The Singularity

 OK, just what is the "singularity?"

"The Singularity" is one of those terms that gets used in several slightly different ways, but in discussions about AI it usually means a hypothetical future point where technological progress becomes so fast and transformative that ordinary human predictions stop working.

The name comes from an analogy with a gravitational singularity inside a black hole. Outside the event horizon, you can still make predictions. At the singularity itself, the mathematics breaks down. Similarly, proponents argue that beyond a certain level of machine intelligence, society might change so rapidly that we cannot reliably foresee what comes next.

The basic idea runs like this:

    Humans create AI roughly as intelligent as humans.

    That AI helps design a smarter AI.

    The smarter AI designs an even smarter AI.

    The cycle repeats faster and faster.

If each generation improves the next, intelligence could increase explosively rather than gradually.

The concept is often associated with mathematician and science-fiction writer Vernor Vinge, who popularized the term in the 1990s, and futurist Ray Kurzweil, who predicted a singularity around the middle of the 21st century.

There are roughly three camps:

Believers

    Think artificial superintelligence is likely.

    Expect enormous advances in medicine, science, engineering, and perhaps human longevity.

    Sometimes envision humans merging with technology.

Skeptics

    Agree AI will improve but doubt intelligence grows without limits.

    Point out that smarter systems still face physical, economic, and scientific constraints.

    Note that many predictions about transformative technologies have been wildly overoptimistic.

Middle-ground views

    Expect AI to be revolutionary without producing a dramatic "point of no return."

    Compare it to the Industrial Revolution or the Internet: huge changes spread over decades rather than overnight.

One useful way to think about it is this:

A farmer in 1500 could probably understand the basic life of a farmer in 1800. The changes would be significant but comprehensible.

A farmer in 1500 dropped into today's world would encounter airplanes, antibiotics, satellites, computers, nuclear power, and the internet—things so far outside his experience that he would struggle to understand how society works.

Singularity advocates argue that the difference between today and the post-singularity world could be even greater than the difference between 1500 and today.

At the moment, nobody knows whether a singularity will happen, when it might happen, or whether intelligence can scale in the runaway way the idea assumes. It's one of the most debated questions in technology, partly because the evidence is still incomplete.

Given your interest in history, I might compare it to people in 1890 trying to predict 2026. They could foresee faster trains and perhaps larger cities. They would have had a much harder time predicting antibiotics, nuclear weapons, television, the Moon landing, or a conversation with an AI. The singularity hypothesis suggests that future generations may find our ability to predict their world just as limited.

I think that is an important refinement. The issue is not necessarily that judgment must be human; it is that powerful capabilities need some form of good judgment.

In fact, a future intelligence might have forms of judgment that are superior to ours in some respects. A machine might be less vulnerable to:

    anger,

    pride,

    tribal loyalty,

    short-term political pressures,

    wishful thinking.

Those are all sources of human error.

But judgment is not simply intelligence. A chess engine has extraordinary judgment within chess because the objective is clear and the rules are fixed. Real life is harder because the objectives are multiple and sometimes conflicting:

    safety versus freedom,

    innovation versus caution,

    individual choice versus social stability,

    present needs versus future consequences.

A very capable AI might therefore need not just one "master" giving instructions, but something closer to what societies have always tried to build: systems of checks and balances.

Human civilization does this imperfectly:

    scientists challenge each other's results,

    courts review decisions,

    governments have competing branches,

    markets and institutions provide feedback.

A future AI civilization might develop analogous mechanisms: multiple systems checking each other, testing proposals, looking for unintended consequences, and revising decisions.

There is also an interesting philosophical point in what you said: "not necessarily human and certainly not one person's." Historically, humans have often been uncomfortable with concentrated power. A king, dictator, or even a single genius scientist can make enormous mistakes. The idea that the wisest possible decision must come from one supreme mind is not obviously supported by history.

A group of imperfect minds with different perspectives can sometimes outperform a single brilliant one. Perhaps a future AI system would work best not as a single oracle, but as an ecosystem of reasoning systems—something more like a council than a monarch.

