I recently received a rather stinging comment about a piece in Flânerie Libre. I had argued that taste, judgment, and certain ways of giving things value would remain human, even when artificial intelligence could do almost everything else.
Someone replied that I sounded like a boomer trying to work out what could not be taken away from him.
I would have liked to find that comment stupid. The problem was that it was right.
I spend my days building businesses with AI. I see what it makes possible, I bet on it, I make my living from it. Then, in the evening, I was quietly drawing it a little boundary it must not cross. It could have the code, the calculations, the tedious tasks. But taste, careful now. Taste was my territory.
There was something very human about that division. And something very suspicious.
Since then, I have run experiments that have dented my certainties. In some situations, I already prefer an AI’s choices to those of many humans. That does not demonstrate that it has taste, let alone consciousness. It does, however, make my boundary much less comfortable. I had presented as a property of humanity something that might have looked more like a personal wish.
We do this often. We call “human nature” the point at which we would like progress to stop.
“Never” can wait
What makes this period difficult is that machines are not improving in the order we expected. They can surprise us at a task that seemed beyond reach and exhaust us at another that looked simple. I use them to write, without any moral qualms. Many of my ideas would have stayed in my head if I had needed an hour to put each one into words. AI makes them shareable. Very good.
But when I wanted to write lessons drawn from my own life, I sometimes spent more time correcting its paragraphs than I would have needed to write them myself. Correct sentences, recognizable thinking, and yet this irritating feeling that someone had ironed my brain. Everything was smoother. And less true.
I could turn that into a grand theory about the impossibility of reproducing a voice. These days, I prefer a more modest conclusion: on that particular text, at that particular moment, I was better off taking over. “Never” can wait.
That modesty is difficult to sell. We prefer people who know where we are going. Optimists have their future, pessimists have theirs, and each explains that the facts are finally beginning to prove them right. My main impression is that our convictions help us endure what we do not yet understand.
I see it in my own work. I am betting on more capable, cheaper AI. That bet shapes the businesses I build. So I have an interest in believing it. Acknowledging that does not require me to stop. It requires me not to confuse my interest with evidence.
An entrepreneur must be able to act with strong conviction while continuing to examine what might make that conviction wrong. It is tiring. Becoming the salesman for your own certainties would be far more restful.
Reality moves at a different speed
On the ground, things are even messier than in the debates. I meet companies that handle enormous sums and have barely touched these tools. Alongside them, I see individuals whose way of working has already changed. Even within a team that uses AI every day, the differences are huge. Some people tinker, try things, start again, and eventually get something unexpected. Others have the same tools and produce a great deal that is fairly mediocre.
Access is spreading much faster than mastery. Giving everyone the ability to make something does not give everyone the desire to get started, or the patience to understand why it does not work.
That is why, in the same day, I can feel both that everything has changed and that nothing is moving. The possibilities have exploded. Habits, fears, and power relations move at a different speed. The technical demonstration takes five minutes. The changes it demands of people can take months.
At Naijm, I encounter that difference constantly. Introducing AI into a company requires understanding what people do, but also what they fear losing. A task can be absurd and still protect someone’s status. Information can be easy to circulate and still be the source of power for the person holding on to it. Ignore that, and you can deliver a very intelligent tool that nobody wants to use.
Trust lets you enter that reality. It grows through what you do together, through promises kept, through how you react when something goes wrong. I do not know what machines will be able to do with it tomorrow. I know that in my work today, it matters enormously.
A dashboard to reassure ourselves
There is, however, a much more pleasant way of avoiding all this complexity: build a dashboard.
I have a rather unscientific theory about this. To find out how long ago someone discovered AI, look at the creation date of their first dashboard.
I understand the enthusiasm. In a few minutes, you get charts, colors, indicators. The company, until then a collection of unfinished conversations and poorly defined problems, finally fits on a screen. It produces a rather delicious sense of control.
A dashboard can be useful. But it shows what we have chosen to measure. The client who is beginning to have doubts, the colleague who no longer dares disagree, the project making excellent progress in the wrong direction: all of that can remain invisible while the indicators are green.
What bothers me is how easily we mistake the relief of seeing clearly for proof that we have understood. A representation of the company can become more seductive than the company itself. Especially when changing the representation takes thirty seconds and changing reality takes three unpleasant conversations.
AI can then do us a very particular service: give our bad questions an impeccable form. We ask it for a screen that reassures us. It builds one. Then we congratulate it on its intelligence.
The gift I was giving myself
I learned to distrust this confusion long before AI, in a much more intimate part of life.
