Technological Singularity

The technological singularity is a speculation about the future, not an observation. It describes a possible point in time at which machines outperform the best humans at all intellectual tasks. The crucial idea: such a machine could then design new, even better machines itself. These, in turn, design even better ones, and at an ever-increasing pace. Humans might eventually be unable to keep up with this chain, because it unfolds too quickly and becomes too complex. The word “singularity” is borrowed from physics, and here it means: a boundary beyond which our usual predictions no longer work.

Why a speculation moves so much money

The idea is old, but it is once again omnipresent. Heads of major technology companies openly talk about wanting to build machines with human or superhuman thinking ability. Such statements are not merely philosophy—they influence investments worth billions. Anyone who believes this point is near is building data centers and buying chips years in advance today.

At the same time, the singularity is a central argument in the debate about risks. If a technology improves itself, a control problem arises. A machine can no longer simply be switched off once it can think through situations better than its operators. That is why governments and companies fund research into how the goals of AI systems can be permanently tied to human interests. This field of research is called alignment.

Many experts, however, consider the thesis to be greatly exaggerated. Their criticism: the singularity is more of an article of faith than a forecast, since it can hardly be verified. They warn that distant doomsday scenarios distract attention from today’s problems. These include discrimination through software, copyright, disinformation, and the electricity consumption of large models.

The logic of the self-accelerating chain

The argument rests on a feedback loop. A feedback loop occurs when the outcome of a process reinforces the process itself. A microphone placed in front of a loudspeaker is the classic example: a quiet noise gets amplified, picked up again, amplified even more, and after a short time it starts to screech. Applied to technology: better AI builds better AI. Experts call this process an intelligence explosion.

A prerequisite for this is that research itself can be automated. Today, AI programs already assist with parts of this, for instance by writing program code or by trying out different model variants. However, practice is still far removed from a system that independently develops new fundamentals. Current models learn from vast amounts of text and do not become smarter on their own after training.

Hard limits speak against the chain reaction. Chips have to be manufactured, and factories for this take years to build. Electricity and cooling water are finite, and experiments in physics or medicine require real time in the lab. Moreover, progress often becomes more expensive the further one advances: the last few percentage points of accuracy cost a multiple of the first ones. An explosion requires that none of these brakes take effect.

From science-fiction motif to headline

One encounters the term mainly in interviews, books, and stock market commentary. It was popularized by the inventor Ray Kurzweil, who has for years placed the timing around the year 2045. In news articles, the singularity frequently appears alongside the abbreviation AGI for artificial general intelligence, meaning a system that does not just master one task but any intellectual task whatsoever.

An important distinction: AGI would be a type of system, while the singularity is an event that could follow from it. One can consider the former possible and the latter unlikely. Another common misconception is confusing the singularity with consciousness. The thesis says nothing about whether a machine feels anything—only about how capable it is.

In everyday life, none of this is noticeable, and that will remain the case for now. Yet when politicians argue over AI laws or companies announce their next models, the idea resonates in the background. Anyone following the debate should therefore check, with every statement, whether it describes a measurement or an expectation.