Through sleepless nights, relentless reading and writing, the roots of my discomfort have become clearer. I share here, through this text, the genesis and development of my questioning about technology and Artificial Intelligence. These reflections lead to a reading grid of AI-related risks that diverges from the usual narrative.
This personal view - which nonetheless converges with certain other voices recently making themselves heard - I hope it will help reinforce our understanding of what's at stake in the use of these technologies, in order to work toward limiting their negative externalities for users.
So, let's fasten our seatbelts and dive again in search of the white rabbit.
I've loved technology for nearly thirty years. I use it, study it, even spread it. It's my profession, my field of expertise. To tell the truth, if you start a conversation with me about technology, you'll probably have to stop me…
It has always been for me synonymous with emancipation, with reflection and construction. A vector of freedom, at least intellectual freedom. It can also be playful: how many weekends have I spent coding "stuff," recompiling a Linux kernel for the 73rd time, learning the n-th fashionable language.
I think I have hundreds, if not thousands of "things" in my closet of stuff. For most of them, they are half thing, half whatsit. Few are elevated to the rank of "object." What is certain is that they are above all perfectly useless, often defective and incomplete. Yet I conceived them all with passion.
It's this kind of tech enthusiast who, almost three years ago to the day, took the announcement of ChatGPT's release head-on. That tipping point in technology, that date so singular whose mere mention triggers indifferently a "already" or an "only." That strange universal impression that space-time had bent; from then on everything would be infinite acceleration.
During these last thirty-six months, I've gone through several phases. After the stupefaction of the first days, the first weeks, I was among the first and most enthusiastic, a real fanboy; like a Swifty, but for AI. In my defense, I also almost immediately understood the major transformative importance this technology would carry. My job, combined with my (delirious) enthusiasm, allowed me to travel to the places where AI is being birthed: Silicon Valley. It's here, in the holy of holies of this technological Mecca, that the pot bubbles; here from which new models emerge at an inhuman pace. In a way, a 2.0 version of the "New World" was being forged and I witnessed it.
These innovations all carried something elusive within them. This technology I knew by heart, which was so predictable and rational to me, was here showing itself in a new light, holding even a kind of magic.
With AI, tech seemed to break free of its chains.
I sensed very clearly a major advance, obviously technologically, but perhaps even more so on the field of its societal implications. Would Rifkin finally be right in predicting the "end of work" in his eponymous 1995 book? I told myself we were dealing with something even more important than the arrival of the Internet; perhaps of the order of the advent of electricity.
I wasn't the only one to feel it, because very quickly, an army of tech enthusiasts (some long-standing, others more opportunistic) rallied under this same AI banner. A kind of planetary ALL-IN. As in the cult film "Interstellar" with that wave that kept growing, AI was going to give birth to swarms of new services; a true new economy. All of it propelled to levels and a speed never before observed in the history of human progress.
In this race, I was in the lead pack of the enlightened, overflowing with enthusiasm. I insisted on putting AI in everything; the nature of the task mattered little. Everything was open to experimentation. These experiences came with a constant justification and a self-persuasion about the absolute necessity of putting it everywhere. And, I don't deny it, this "without moderation" consumption of AI, so intoxicating, brought me real pleasure.
But what genius had designed this cocktail that was at once holy water by its virtues and fentanyl by its addictive power?
I first wore out and exasperated my family, then the few friends I had left and, finally, of course, my colleagues. But thank God, most could be converted to this new drug without too much difficulty; swelling the ranks of the apostles who would continue spreading this good news around the world. A kind of religion, a cult. Alongside this fit of madness, which had literally taken control, my TV appearances, conferences, and my "little" notoriety also benefited from this technological swell; all forming a feedback loop between my ego and my delirium. Taking photos with demi-gods during my travels in the Olympus of AI. I reveled in the nectar of being recognized in this small circle, of "being part of the club"; this club whose members also included accelerationists, techno-solutionists, transhumanists and other even more obscure cliques.
Yes, I was completely sucked into this vortex, carried away by this whirlwind that seemed to accelerate endlessly.
My personal tipping point was the realization that AI wasn't just going to eliminate a few jobs, but redraw the contours of entire industries, going so far as to raze some, perhaps many of them. It's not the little genius Schumpeter, you know the one who is systematically invoked in these cases, who reassured me. Maybe it's finally time to invent a "Schumpeter point" like Godwin's point. Because let's not fool ourselves, there will be no creative destruction. No more than there was, by the way, during the previous rupture. This revolution would be a social disaster, I was certain. The Titanic's iceberg, except the size of Australia.
Wait… To tell the truth, I'm not being honest with you.