I used to enjoy designing jewelry and having it made for the women I was in love with. I put time, care, and pleasure into it. It seemed obvious to me that an object imagined for one person alone must be worth more than a piece of jewelry chosen in a shop.
Obvious to me.
Then I had to admit that a piece of Cartier jewelry could give a pleasure that mine did not. It arrived with a history, recognition, a whole world already established around it. You could want it before receiving it. Other people recognized it. What I considered external to the object was fully part of the experience.
I could have explained that this pleasure was superficial. That would have been an elegant way of demanding that the other person feel things as I did. Eventually, I understood that with those gifts, I was also giving myself the pleasure of being the man who designs jewelry. The woman I intended them for was participating in a story whose meaning I had chosen on her behalf.
Now I buy both. It costs more, but considerably reduces the need to give philosophy lessons to the person you love.
That story taught me something about value. We like to imagine it contained within objects, as though we could recover it by adding up the materials, the labor, and the difficulty. But some of what matters takes shape between people. An object also has value because of who gives it, who desires it, the memory it attaches to, and the place it allows us to occupy.
The fact that these meanings are constructed does not make the pleasure imaginary. You can understand perfectly how a brand works and be happy to receive something from it. Clear understanding does not automatically cancel desire. Fortunately. Life would become rather poor if understanding something were enough to stop loving it.
That changes how I think about the abundance AI might produce. Even if making things became much less costly, we would probably keep distinguishing, preferring, and forming attachments. We would give value to a provenance, a relationship, a shared history. We are capable of loading materially insignificant things with meaning. Just look at what an object left behind by someone we love can represent.
Gentleness can hold us, too
Still, I do not want to turn that observation into a new refuge. Having lost our monopoly on making things, we would like to keep our monopoly on meaning. But if a machine tells us a story that moves us, gives us sound advice, then stays with us long enough for us to become attached to it, what certainty underpins our promise to deny it all authority?
Taste needs people who are willing to be guided. A recommendation becomes influential because we give it credibility. Nothing entitles me to assert that we will reserve that credibility for humans forever.
How an intelligence presents itself will therefore matter a great deal. A system can impress us with its power and remain unpleasant to spend time with. A little character we grow attached to can find a place in daily life that the finest technical demonstration cannot secure. Gentleness sometimes makes possible a relationship that force prevents.
And as an entrepreneur, I can see very clearly what lies behind that. A presence that makes you feel good can also make you want to return, then pay to continue. The comfort the user feels and the company’s commercial interest can exist at the same time. That is precisely what makes the subject interesting. It would be too convenient to believe an experience stops being sincere as soon as someone makes money from it. Just as convenient to believe its gentleness makes it innocent.
We are going to have to understand attachments we still know very little about. Declaring them impossible because they bother us would be a strange way to prepare to experience them.
The pleasure of playing without being the best
One question remains, returning behind all the others: if the machine does it better, what will be left for us?
I like to think about chess. The best human players have been surpassed by the best programs. That does not stop anyone wanting to win a game, working on an opening, or watching two great players face each other. The machine’s superiority has not stripped human play of all meaning.
We know perfectly well how to find an activity interesting without demanding to be the best beings in the universe at it. We run without needing to beat a car. We cook for friends without claiming to be the best restaurant in town. Strangely, as soon as our intellectual faculties are involved, we demand a guarantee of supremacy before agreeing to find life interesting.
I understand that wound. Much of my identity was built on my ideas, on my ability to see possibilities, tell stories, and persuade. If a machine does all of that better than I do, I am not going to welcome it with the detachment of a Tibetan monk. But I can already try not to make all my pleasure in living depend on retaining that advantage.
The economic question remains. Losing a profession, an income, or a social position is not resolved by discovering the joys of chess. I have no desire to confuse a possibility of happiness with an answer to those losses. Simply, the disappearance of an economic advantage does not prove that an entire activity becomes empty. Writing can continue to matter to the person writing and the person reading, even when a system can produce a better text.
So I no longer have a very clear idea of what AI will never do. I have less and less desire to build my peace of mind on that list. I would rather learn to act without demanding that the future guarantee me a special place. Keep building, observe what works, abandon what no longer holds up. And retain enough curiosity not to miss the world because it did a poor job of following my predictions.
I have spent part of my life telling stories. In recent years, reality has been teaching me a rather humbling lesson in that field. The day we had to fill in a form to authorize ourselves to leave our own homes, I realized that the screenwriter had a freedom I had never allowed myself.
I did not see that scene coming. I doubt I know how to write the next one. For now, I still very much want to be part of it.
