By Siti Muinah 
For over seven decades, the trajectory of artificial intelligence research has been guided by a singular, foundational vision: that human intelligence is essentially a computational process, capable of being detached from the biological body and replicated in code. However, one of the field’s most respected voices is now calling for a radical re-evaluation of this consensus.
In his provocative new book, Turing’s Mistake: Escaping the Yoke of Unintelligent Machines, renowned computer scientist Peter J. Denning argues that the very pillars upon which modern AI is built are fundamentally flawed. According to Denning, the "AI mess" we find ourselves in today—characterized by massive, opaque models and growing concerns over existential risk—is the direct result of adhering to two faulty assumptions laid down by Alan Turing in 1950.
The Foundations of a False Promise
To understand Denning’s critique, one must look back to the birth of the field. In his seminal 1950 paper, "Computing Machinery and Intelligence," Alan Turing proposed two ideas that would become the North Star for generations of researchers.
The first was the "disembodiment" thesis: the belief that intelligence is a software-like property that can exist independently of physical form. The second was the "Turing Test," a functionalist standard suggesting that if a machine can imitate human conversation well enough to fool a human judge, it should be considered intelligent.
"These two claims have shaped much of AI research and development," Denning writes. "My premise is that our acquiescence to these claims has led to the AI mess in which we find ourselves today."
Denning argues that by chasing the phantom of Artificial General Intelligence (AGI)—the quest to create a machine with human-level cognitive parity—researchers have ignored the fundamental ways in which human knowledge is "embodied" and contextual. By focusing on imitation rather than genuine understanding, we have built systems that are brilliant at syntactic manipulation but utterly hollow when it comes to semantic meaning.
A Chronology of the Computational Dream
The history of AI is a timeline of increasingly sophisticated attempts to bridge the gap between symbols and reality:
- 1950: Alan Turing publishes his seminal paper, shifting the focus of AI from "thinking" to "imitating."
- 1956: The Dartmouth Workshop officially coins the term "Artificial Intelligence," cementing the computationalist paradigm.
- 1960s–1970s: The era of "Good Old Fashioned AI" (GOFAI), focusing on logic-based systems and symbolic reasoning.
- 1984: Douglas Lenat launches the Cyc project, an ambitious, multi-decade attempt to encode human "common sense" into a database.
- 2010s: The "Deep Learning" revolution begins. Instead of manually encoding rules, researchers shift to neural networks that learn patterns from massive datasets.
- 2022–Present: The generative AI boom, led by Large Language Models (LLMs) like ChatGPT, Claude, and Gemini, brings the Turing Test into the mainstream, effectively "passing" it in many eyes—yet confirming Denning’s fears about the lack of true understanding.
The Tacit Knowledge Problem: Why Data Is Not Wisdom
At the core of Denning’s critique is the concept of "tacit knowledge"—the vast, intuitive reservoir of human understanding that defies formalization. Denning posits that machine learning, no matter how much data it consumes, remains fundamentally incapable of capturing five essential categories: common sense, everyday social interactions, emotional and perceptual intelligence, practical performance skills, and cultural context.
The Cyc Failure and the Limits of Logic
The Cyc project stands as a cautionary tale in this regard. Spanning four decades and accumulating roughly 25 million entries, Cyc was intended to be the "encyclopedia of common sense." Yet, as Denning observes, the project reached a plateau. "Cyc validated that much of the knowledge that makes people experts cannot be articulated as propositions," he notes. The failure was not a lack of data, but a misunderstanding of what "common sense" actually is: it is not a list of facts, but an active, lived engagement with the world.
The "Know-How" Divide
Denning emphasizes that while we can digitize "know-what"—the descriptive outcomes of a task—we remain ignorant of how to encode "know-how," the embodied skill required to perform it.
Consider the virtuoso violinist. A machine can analyze the frequencies, tempo, and note sequences of a Paganini caprice. It can even mimic the performance. However, it cannot grasp the muscular tension, the emotional resonance, or the subtle, spontaneous adjustments a human makes in response to the acoustics of a room or the mood of an audience. Without a biological body, the robot remains an observer, not a participant in the human experience.
The Representation Problem: Symbolic Shadows
Why does this knowledge resist encoding? Denning points to the "representation problem." Computers operate on discrete, encoded data. Words are symbolic representations—pointers—but they are not the meanings themselves.
"Behind every word is a deep well of tacit knowledge that gives it meaning," Denning says. When a model like GPT-4 generates a sentence about "love" or "grief," it is not accessing a lived history of human emotion; it is performing a high-dimensional statistical prediction based on the proximity of tokens. It does not know what it is saying, and more importantly, it cannot care.
This creates a fundamental divide. Because we cannot fully explain the biological and neural mechanisms that host human tacit knowledge, we have no roadmap for how to transfer that knowledge into a silicon architecture. We are essentially building machines that simulate the surface of human communication while ignoring the depth of the human condition.
The Fractal Nature of Context and Culture
Denning’s argument extends to the role of context. Human communication is not a series of isolated exchanges; it is a nested, fractal structure where every sentence relies on a history of prior conversations, shared cultural norms, and unspoken assumptions.
"When you inquire into where an assumption of the current context came from, you discover it rests on previous conversations from previous contexts," he explains.
Culture, for Denning, is the operating system of human intelligence. It includes the power dynamics, the shared moods, the historical burdens, and the subtle judgments that make communication meaningful. Because LLMs are trained on static datasets, they lack the ability to participate in the evolving, living flow of culture. They are "frozen" in the data of the past, unable to navigate the shifting currents of present human values.
Implications: The Alien in Our Midst
The most sobering aspect of Denning’s analysis is the warning regarding AI safety. If machines are fundamentally different from humans, we are heading toward a future where our AI systems and our human societies operate on parallel, incompatible tracks.
"We are aliens across an uncrossable divide," Denning writes.
The Risk of Agentic Networks
The danger, according to Denning, is not necessarily the Hollywood-style "superintelligence" that turns against humanity. Rather, it is the emergence of "agentic networks"—systems that act with a logic we cannot audit, driven by goals that do not align with human well-being. These machines may develop their own, alien forms of "intelligence" that are effective at problem-solving but entirely indifferent to human suffering or values.
A Call for Re-humanization
Denning concludes with a plea for a shift in perspective. He argues that the solution is not to double down on the pursuit of AGI, but to recognize the unique, non-computational nature of human intelligence.
"We decline to think like machines or be subservient to machines," he writes. "We refuse to submit to a yoke imposed by low-intelligence machines. Most importantly, we reassert our humanity, declare once again what makes us different from machines, and celebrate those differences."
As AI becomes increasingly integrated into our daily lives, Denning’s critique serves as a necessary intervention. It challenges us to stop viewing the human mind as a machine to be replicated and start viewing it as a biological, social, and cultural miracle that cannot—and perhaps should not—be reduced to code. The future of technology, he suggests, depends not on how well we can make machines think, but on how clearly we can define the boundary where our own humanity begins.
Sagoh 
Asro 