My doubts started well before all that. As early as 2015, they began to appear in my sleep, not about the technology itself, but about those who design it and those who control it.
Make no mistake, I repeat, I deeply love technology, which in itself is neither good nor bad. Every technological advance generates positive effects and negative effects in its wake. While we logically try to maximize the former, we unfortunately haven't yet found a way to reduce the negative effects to zero, mainly because a good portion of them are unpredictable (see Jacques Ellul and his many studies on the subject) or their risks are low and we don't consider them.
It's reading "Hooked" by Nir Eyal in 2015 that struck me deeply: he describes in detail the near-scientific methods for designing extremely addictive applications and software. The principles laid out served as the design brief for the development of platforms like Facebook, Instagram, TikTok, Twitter and so many others. Their addictive nature is the direct cause of their harmfulness to mental health, as attested by numerous scientific studies.
Then, still in the "memorable" category, came the Cambridge Analytica episode, Brexit, the manipulation of opinion by tech and platforms. That big data and that cloud presented as so innocent, that liberating Web 2.0, supposed to connect people, transform the visitor into a web actor…
The puzzle was assembling piece by piece, but the image promised on the box looked nothing at all like what was actually being plotted.
So, forgive my naivety, technology can also - and this is what is worrying / revolting - have negative effects that are wanted, conscious, even orchestrated.
Today, a significant part of tech aims to capture as much of users' time and attention as possible, to track and profile their every action, to predict and influence their opinions. This kind of tech, which I personally consider abject, was springing up like the poisonous mushroom in Tintin's "The Shooting Star," day after day, more powerful and threatening.
In short, my doubts have existed for 10 years.
And these new questions linked to AI now revive the first "bitterness" that had perhaps faded a little or lost some of its force.
It's incredible, this capacity we have to forget, sometimes even the most enormous things…
Today, I'm in the most introspective period of my life and, no, it's not just a pre-mid-life crisis.
Despite everything, something is off in all this noise.
Never in history has a technological development seen such enthusiasm, not even close. Not the internet, not the computer, not the butter knife despite being very disruptive and of capital importance to humanity. It's true that the hyper-connection of our information society allows AI to spread exponentially, which wasn't the case for previous ruptures.
This massive dynamic is orchestrated first and foremost by the designers of these AIs and all those who gravitate around them (infrastructure, hardware, software solutions, etc.). The "usual suspects" are at the controls: Microsoft, Google, OpenAI, Anthropic, Amazon, Apple and of course, the sun above them all, NVIDIA.
To do this, they've pulled out the good old carrot and stick.
Artificial General Intelligence, AGI, is the utopia of reaching an AI with the level of intelligence of a human being, capable of being generalist (today AIs are most often confined to restricted domains). It is indeed a carrot for employers, business leaders and the patronage. It is also one for the markets (delirious at the idea) and more generally for Technocapitalism.
Because indeed, in the hypothesis - which is not certain given the current state of the art - that science would manage to develop such an AI, then this instrument would be of a nature to transform the human labor of hundreds of millions of white-collar workers into GPU cycles (AI-specific computations). From human agents we would move our economy to silicon and AI agents.
Consequently, it could be the greatest transfer of value (some will say dispossession) in the history of humanity, and also, very certainly, a multi-systemic earthquake. I can hardly see a scenario in which (techno)capitalism, as we know it, would continue its course "normally."
Put differently: it would put back into play economic, societal, and even anthropological fundamentals on a global scale.
You'll grant me that this constitutes a pretty big stick (at least as big as its carrot counterpart) for the few hundred million white-collar workers our planet counts. Roughly, everyone who works in front of a screen, you, me, and all the others.
Close your eyes, imagine that AGI, in fact, is Thanos in Avengers Infinity War, the scene where one person in two turns to dust. What's sad here is that even if I'm right, it will never be as cinematically beautiful.
Despite all this… Yes, despite all this, reassuring discourses also make themselves heard here and there to avoid general panic, emanating notably from public authorities, institutions and the big Tech themselves.
Catchphrases or gimmicks like "AI won't replace you but someone using AI will" or the famous "memes" proliferate to overdose levels on the internet. The point here is to make people "#just-scared-enough," but not too much, because panic is never good for the market.
It recalls the Titanic's captain shouting: don't worry, there are lifeboats… Of course, except everyone knows that at the end of the film Jack dies because there isn't enough room on the plank, sorry for the spoiler.
AGI is therefore a powerful marketing tool, supporting a storytelling at the measure of its excess.
But it is also, and this is new, a major geopolitical weapon.
