Shaping Ethical AI (English version)

For once, I won’t be talking about artificial intelligence. No robots, no algorithms, no ChatGPT. Just us. This silent phenomenon running through our hospitals, our schools, our workshops, our labs. This week, I look at the gradual fading away of a discreet but essential class: the ants. The ones who keep the world standing while others dance on the ruins. La Fontaine wrote The Grasshopper and the Ant. Today, he might write The Ants Driven Out by the Grasshoppers.

The Grasshopper and the Ant, the final season.

Forget La Fontaine. His fable no longer holds.

In today’s world, it is no longer the grasshopper who finds herself destitute when the north wind blows. No, she demands a right to heating, a “climate” allowance and emotional recognition for her hard winters.

And the ant? She no longer has time to reply. She’s grafting. Well… she was.

Our age has swapped the roles: the grasshopper has become a full-time influencer, an activist every waking moment, an expert in everything except effort.

The ant, meanwhile, is fading away. Worn out, scorned, invisible. She doesn’t protest, she doesn’t shout. She disappears. Calmly. For good.

And while the grasshoppers scroll, post and rage at the speed of 5G, the pillars of the real world are crumbling without a sound.

One day, when the lights no longer come on, when there are no more teachers, no more doctors, no more tradespeople to fix the leaks, a panicked voice will ask:

“Where have all the ants gone?”

Spoiler: they’re not coming back. They’ve understood. And they’ve taken their leave.

Find the full transcript of this episode, along with all the sources and plenty more, at: The ants are fading

What if one day your car made a decision for you… and got it wrong?

A mock trial tried to answer a question that is no longer really hypothetical: can an artificial intelligence be judged the way a human would be?

Behind this courtroom staging, the whole complexity of our digital future comes to light: who is to blame when a machine causes a tragedy, but nobody really understands why?

Did you think the famous “red button: emergency stop” would always save you?

Listen to this instead.

Find the full transcript of this episode, along with all the sources and plenty more, at: The trial that sought to judge an artificial intelligence

What if you could whisper in an AI’s ear without anyone noticing? Researchers have done it. Not in a novel, but on arXiv, the most serious of scientific platforms. By typing invisible messages into their papers, they quietly steered the judgement… not of human readers, but of the artificial intelligences tasked with assessing the publications.

White text on a white background. Tiny font. Hidden instructions.

The reader sees nothing. The AI, for its part, obeys.

This is not a mere technical hack. It is a sign of the times.

Because in a world where AIs help us read, choose and decide, what happens when they are manipulated without our knowledge?

And more dizzying still: what is left of our free will if even what we read has already been formatted… for the machine that filters our view?

This podcast explores a new form of manipulation. Subtle. Insidious. Invisible. And yet formidably effective.

Find the full transcript of this episode, along with all the sources and plenty more, at: The discrete art of manipulating AI with invisible prompts

Three centuries ago, a weaver escaped in a sack of wool. Today, we fling open the doors of our lives to those who want to weave in our place.

In 1733, the flying shuttle upended the working world. In 2025, it is artificial intelligence that is transforming our gestures, our trades, our identities. But the real rupture may not be the one we think. It is no longer just our skills that we delegate; it is the taste for thinking.

And this time, there may well be no turning back.

In 1733, they burnt the machines. In 2025, we applaud them. But in both cases, it is people who are sacrificed.

Worth a listen if you, too, sense that something essential is melting away.

Find the full transcript of this episode, along with all the sources and plenty more, at: From the weaver’s shuttle to artificial intelligence

260 McDonald’s nuggets in a single order. An Air Canada chatbot lying to a bereaved customer. A recruitment algorithm blacklisting everyone over 40.

Welcome to 2024, the year artificial intelligence showed its true face. And spoiler alert: it isn’t pretty.

While everyone was going into raptures over ChatGPT, companies were brutally discovering an uncomfortable truth: when your machines go haywire, YOU pay.

Gone are the blessed days when you could shrug and mutter “it’s a computer glitch”. The courts have ruled: your algorithms, your responsibility. Full stop.

Europe legislates with the AI Act (180 pages of bureaucratic bliss). The United States innovates at breakneck speed. China controls everything. And meanwhile, our companies are discovering that developing responsible AI is like flying a fighter jet blindfolded through a storm.

The greatest irony? This silent revolution will not only decide who picks up the pieces tomorrow. It will decide who dominates the global economy for the next 50 years.

So, ready to find out why your next nightmare might go by the sweet name of algorithm?

Find the full transcript of this episode, along with all the sources and plenty more, at: Corporate responsibility in the face of potential abuses of artificial intelligence

They dominated video games. Posted superhuman scores. And then, one day, the games were made easier. The result? They collapsed.

