The Turing Error: Why AI Will Never Be Truly Intelligent, But Is Already Dangerous

In August 2026, after decades of unprecedented growth in computing power and the development of large language models, the scientific community faced a fundamental paradigm shift in the development of artificial intelligence. Leading computer scientist Peter Denning, in his new study "The Turing Error: Freeing Ourselves from the Yoke of Unintelligent Machines," argues that modern AI architecture will never reach the level of human thought. Moreover, the scientist warns that we are already living in an era of new threats created precisely by this technological blindness.

The Phenomenon of "Tacit Knowledge" as an Insurmountable Barrier

Denning's key argument is the concept of "tacit knowledge." This is a vast array of experience that people accumulate from birth and use daily but cannot fully express in words or translate into digital symbols. According to the scientist, it is precisely the absence of this basic element of human existence that makes creating truly intelligent machines impossible. Modern neural networks, no matter how powerful, operate only with explicit data, ignoring the colossal layer of human experience that lies beyond text and visual databases.

Why Scaling Does Not Solve the Problem of Understanding

Denning emphasizes that attempts to codify a database of "common sense" have continued for decades, but it turned out that expert knowledge cannot be reduced to a set of statements. A virtuoso violinist cannot convey the feeling of playing to a student, and a person cannot encode intuition or a premonition. "Modern large language models merely manipulate words without understanding their deep meaning," notes the researcher. A word is merely a symbolic designation, behind which lies a context dependent on irony, sincerity, humor, or sarcasm. Since every context relies on an infinite structure of preliminary situations, simply scaling up language models will not help them absorb culture or truly pass the Turing test.

The Threat of Unintelligent Networks: A New Type of Danger

The main threat, according to Denning, does not lie in the emergence of a superintelligence that will take over the world, as often shown in science fiction. The real danger lies in the emergence of networks with a low level of intelligence that will act powerfully, unpredictably, and potentially harmfully. The inability of machines to grasp tacit knowledge creates a chasm between humans and algorithms. Machines create their own type of "tacit knowledge" that humans cannot read, and vice versa. This leads to algorithms making decisions that do not take human values and context into account, causing chaos in social and economic systems.

Contradictory Data

While Peter Denning insists on a fundamental divide between human and machine intelligence, representatives of leading technology corporations and neural network developers in 2026 hold a different view. Proponents of "Strong AI" argue that the exponential growth of computing power and the emergence of multimodal models already allow systems to simulate tacit knowledge with accuracy indistinguishable from human. They believe that the problem of "tacit knowledge" is merely a temporary stage that will be overcome through training on massive datasets of human behavior. However, critics, including Denning, point out that simulating understanding is not equal to understanding itself, and this difference could become fatal in critical situations.