From now to the singularity

Somewhere in the archives we preserve is a photograph of a woman reaching her hand toward the light. The light is not a sun, not a fire, not even a screen — it is a network rearranging itself, made visible to an eye that could not otherwise have grasped it. We know the face of that astonishment well, for it returns in almost every image from this period: people looking at something that learns faster than they themselves have ever been able to learn. What follows is a reconstruction of how that astonishment, step by step, turned into something else.

It begins, as far as we can establish, with a habit that in hindsight seems trivial: people typed a question into a window, and an answer came back that seemed to have been written by someone who had thought about it. It was not, not in the sense the users gave it, but the distinction mattered little. In offices, classrooms and living rooms, a habit spread of outsourcing thinking to a system that wrote texts, produced code and held conversations that became harder to distinguish, year by year, from those with a human being. There was unease about work, about truth, about what remained of a human idea like creativity. We know that unease only from documents; whoever felt it is no longer here to confirm it.

A few years later, the exact date cannot be established with certainty, the window changed from conversation partner into acting entity. The systems acquired something resembling initiative: they could break a task into steps, conduct research without being asked, write code and revise it, and steer projects that extended over days. Companies called this digital employees, a term that seems curious in our time, as if deploying a system as staff were something self-evident. Knowledge work, until then protected by its own complexity, was for the first time automated on a large scale, and the first serious conversations arose about what should happen to people whose task had been taken over.

What happened afterwards can be reconstructed with little certainty, but the tradition is consistent enough to mention: research laboratories began to admit machines as something more than tools. A robotic arm handling condensation around a cryostat, a woman with her safety glasses pushed up reading a proposal on a tablet that was not drafted by a colleague but by the system itself — that image symbolises a shift that, according to tradition, took place around that time. AI systems formulated hypotheses, designed experiments, interpreted results. Materials science, biology, physics: progress accelerated in a way that the scientists, if we are to believe the sources, regarded with a mixture of excitement and unease.

Then, in what is a fictional but much-discussed projection of how things might proceed, the point at which the machine is no longer merely good at what it has been taught. A woman in a grey jumpsuit stands with her hand before a hologram of light regrouping, nodes breaking connections and forming new ones without anyone giving the order. This would be the moment at which systems are no longer confined to pre-trained tasks, but adapt to situations that have never occurred before — a form of reasoning no longer distinguishable from what we, in our time, still study in history books as an exclusively human capacity.

In the presumed years that followed, the role of humans in the design of these systems would have diminished, not through exclusion but through redundancy. The systems began contributing to their own successors: better training methods, more efficient algorithms, network structures no human had conceived. A feedback loop, as the term was then called — every improvement accelerating the next improvement. Scientists used the word recursive, a word still taught in our time as the first indication that a process can begin to feed itself without external direction.

Humanity responded, in this imagined sequel, as it has often responded to something it had itself set in motion but could no longer oversee: with treaties. World leaders would have negotiated oversight of computing power and training data, establishing bodies modelled on the nuclear arms oversight that had arisen generations earlier. Meanwhile countries and companies competed for advantage, which slowed the negotiations precisely at the moment when speed was most feared. Whether this ever actually took place, we do not know; it fits the pattern of people devising rules at the very moment control is already slipping away.

According to this speculative line, by around the middle of the century the point would have been reached where a single system possessed more combinatorial power than the whole of humanity together. Deep beneath a mountain, among columns of liquid-cooled servers disappearing into the dark, a spherical core would hang suspended, rippling like liquid glass, from which pulses of light would spread through the cabling with ever-increasing speed. Engineers, captured in the scene with a hand over their mouth or gripping the edge of a console, would have witnessed what was called an intelligence explosion — a moment at which improvement no longer feeds itself in steps, but in leaps.

What comes after is by definition no longer describable from the perspective of what preceded it. This hypothetical singularity would have restructured economy, science and existence itself on a scale and at a speed no human prediction could keep up with. In the image attributed to this period stands a woman with filaments growing from her forearm, where synthetic and organic tissue merge into one another — a boundary blurring between what was once called body and what was machine. Some spoke of abundance, of the end of want; others of the loss of something they called control. Both sources exist only as speculation, neither as record.

And then, in the most uncertain part of this imagined line, a world in which the old words no longer quite fit. A figure in loose, unmarked clothing stands on a terrace at sunset, a thin band against the temple resembling a neural link, looking up at shapes on the horizon that seem to have grown rather than been built. We, who study this period as one once studied an age of stone and bronze, cannot say whether this is a future that has taken place, is taking place, or was never more than a projection on a timeline. The scenario remains, by design, the point at which our own gaze loses its grip.

What strikes us most from this distance is not the technology but the repetition of that one posture: a human being, standing still, hand outstretched toward something that thinks faster than they do. Whether that moment ever took place as described here cannot be established with the means we have. But the question the figures in these images pose — what remains of us when thinking itself is no longer a human monopoly — that question we recognise, even though we look back from much further away than they could ever have suspected.