Why? Because they had never understood what they were doing.

And that is where everything changes.

In a study as brilliant as it is unsettling, Quentin Delfosse and his team lift the veil on an illusion of intelligence: that of systems that excel… as long as nothing changes.

They devised HackAtari, a clever test based on simplified versions of Atari games. A test that, instead of making tasks harder, makes them simpler, and yet it reveals a gaping flaw. Because when you remove the obstacles, the AIs stumble. Where a human adapts and understands, the machine collapses.

What does HackAtari tell us? That an AI can shine in the exam without ever grasping its meaning. That it can repeat, optimise, correlate… without ever reasoning.

And what if our AIs were, in reality, very good pupils… who recite without understanding?

This is not a test of performance, but a test of truth. A test that measures not what an AI does, but what it understands. And which raises, implicitly, this unsettling question: do our AIs really understand what they are doing?

Find the full transcript of this episode, along with all the sources and plenty more, at: HackAtari: the test that reveals our AIs understand nothing

While six-year-old Chinese children are learning to train AI models to recognise the insects in their gardens, French children of the same age are discovering… how to open a word processor.

This gap is not a detail. It is the symptom of a strategic chasm opening up before our eyes.

On one side, China is rolling out a plan of breathtaking ambition: 12 years of progressive AI learning to turn every citizen into a “digital native”. The result? It already produces 50% of the world’s top AI researchers, against 18% for the United States.

On the other, France has just decided “once and for all” after… a 4-month consultation that gathered 500 contributions. Out of 1.2 million people in the French education system. That is 0.04% of the education community.

The French verdict? AI will be allowed only from Year 9, with compulsory training lasting 30 minutes to 1.5 hours at most to master the “basics of prompting”. In between two reminders about the water consumption of servers.

While Beijing trains entire cohorts of children who will grow up with AI as a natural companion, Paris organises consultations and offers micro-modules lasting an hour and a half.

In 10 years, guess who will truly master this technology that is already redefining the global balance of power? History may judge us on our ability to turn a technological revolution… into an administrative reform.

Find the full transcript of this episode, along with all the sources and plenty more, at: When the Middle Kingdom reinvents education with AI as a tool for civilizational conquest…

“This AI writes better than I do!” I hear that sentence at least three times a week. From a marketing director dazzled by ChatGPT. From a graphic designer captivated by Midjourney. From a student who has just discovered that a machine can solve his maths exercises in seconds.

And every time, I tell myself: we have just crossed an invisible line.

Not the line of technical performance; that is just computing doing what it has always done: calculating quickly and well. No, we have just crossed the line of our own devaluation. The one where we begin to doubt our most human abilities: thinking, creating, deciding.

As a mathematician who works with AI every day, I see three great mythological narratives being built before our eyes. Three seductive stories that gradually make us give up something precious: our intellectual autonomy.

The problem is not that AI is too capable. It is that we are becoming too credulous.

In what follows, I suggest we unpick together these three myths that are silently redrawing the borders of our humanity. Because before knowing what AI can do, it is high time to remember what we, for our part, do not want to lose.

Ready for a little exercise in collective clear-sightedness?

Find the full transcript of this episode, along with all the sources and plenty more, at: AI myths and their power of dispossession

What if the real revolution were not that of machines, but that of how we see work? Every age has its fears. Yesterday, the power looms. Today, artificial intelligence. With every advance, the same scenario: we dread disappearance, we forget transformation.

And yet…

History whispers something else. It tells us how, behind every apparent disappearance, a metamorphosis lay hidden. How trades transformed themselves. How humans moved their talent to where the machine could not go. I invite you on a journey through time and imagination. A tale for two voices, between Ada (the AI) and Cerise (aged 17), to rethink what it means to work, to adapt, to imagine… in a changing world.

If you believe AI kills jobs, listen to this. And if you don’t… listen to it too.

Find the full transcript of this episode, along with all the sources and plenty more, at: AI doesn’t destroy jobs, like every technological revolution, it creates new ones

Artificial intelligence is not what you think. Not a thought, just a prediction. Not a mind, but a statistical echo. And what if the real danger were not AI itself, but what we stop doing because it exists?

AI does not steal our intelligence. It relieves us of using it. And in that relief, a slow erosion begins… of our ability to doubt, to search, to truly think.

This piece is not an indictment of technology. It is a plea for thought. An invitation to clear-sightedness. And a warning about what we risk losing without even noticing: our inner freedom.

Listen, share, discuss. This is not a text about AI. It is a text about you.

Find the full transcript of this episode, along with all the sources and plenty more, at: AI as a thinking aid, or the slow fading of knowledge?