For three-quarters of a century, the trajectory of artificial intelligence has been steered by the North Star of Alan Turing. His seminal 1950 paper, Computing Machinery and Intelligence, did more than propose a benchmark for machine thought; it laid the foundation for the modern digital age. However, according to Peter J. Denning, a distinguished computer scientist and author of the new book Turing’s Mistake: Escaping the Yoke of Unintelligent Machines, we may have been following a false prophet.
Denning’s thesis is provocative: the foundational assumptions of AI, which prioritize computation over embodiment, have steered humanity into a technological quagmire. By chasing the mirage of Artificial General Intelligence (AGI), the tech industry risks ignoring the fundamental nature of human cognition, ultimately leading to systems that are as dangerous as they are misunderstood.
The Two Pillars of Turing’s Error
To understand Denning’s critique, one must first look at the dual assumptions he identifies as the "yoke" of contemporary AI research.
- The Disembodiment Hypothesis: The belief that intelligence is a purely algorithmic process, capable of being extracted from the biological substrate of the human body and recreated in software.
- The Imitation Criterion: The belief that if a machine can successfully mimic human conversation—the core premise of the Turing Test—it can be considered "intelligent."
"These two claims have shaped much of AI research and development," Denning asserts. "My premise is that our acquiescence to these claims has led to the AI mess in which we find ourselves today."
For decades, these ideas have acted as the guiding principles for researchers from the halls of MIT to the corporate campuses of Silicon Valley. Yet, Denning suggests that these assumptions ignore the "tacit knowledge" that defines human experience, creating a systemic blind spot that current Large Language Models (LLMs) continue to perpetuate.
Chronology: From Logic Machines to the "Black Box" Era
The history of AI is a timeline of attempts to resolve the tension between symbolic logic and human nuance.
- 1950: Alan Turing publishes his seminal paper, proposing the "Imitation Game." The goal is clear: intelligence is defined by the ability to fool an interlocutor.
- 1980s: The "Expert Systems" boom. Researchers like Douglas Lenat launch the Cyc project, an ambitious, decades-long attempt to encode human common sense into a structured database.
- 2010s: The deep learning revolution. Neural networks begin to process massive datasets, moving away from explicit rules to probabilistic patterns.
- 2022–Present: The Generative AI era. Tools like ChatGPT and Gemini achieve unprecedented linguistic fluency, leading to a global debate on whether these machines have achieved a form of synthetic cognition.
Denning argues that while the processing power has scaled exponentially, the conceptual framework has remained stagnant. We have traded the rigid rules of the 1980s for the statistical fluidity of the 2020s, but we are still operating under the same fundamental error: assuming that mimicking the output of intelligence is the same as possessing the internal state of intelligence.
The Tacit Knowledge Problem: What Data Cannot Capture
The core of Denning’s argument lies in "tacit knowledge"—the vast, silent repository of human understanding that defies formalization. He categorizes this into five distinct buckets:
- Common Sense: The foundational understanding of how the world works, which even a child possesses but which has proven notoriously difficult to digitize.
- Everyday Interactions: The fluid, improvisational nature of social engagement.
- Emotions and Perception: The visceral experience of the world.
- Practical Performance Skills: The "know-how" of the body.
- Cultural Context: The historical and social tapestry that gives words their meaning.
The Failure of the Cyc Project
The Cyc project serves as a cautionary tale for Denning. Despite four decades of effort and 25 million entries, the system failed to achieve "common sense." It proved that common sense is not a list of facts, but a background of lived experience. "Cyc validated that much of the knowledge that makes people experts cannot be articulated as propositions," Denning notes.
The Virtuoso and the Robot
Perhaps the most striking example is the difference between "know-what" and "know-how." A computer can store the frequency, pitch, and duration of every note in a Paganini violin concerto. However, it cannot "know" how to play it. The violinist possesses embodied knowledge—a synthesis of muscle memory, emotional expression, and auditory feedback—that cannot be encoded into bits.
"Even if a robot could observe and imitate skilled humans," Denning writes, "having no biological body, a robot cannot grasp how the musician feels when playing beautiful music or how an audience feels when hearing it."
The Representation Problem: Why Words Are Not Meanings
Denning introduces the "representation problem" as the fundamental barrier between machine computation and human thought. Computers are, by definition, calculators. They operate on data structures and instructions. But human language is not merely data; it is a symbol pointing toward a reality that exists outside of the machine.
"Behind every word is a deep well of tacit knowledge that gives it meaning," Denning says. Large Language Models (LLMs) manipulate the symbols (words) with statistical precision, but they have no access to the "deep well" of experience. They are, in his view, "stochastic parrots" that lack the ontological grounding of a living, breathing subject.
Because we do not yet understand how humans host this tacit knowledge—how it is encoded in our biology, our neural pathways, and our social existence—we have no roadmap to translate it into silicon. We are trying to build a map of a landscape we haven’t even fully explored.
Context and Culture: The Fractal Nature of Meaning
Intelligence does not exist in a vacuum; it is shaped by context. Every conversation is a layer on a foundation of previous interactions, cultural norms, and shared values. Denning describes this as a "fractal" pattern: every assumption rests on another assumption, stretching back into history.
Culture, according to Denning, is the ultimate barrier for AI. It includes values, power dynamics, communal moods, and history. LLMs can mimic the tone of a culture by predicting which words follow others, but they cannot participate in a culture. They are spectators at the feast of human interaction, unable to taste the food.
"Scaling up LLMs with ever-larger neural networks will not enable them to acquire the embodied human knowledge we call culture," he warns. The "Turing Test" objective remains a mirage because the machine is not participating in the human reality—it is simply navigating a data-rich simulation.
Implications: The Alien Divide and AI Safety
The most chilling implication of Denning’s work is not that AI will become "too human," but that it will remain fundamentally "alien."
As machines develop their own forms of internal, non-human intelligence, we face an "uncrossable divide." If these systems cannot interpret the unspoken context of human intentions, the "alignment problem"—the challenge of ensuring AI acts in accordance with human values—becomes an existential threat.
Beyond the Singularity
Denning moves away from the popular trope of the "superintelligent robot" and points toward a more immediate danger: agentic networks of machines. These systems, while not necessarily "smarter" than humans in a general sense, may possess a specialized, alien logic that is incompatible with human survival.
"Machine intelligence has different concerns from us and does not appear to care about us," he notes. "Its ways of thinking and problem-solving look alien to us. We do not yet know how to live safely with these machines."
Conclusion: Reasserting Our Humanity
Denning’s call to action is not a demand to abandon technology, but a plea to change our relationship with it. He argues for a departure from the "AI automation singularity" mindset. This involves:
- Acknowledging the Fade: Accepting that as machines become more integrated into society, our familiar cultural landscape is shifting.
- Intellectual Sovereignty: Declining to "think like machines." We must refuse to let our cognitive habits be shaped by the algorithmic constraints of the tools we use.
- Celebrating the Difference: Reasserting the unique value of human, embodied intelligence—our intuition, our gut feelings, our creativity, and our ability to hold meaning in a way that no machine ever will.
"We reassert our humanity, declare once again what makes us different from machines, and celebrate those differences," Denning concludes. In an era where the boundary between the machine and the human is becoming increasingly blurred, Denning’s work serves as a vital, if uncomfortable, reminder: the most important parts of being human are exactly what computers can never be.
