND trajectory on the highway, but where you have to stay attentive: what level? » → Level 2.

3.4 🎓 Stopover 3 — Education (5 min)

Three uses to present:

Personalized tutoring: an AI available 24 hours a day which explains as many times as necessary, without ever getting annoyed, and adapts its explanations to the level of the student (examples: Khanmigo from Khan Academy, Duolingo ⚠). The luxury of a private teacher, for everyone.
Automatic correction: MCQ of course, but also helps with writing corrections (spelling, structure). The teacher saves time on the repetitive → more time for the human.
Difficulty detection: by analyzing a student's responses, the AI can spot EARLY that he or she is stuck on a specific concept (e.g. fractions) and alert the teacher before the delay accumulates.

Lightning talk (2 min): “Could an AI replace your favorite teacher? » → Bring to light: the AI explains, but it does not motivate, does not reassure, does not know your story, does not believe in you. Teaching is about relationships, not just information.

3.5 🌾 Stopover 4 — Agriculture (5 min)

The sector that surprises the most — AI in rubber boots:

Surveillance drones: they fly over the fields and photograph everything. The AI analyzes the images: here there is water missing, there an area is abnormally yellowing. The farmer knows exactly where to act instead of treating the entire field.
Harvest prediction: weather + soil condition + satellite images → AI estimates the harvest in advance. Useful for planning, selling, anticipating shortages.
Plant disease detection: mobile apps (e.g. PlantVillage ⚠) where the farmer photographs a diseased leaf and the AI identifies the disease — even in villages without an agronomist within miles.

Key message: it’s “precision agriculture” — less water, less pesticides, more yield. AI serving the planet AND the plate.

3.6 🎬 Stopover 5 — Entertainment (5 min)

The sector they know best — reverse: get them to tell the story.

Recommendations: Netflix, YouTube, Spotify, TikTok. AI observes what you watch, like, spend — and predicts what you'll like. Reminder of session 6: this is also how people are born filter bubbles .
Video games: the AI animates the non-player characters, adapts the difficulty, and even helps to generate worlds (landscapes, textures, dialogues). ⚠
Special effects in cinema: rejuvenation of actors, generated crowds, dubbing where the lips follow the new language. Opportunity to recall the question of deepfakes (session 5) and the debate on digital actors.

3.7 🔬 Stopover 6 — Science (5 min)

AI as a super-assistant for researchers:

New molecules and materials: beyond medicines (see health), AI offers materials for better batteries and more efficient solar panels.
Climate : AI digests mountains of data (satellites, oceans, atmosphere) to refine climate models and predict extreme events.
Space : automatic sorting of millions of images from telescopes (AI has already spotted exoplanets in data that humans had already looked at!), Martian rovers which choose their route – essential when a message takes many minutes to travel between Earth and Mars.

Anecdote that works well: citizen science + AI projects team up — volunteers classify galaxies, AI learns from them, then classifies the remaining millions. Human + AI = winning team, again and again.

3.8 💰 Stopover 7 — Finance (5 min)

Fraud detection: your bank knows your habits. An unusual payment (amount, country, time) → the AI identifies it by a few milliseconds and can block or request confirmation. It is one of the oldest and most effective applications of AI.
Algorithmic trading: a large part of stock market orders are placed by programs, in fractions of a second. Honestly mention the downside: “flash crashes” have occurred when algorithms run amok on-chain. Speed ≠ wisdom.
Banking chatbots: common questions 24 hours a day (“what is my balance?”, “how do I object?”). Useful for the simple, frustrating for the complicated – hence the famous “talk to an advisor” button. Once again: AI manages the repetitive, humans manage the delicate.

3.9 🌍 Stopover 8 — The environment (5 min)

Ending the world tour on a note of hope:

Deforestation monitoring: satellites continuously photograph forests; the AI compares the images and near real-time alert when trees disappear (e.g. Global Forest Watch ⚠). Before, we discovered deforestation months later.
Improved weather forecast: Recent AI models (GraphCast from DeepMind, among others ⚠) predict the 10-day weather faster – and often better – than classic supercomputers. Better anticipation of cyclones = lives saved.
Energy optimization: AI adjusts the heating/cooling of buildings and data centers (Google has reduced the cooling energy of its data centers by around 40% thanks to AI ⚠), balances electrical networks, better integrates solar and wind.

Honest nuance to give: the AI itself consumes a lot of energy (drive of large models, data centers). It is both part of the solution AND part of the problem. Good critical thinking reflex inherited from session 6.

3.10 ⚖ Limits — AI assists, does not replace (13 min)

The most important sequence. After the exciting world tour, we rebalance.