Reaching AGI could - one more promise from big Tech - lead toward ASI (Artificial Super Intelligence), an AI that would supplant any human intelligence in any domain. So, to keep up with me, ASI is the "Final Boss" version of AGI, except you can never, ever beat it. (This poses, in my view, a major ontological problem but I still need to work on this philosophical point).
In any case, this is why the great powers of this world (mainly US and China) approach AI with an effort similar to the Manhattan Project last century. Here, it's not about reaching the atomic bomb but AI supremacy, hence AGI which will open wide the doors to ASI. (A bit like reaching first the A-bomb then the H-bomb).
AGI is a national security issue that must be reached before the opposing camp, whatever the cost. In the blink of an eye, AI has come out of its laboratory to be elevated to the rank of a major international strategic issue. And a few additional seats are being added at the table of the powerful to welcome the bosses of Big Tech.
Let's briefly summarize the situation: we have the Big Tech, for whom humanity's well-being is not precisely the first concern. These companies are driving the greatest technological (and financial) movement in history, dragging along the two world superpowers.
At bottom, this storytelling is finely chiseled, isn't it?
The media spread it just right, the Big Tech blow hot and cold, the politicians surf the little waves it generates or play a life-size game of Risk.
All of it heightened by a generalized hypnosis: those who don't kneel before AI are quickly catalogued as decelerationists, degrowthers, or worse, boomers. A little more and you'd say the world is running to its end (this is where I open a second bottle of viognier…). These apocalyptic hypotheses, all hammered home with force, seem highly probable and irrationally seductive. Almost spellbinding. They have kept me awake, or curled up in fetal position in my bed, on certain nights.
But through endless reflection (some of it in fetal position), bulimic reading (some accompanied by viognier), therapeutic writing and many discussions with real people and not just with Claude, I've ended up understanding something.
The real danger is absolutely not where these professional storytellers want us to look.
When you no longer know who you are, you no longer understand the world or what you're doing in it, the solution is always the same: go reread the philosophers. They were in exactly the same state. Except now they're dead.
Hannah Arendt wrote:
"The living humanity of a man declines insofar as he renounces thinking."Hannah Arendt
What if the real danger were quite simply: our thinking?
AI interfaces as they were designed, all the chatbots, were conceived to make us believe, to simulate, to feign that we are dialoguing with another consciousness and not a machine or an algorithm.
Even those who know better fall for it because appearances are now too deceiving. As proof, my last article.
The lexical field of all messages between these AIs and the user are actions or adjectives attached to the human. When they say "thinking" before "responding" to you; it would be less interesting for the user to read "inference in progress, please wait." But it would be more accurate and honest.
After launching the "memory" feature, the Claude chatbot just added a new one: skills (compétences in French).
And what to say, finally, of the eternal flattery with which these AIs respond to us? Whatever you type into the interface, you are anyway "perfectly right," "incredibly lucid"…
But wait, doesn't this remind you of something?
Yes, it's positive feedback supposed to release a small dopamine hit in users' brains. The exact same mechanism that likes, followers and other vanity metrics on Facebook, Insta, TikTok and others used to addict the planet…
You also find this anthropomorphic narrative in the mouths of designers and executives. Take Sam Altman for example, who tells the world it's preferable to say "Please" and "Thank you" to ChatGPT.
Yet it is indeed a machine, a program, an algorithm and NOT a consciousness. And even if it were a consciousness, let's grant it for a moment, it is absolutely not "your consciousness."
It is external to you. But everything was thought from the design stage with anthropomorphism as a goal of its creators. Because anthropomorphism is a feature of AI chatbots. Their storytelling and marketing are also profoundly anthropomorphizing.
For Socrates, thinking is "a conversation the soul holds with itself."
This Socratic "two-in-one" hypothesis, taken up later by Arendt, indicates that thinking implies being "in one's own company" and that from this discussion with oneself a minimal moral conscience can be born.
She adds that this Socratic "two-in-one" is the very structure of consciousness.
This inner dialogue between self and self is human consciousness.
The risk lies in the fact that with these chatbots disguising themselves as "another" pseudo-external consciousness, the inner dialogue gradually loses ground and we systematically "delegate" it to a system (AI, bureaucracy, ideology).
That we no longer experience ourselves as a two-in-one but as a "one-with-a-system" which could then morph into an "assistant-of-the-system."
Pushing the reasoning even further, the absolute risk would be the systematic replacement of inner dialogue by an external authority "that knows better than oneself" (AI).
This is what Arendt calls "thoughtlessness" (when one loses this dialogue with oneself).
I'll end with this quote found in a recent thesis (Leifer, V. A. (2024))
"If the absence of thought becomes massive, totalitarian movements have ideal ground and democracy rapidly erodes."V. A. Leifer (2024)