Each time a machine learns to perform a task we believed to be our own, there are those who hasten to claim that the task was never truly significant. When they calculated better than us, we asserted that calculation was mechanical. When they triumphed at chess, we maintained that chess was merely calculation governed by rules. When they began to compose coherent paragraphs, we insisted that writing is not thinking. Historians of artificial intelligence identified this reflex decades ago: every achievement of the machine ceases to be regarded as intelligence as soon as its mechanism is understood, and the truly human element retreats one boundary further, always to that which they cannot yet perform.
It is easy to interpret this reflection as a retreat, as if we were losing ground and seeking consolation. I interpret it in reverse. I have no interest in joining those who fear that artificial intelligence will steal what makes us human, not least because I do not believe it to be true and because I see how much it contributes: there are systems that have accelerated the prediction of protein structures, that solve mathematical problems of olympiad level with every step verified, that refine diagnostic imaging. I do not see it as a threat; it is an extraordinary tool in hands that know how to use it. And the question I ask myself concerns us.
If every capacity the machine acquires ceases to appear essential to us, perhaps the issue is not that we are being emptied: it is that we have spent centuries searching for ourselves in the incorrect location. We have catalogued the human as a list of proficiencies: calculating, memorising, reasoning, executing with dexterity. And one by one, those proficiencies prove to be replicable. The machine does not remove anything that was truly ours; it removes the comfortable answers, those that conflated what we are capable of doing with what we are. Each boundary that falls is a false definition that collapses with it. What remains thereafter is not less, but more refined: the unadorned question of what a human being is when they can no longer be defined by any task that a system cannot imitate.
I will not answer that question because to answer it would be to repeat the error. If I were to say now that “what makes us unique is consciousness”, or creativity, or bonds, I would be doing the same thing again: placing a piece on a square, waiting for someone to build the machine that occupies it and forces me to move again. Human singularity, if the word means anything, may not be a property that one possesses and can lose, like someone losing a position in a competitive examination. Perhaps it is something else: not a skill, but the fact that no one occupies the place from which one acts. Hannah Arendt distinguished between what someone is, which is described by capabilities and attributes, and who someone is, which only appears when acting and speaking among others, and which cannot be reduced to a list. The machine can equal the what; the who is not a superior skill: it is of a different order.
I do not maintain that “the human” has ever been a pure and stable essence that is now under threat. It was not: for centuries, the definition of what is fully human excluded women, colonised peoples, and those who did not fit the reason of the one defining. The machine does not destroy a clean boundary; it illuminates that the boundary was never innocent. Nor do I believe that it is the machine that “forces” us to do anything, as if it possessed a will. We are compelled by human decisions regarding what to build, for what purpose, and for whose benefit. Technology does not think for itself; it is used by corporations, States, and individuals. Attributing pressure to it is a convenient way to avoid looking at those who decide.
And here lies a matter of responsibility, not nostalgia. If the singularity of each person does not lie in what they produce but in the fact that they occupy a place that no one else occupies, safeguarding it does not consist of slowing down the machine or competing with it on its own ground, which I believe is a lost and, furthermore, absurd battle. It consists of not organising life in common as if the only thing that mattered were what is replicable, measurable, and productive. A society that only recognises what can be automated will end up treating people as slow versions of a machine. The task is not to protect skills that are no longer scarce, but to sustain the conditions in which someone can continue to appear as someone: non-productive time, care, the word that does not yield, the gesture that optimises nothing. That is not saved by technology, nor is it threatened by it: we decide that ourselves.
I do not know where human singularity lies, and I am suspicious of anyone who claims to be certain. But I am grateful that a machine capable of almost everything has stripped me of easy answers. We spent centuries believing that we were special because of what we knew how to do, and it turns out that almost all of that can be taught to something else. What has not been copied, yet, is not a more difficult skill: it is the very fact of being here, occupying a place, answering for what one does. It may be that there is nothing there to defend, no fortress. Only someone, unrepeatable, whom the machine knows how to imitate in every way except for the fact that it is him.
On open conversation
This text is not a critique of artificial intelligence, which I consider an invaluable tool in science, research, and many fields, but a question about ourselves: what is human singularity when the machine equals, one after another, the capabilities with which we defined it? I maintain that it does not subtract from it, it reopens it, and that the answer does not lie in taking refuge in the next skill not yet imitated. I distinguish the verified fact (the effect described by historians of AI, the real advances and their limits) from the philosophical interpretation, which I support with Arendt, Haraway, and Braidotti without attributing to them theses on AI that they did not formulate. If anyone wishes to intervene from the perspective of philosophy, the ethics of technology, science, or work with these systems, this notebook remains open.
Sources
Pamela McCorduck, Machines Who Think (1979), on the pattern by which every achievement of the machine ceases to count as thought (the “AI effect”). The formulation “AI is whatever hasn’t been done yet” is attributed to Larry Tesler; Douglas Hofstadter popularised it in Gödel, Escher, Bach (1979).
Hannah Arendt, The Human Condition (1958): distinction between “what” someone is and “who” someone is; singularity in action and plurality.
Donna Haraway, A Cyborg Manifesto (1985) and Rosi Braidotti, The Posthuman (2013): on the historical instability of the boundary of “the human”.
Verified contributions of AI with their limits: AlphaFold (Kovalevskiy et al., PNAS, 2024); AlphaProof and AlphaGeometry 2 at the 2024 Mathematical Olympiad (Hubert et al., Nature, 2025); AI in clinical practice (Han et al., The Lancet Digital Health, 2024).