The golden rule (to be repeated to the group): “AI does not replace humans – it assists them. »

In ALL sectors visited, the pattern is the same:

Health → AI signals, the doctor decides
Transport → AI assists, the driver remains responsible (levels 1-3)
Education → AI explains, the teacher motivates and supports
Finance → AI alerts, the advisor manages delicate cases

Where the AI fails (the 3 red zones):

Cultural context and common sense. AI learns from data; what is rare or local escapes him. A joke, an innuendo, a regional custom, irony — it misses the point. A telling example: a recruitment AI trained on past CVs reproduced past discrimination (Amazon case, abandoned in 2018).
True emotions. An AI can simulate empathy (“I understand that it is difficult”) but feels Nothing. Breaking bad news to a patient, consoling a student, negotiating a conflict: human, human, human.
Pure creativity and the unexpected. AI excels at recombining what exists. Inventing an entirely new artistic genre, reacting to a never-before-seen situation (the famous road under construction with an agent making gestures): this is where it stalls. She interpolates, humans invent.

Summary formula to remember: “AI is unbeatable in the repetitive, the massive and the fast. The human remains unbeatable on context, emotion and the unprecedented. The best teams combine the two. »

Flash activity — the impact meter (web page): for a series of tasks (“spot a tumor on 10,000 x-rays”, “consolate a patient”, “predict the weather”, “write a sincere poem about your grandmother”…), the participants place the cursor between “AI excels” and “humans are irreplaceable”. Discuss borderline cases — that’s where the learning happens.

3.11 🛠 Final activity — “Invent your AI” + mini pitches (25 min)

See exercise 2 of the exercise file for the complete procedure. In summary:

Choose a sector (among the 8, or another: sport, justice, cooking, fashion, etc.)
Identify a real problem in this sector
Imagine an AI application who helps (without replacing humans!)
Prepare a 2 minute pitch with the canvas: Problem → AI solution → Necessary data → What the human keeps → Name of the app
Pitcher in front of the group

The “AI design” of the web page can serve as a warm-up: it manipulates the logic problem → data → method → result in a fun way before the paper exercise.

Role of the instructor during pitches: time strictly (2 min!), applaud each passage, and ask ONE question per pitch, always the same: “And what does the human keep in your system?” » If the student has an answer, the lesson of the session is acquired.

4. Frequently asked questions from participants (and ready answers)

“Will AI eliminate jobs? » Honest answer: it will transform many, eliminate some (very repetitive tasks) and create others (like every great technology: the computer has eliminated typist jobs and created millions of IT jobs). Jobs combining expertise + human relations are the strongest. The best part: knowing how to work WITH AI — that’s exactly what we learn here.

“Why aren’t self-driving cars everywhere yet? » Because the 1% of unforeseen situations are extremely difficult, liability in the event of an accident is a thorny legal issue, and public trust is earned slowly. Driving is 99% easy, but it's the last 1% that counts.

“Can AI make mistakes in medicine? » Yes — that’s exactly why the doctor keeps the decision. The AI can make mistakes on rare cases or populations that are poorly represented in its training data. Two looks (AI + doctor) are better than one.

“What is the most protected profession in AI? » Nice trap: there is no “protected” profession, but tasks more or less automatable. Nurse, plumber, educator, craftsman: lots of context, unforeseen events and human relationships → very difficult to automate.

5. Educational pitfalls to avoid

The catalog effect. 8 sectors in 50 minutes = risk of indigestion. The remedy: ONE notable story per sector, not an exhaustive list. Better that they remember AlphaFold and Zipline than 40 skimmed names.
The techno-smug. Only success stories = dishonest. The nuances (non-existent level 5 car, biased recruitment AI, energy consumption) make the point credible.
Reverse catastrophism. “AI will destroy everything” is no more accurate. Hold the line: great assistant, bad replacement.
Skip the pitches. If time is running out, cut the presentations, NEVER the final activity. She is the one who anchors everything.
Give outdated figures. Everything marked ⚠ is checked the day before. A false number detected by a participant ruins your credibility (and it's a great lesson from session 6 that they would apply against you!).

6. Materials and preparation

Video projector + session slides
Interactive web page (webpage/index.html ) tested — works offline
A4 sheets + markers for pitch canvases
Visible stopwatch for pitches (phone or online timer)
Printouts of the pitch canvas (exercise 2) — 1 per participant or pair
15 min of news monitoring on topics ⚠ (autonomous cars, drones, tools mentioned)
Optional: small symbolic rewards for the “public’s favorite pitch”

7. The final word (to be said at the end)

“We traveled around the world: hospitals, roads, fields, cinemas, laboratories, banks, forests. Everywhere, the same story: AI does the repetitive, massive and rapid work — and humans keep the context, the emotion and the decision. The question is no longer “will AI change your future job?” — she will. The real question is: will you be the one who knows how to work with her? After tonight, you have a head start. »

Bridge to session 8: “Next time, we’ll talk about what all this involves: ethics, privacy, and how AI should be regulated. Because such a powerful tool deserves rules. »