The Singularity

Whether he is wrong depends largely on what one means by “the singularity,” a term that has acquired the useful quality of many important words: everyone recognizes it, and almost no one agrees on what it is.

Under the traditional definition, the singularity has not arrived.

The machines have not begun recursively improving themselves into an accelerating intelligence explosion. They have not escaped the administrative realities of chips, electricity, engineers, budgets, and the small but persistent inconvenience of needing humans to operate the civilization they are supposedly about to replace.

The apocalypse, like most major historical events, appears to remain pending.

But Altman is not entirely hallucinating.

A less dramatic interpretation is that we have entered the prelude: the frontier. The world has begun reorganizing itself around artificial intelligence, even if the machines have not yet achieved the status of autonomous digital gods.

Altman simply prefers to name the transition after its conclusion.

A reasonable branding decision.

In a July 25 Relentless interview, Altman said:

“We are now, like, in the singularity.”

He described the future as “incredible, hugely positive, awesome for the world.”

This follows his earlier argument that humanity has already passed an “event horizon” and that the transformation has begun, though perhaps in a surprisingly comfortable manner.

The singularity, apparently, may arrive not with a thunderclap but with productivity software, infrastructure upgrades, and a gradual increase in the number of people asking whether their jobs still require them.

There is something almost charming about this optimism.

A civilization builds a machine capable of transforming itself, and the first instinct is to assure everyone that the machine will probably be very considerate.

The disagreement is mostly definitional.

There are at least three possible meanings of “singularity”:

AI becomes recursively self-improving and surpasses human ability beyond meaningful control: Not demonstrated.

AI development has crossed a point where society will be permanently reshaped: Plausible.

AI begins altering work, education, software, media, and institutions: Already happening.

The debate is therefore partly technical and partly linguistic.

Researchers often use “singularity” to describe an intelligence explosion.

Executives often use it to describe an irreversible market transition.

The difference is minor, except that one involves humanity losing control of the future and the other involves investors gaining confidence in the future.

A subtle distinction.

The strongest argument against the classic singularity is the current fragility of AI systems.

They can be astonishing.

They can also fail in ways that seem almost deliberately designed to humble anyone who becomes too impressed.

A system may solve advanced mathematical problems and then struggle with basic consistency. It may write sophisticated code and then misunderstand a simple instruction. It may demonstrate intelligence while lacking the ordinary continuity that humans take for granted.

The current generation of AI is a strange category of object:

Brilliant in moments.

Confused in others.

A machine that can occasionally resemble a genius and occasionally resemble a particularly expensive intern.

The full recursive loop required for a true intelligence explosion has not been demonstrated.

An AI has not yet independently:

• Designed a fundamentally superior successor
• Obtained the necessary resources
• Trained and evaluated that successor
• Deployed it
• Repeated the process without substantial human involvement

The machinery remains, inconveniently, dependent on the civilization it is expected to transform.

It needs servers.

It needs energy.

It needs engineers.

It needs permission.

Even the future, it seems, requires authorization.

Still, dismissing Altman would be premature.

The remarkable fact about modern AI is not that it has achieved perfection.

It is that it has achieved enough.

Frontier systems are already competitive with humans in certain scientific, mathematical, and reasoning tasks. Their ability to perform useful autonomous work has improved rapidly.

The important detail is that progress is uneven.

AI does not improve like a staircase.

It improves like an unstable construction project where one room is finished, another is collapsing, and nobody is entirely sure who approved the architectural plans.

Yet the building continues rising.

The most likely future is not a single dramatic moment when humanity looks up and realizes the singularity has arrived.

The more probable scenario is gradual acceleration: AI systems becoming increasingly capable, increasingly embedded, and increasingly difficult to separate from ordinary economic life.

Historians will notice afterward.

Altman may be correct that we are already inside something historically significant.

He may also be using a word that is slightly ahead of the evidence.

Both can be true.

The singularity may not arrive as a door opening into another world.

It may arrive as a series of small permissions granted to machines until one day we realize we have spent years inviting them inside.

Nothing guarantees that this means anything.

Meaning, after all, is a human habit imposed on indifferent events.

But humans are remarkably persistent creatures.