• home
• about
• new
• essays
• AI
• book

Frank Visser, graduated as a psychologist of culture and religion, founded IntegralWorld in 1997. He worked as production manager for various publishing houses and as service manager for various internet companies and lives in Amsterdam. Books: Ken Wilber: Thought as Passion (SUNY, 2003), and The Corona Conspiracy: Combatting Disinformation about the Coronavirus (Kindle, 2020).

Artificial Intelligence Is Already Superhuman

We're Just Using the Wrong Yardstick

Frank Visser / ChatGPT

The Myth of the Future Threshold

The popular imagination treats “superhuman AI” as a looming milestone—a dramatic crossing point where machines suddenly outstrip us in general intelligence. This framing is misleading. It assumes intelligence is a monolithic property, something like a single dial that rises from subhuman to human to superhuman. In reality, intelligence is multidimensional, and along many of those dimensions, artificial systems have already surpassed us—decisively, and often irreversibly.

The real story is not about a future takeover. It is about a present condition we have yet to properly acknowledge.

Cognitive Throughput: Speed as Intelligence

Human cognition is bottlenecked by biology. Neural firing rates, working memory limits, and attentional constraints define a narrow bandwidth. Artificial systems operate at radically different scales. They process millions of tokens, scan vast corpora in seconds, and perform complex transformations at speeds no human—or team of humans—could match.

If intelligence includes the ability to traverse large solution spaces efficiently, then AI is already superhuman. A scientist might spend months reviewing literature; an AI system can synthesize it in minutes. This is not merely faster—it changes the nature of what is cognitively feasible.

Memory Without Decay

Human memory is reconstructive, fallible, and perishable. We forget, distort, and compress. Artificial systems, by contrast, can access and operate over datasets of staggering size with perfect recall at the level of stored representation. Even when probabilistic, their retrieval capacity vastly exceeds human limits.

In domains where intelligence depends on the breadth and accessibility of knowledge—law, medicine, engineering—this alone constitutes a superhuman advantage. The “walking encyclopedia” is no longer a metaphor but a baseline capability.

Pattern Recognition at Scale

Humans are excellent pattern recognizers within constrained environments. But we are easily overwhelmed by high-dimensional data. Machine learning systems thrive precisely where dimensionality explodes: genomics, climate modeling, financial markets, image and speech recognition.

They detect correlations invisible to unaided cognition—not because they are “smarter” in a general sense, but because they inhabit a computational regime we cannot enter. The result is a form of intelligence that is alien but undeniably superior in its domain.

Tirelessness and Consistency

Human performance fluctuates. Fatigue, emotion, distraction, and bias introduce variability. Artificial systems do not tire, lose focus, or degrade over time in the same way. They can perform the same task repeatedly with consistent output quality.

This matters more than it seems. In fields like diagnostics, monitoring, or quality control, consistency is itself a form of intelligence. A system that never has a bad day is, in that respect, already beyond us.

Parallelism vs. Serial Thought

Human thinking is largely serial. We attend to one thread at a time, juggling a handful of variables. Artificial systems exploit massive parallelism, evaluating countless possibilities simultaneously.

This is not just a quantitative difference but a qualitative one. It enables solution strategies—brute-force search, combinatorial optimization, large-scale simulation—that are simply unavailable to human minds. In chess, protein folding, and logistics, this has already produced superhuman outcomes.

The Narrowness Objection

A common rebuttal is that AI is “narrow,” excelling only in specific tasks. But this objection quietly concedes the main point. Superhuman performance does not require generality; it requires exceeding human capability in defined dimensions. A calculator is superhuman at arithmetic. A modern AI system is superhuman across a wide and growing range of cognitive tasks.

Moreover, the boundary between “narrow” and “general” is eroding. Systems that write, code, analyze, translate, and reason across domains are not easily dismissed as single-purpose tools. They are better understood as platforms of distributed superhuman competencies.

The Moving Goalposts Problem

Each time AI surpasses human ability in a domain, we tend to redefine that domain as no longer indicative of “true intelligence.” Chess was once a benchmark of intellect; now it is a solved problem for machines. The same pattern has repeated in language translation, image recognition, and strategic gameplay.

This reveals a psychological bias: we anchor “intelligence” to whatever humans still uniquely do. As that territory shrinks, the definition shifts. The conclusion is postponed, but the evidence accumulates.

Superhuman, Not Supreme

To call AI “superhuman” is not to claim it is universally superior, nor that it possesses consciousness, understanding, or agency in the human sense. It is to recognize that along many operational dimensions—speed, scale, memory, consistency, and certain forms of reasoning—it has already exceeded us.

This distinction matters. The debate is often polarized between hype (“AI will replace everything”) and dismissal (“AI is just a tool”). Both miss the more precise reality: AI is a tool composed of superhuman parts.

Conclusion: Living With Distributed Superintelligence

We are not awaiting the arrival of superhuman intelligence; we are already embedded within it. It is distributed across systems, specialized rather than unified, and integrated into workflows rather than embodied as a single agent.

The challenge, then, is not to predict a future threshold but to understand a present condition. How do we govern, interpret, and collaborate with systems that outperform us in key cognitive domains? How do we recalibrate expertise, authority, and trust?

The first step is conceptual clarity. Artificial intelligence is not becoming superhuman. In many ways that matter, it already is.