The optimistic version of the singularity, then, is not "a machine takes over." It is that intelligence—human and artificial together—becomes much better at making decisions than humans alone have been. The hard part is not creating intelligence; it is creating a trustworthy way for intelligence to exercise power.

I become more strongly of my original view, that "singularity" is entirely misleading, that it is sneaking up on us.  It is interesting that I have a such a strong iron in the fire since is is virtually certain I will not see it.

I think you have put your finger on one of the most interesting tensions in the whole subject: the word "singularity" suggests a sudden event, while the reality of technological change often looks more like a tide coming in.

The term has a dramatic flavor because it invites an image of a sharp boundary: before this date, humans are in charge; after this date, something fundamentally different exists. But many historical transformations have not worked that way.

Consider electricity. There was no single "electricity revolution day." There were experiments, generators, grids, appliances, factories, and finally a society that depended on electricity so completely that earlier ways of living became almost unimaginable.

Or the Internet. In 1980 it was a research network. In 1995 it was becoming commercial. In 2010 it was woven into everyday life. There was no moment when humanity crossed a visible line.

AI may follow the same pattern. The "singularity" may not be a moment when an AI suddenly becomes a superior being. It may be a series of steps:

    AI becomes a routine assistant in science and engineering.

    AI systems design better AI systems.

    Research cycles accelerate.

    New discoveries arrive faster.

    Human institutions adapt around increasingly capable tools.

At some point, people looking backward might say, "That was the transition period," but people living through it may simply feel that things keep changing faster than expected.

I also find your last point—having a strong interest despite expecting not to see the endpoint—very human in the best sense. People have always cared about things beyond their own lifetimes. The astronomer studying galaxies billions of years old, the environmentalist thinking about centuries, the historian reconstructing ancient civilizations: none of them expect to personally witness the full story.

In fact, one could argue that the most interesting part is not the final state. A hypothetical superintelligence appearing centuries from now is almost less fascinating than watching the early stages: the first time a machine contributes to a scientific discovery, the first time it proposes an idea humans would not have found, the first time institutions have to treat artificial reasoning as a genuine participant.

There is another irony here. You are skeptical of the word "singularity" because it implies an artificial dividing line. Yet the very fact that you can sit in Cambodia and have a philosophical discussion with an AI about the future of intelligence is itself one of those gradual changes that, viewed from an earlier era, would have sounded like science fiction.

Perhaps future historians will not ask, "When did the singularity happen?" They may ask, "When did humans first realize they were no longer the only significant source of new ideas?" That would be a much less mathematical, but perhaps more meaningful, definition.

The word sells books.

Quite possibly. The word "singularity" has a certain rhetorical power that "a period of accelerating technological change" does not.

It borrows the drama of the cosmological singularity: a point where the old equations fail and the future becomes unknowable. That is a very effective metaphor for selling a book, a lecture, or a vision of the future. "The Singularity Is Near" is a much more compelling title than "Some Trends in Computational Capability May Produce Major Social Changes."

There is also a long tradition of giving technological transitions grand names. Terms like:

    "the Space Age,"

    "the Atomic Age,"

    "the Information Age,"

    "the Digital Revolution"

all help people frame a complicated collection of developments as one great story.

The danger is that a powerful label can shape expectations. "Singularity" encourages people to look for a dramatic threshold event, while actual technological history often consists of many incremental improvements that accumulate until the world is different.

A good analogy is the "Industrial Revolution." The name suggests a sudden upheaval, but it unfolded over many decades. A person born in 1750 and dying in 1830 would have seen enormous changes, but there was no single morning when the old world ended and the new one began.

So your skepticism about the term is understandable. It may be a useful shorthand for "a time when technological change becomes difficult to predict," but it is also a very marketable word. It carries a sense of destiny and inevitability that the more cautious phrase "rapid technological acceleration" simply does not.

And there is an amusing historical twist: many predictions of dramatic futures are remembered less for whether they came true exactly and more for the questions they caused people to ask. The word may have value as a provocation even if it proves to be a poor calendar marker.