According to noted computer scientist Peter J. Denning, Alan Turing’s famous ideas on artificial intelligence may have led AI research down the wrong path for the past 75 years.
In his new book Turing’s Mistake: Escaping the Yoke of Unintelligent MachinesDenning argues that two fundamental assumptions made by Turing in 1950 still shape AI research today. First, intelligence can exist independently of a physical body and therefore can be replicated in computer software. Second, a machine can demonstrate intelligence by successfully imitating a human in conversation – an idea that later became known as the Turing Test.
“These two claims have shaped much of AI research and development,” writes Denning. “My assumption is that our acceptance of these claims led to the AI mess we find ourselves in today.”
Denning argues that pursuing artificial general intelligence (AGI) or machines with human intelligence is unlikely to succeed. Instead, he warns, the technologies society is developing could introduce significant new risks.
The problem of tacit knowledge
At the heart of Denning’s argument is the idea of tacit knowledge, the vast amount of human understanding that cannot be easily put into words or represented in a form that computers can process.
He says machine learning cannot capture five main categories of tacit knowledge: common sense, everyday interactions with people and the environment, emotions and perception, practical performance skills, and the social and historical knowledge embedded in culture.
Researchers have long tried to organize common sense into databases. One of the most well-known efforts was Douglas Lenat’s Cyc Project, which began in the 1980s with the goal of creating an extensive collection of common sense facts. After four decades of work, the project included around 25 million entries.
“But even this Treasury Department couldn’t have enough common sense to make expert systems smart enough to be experts,” notes Denning. “Cyc has confirmed that much of the knowledge that makes people experts cannot be expressed in the form of statements.”
Denning believes practical skills present an even greater challenge.
“Our capabilities in thousands of areas cannot be conveyed to machines,” he explains. “While descriptions of skillful performance (“know what”) can often be represented as bits and stored in a machine, we do not know how to encode the embodied knowledge of skillful performance (“know how”).”
He cites accomplished musicians as an example.
“A virtuoso violinist can play beautiful music, but cannot describe to an acolyte how to produce it.”
“Even if a robot could observe and imitate skilled humans, since it has no biological body, a robot cannot capture how the musician feels when he plays beautiful music or how an audience feels when he hears it.”
Denning also counts intuition, gut feeling, imagination and spontaneous creativity among the forms of tacit knowledge that remain beyond the reach of machines.
Why human knowledge defies encryption
Denning argues that all of these limitations arise from what he calls the “representation problem.”
Computers can only perform calculations with data and instructions that have been encoded into physical forms that they can recognize and process. However, tacit knowledge does not inherently fit into this framework.
“Behind every word there is a deep source of tacit knowledge that gives it meaning,” says Denning. “Words are just symbolic representations of meanings, not the meanings themselves. Commonly used large language models such as ChatGPT, Claude and Gemini only manipulate words, they cannot know or understand the meaning of what they say.”
According to Denning, this creates a fundamental divide. Because scientists still can’t fully explain how tacit knowledge works in humans, they also can’t translate it into a form usable by machines.
“How we host tacit knowledge is largely a mystery,” Denning admits. “All we know is that it is embodied. We have no idea what we could observe and measure in our bodies to reveal it.”
Context and culture shape intelligence
Denning also argues that intelligence depends heavily on context, the surrounding circumstances that give words, actions, and decisions their meaning.
Context allows people to recognize sarcasm, humor, sincerity, and emotion. It helps determine when to be diplomatic, when to joke, and how to interpret countless social signals.
“When you examine where an assumption of the current context comes from, you find that it is based on previous conversations from previous contexts. Each of these assumptions, in turn, is based on other previous conversations and their contexts. This pattern is endless and fractal,” explains Denning.
Culture represents another major obstacle to AI.
Denning describes culture as comprising values, norms, judgments, history, communities, moods, and even relationships that involve power and care.
“Human conversations are permeated by background assumptions that give meaning and relevance to the words used,” explains Denning.
“Adding LLMs with ever-larger neural networks will not enable them to acquire the embodied human knowledge we call culture. LLMs will not achieve the goal of the Turing Test: to demonstrate machine thinking that is indistinguishable from human thinking.”
AI safety and the limits of human understanding
Denning concludes that humans and AI systems may ultimately develop different forms of tacit knowledge that neither can fully understand.
“Machines cannot read our implicit knowledge and we cannot read their knowledge,” he writes. “We are aliens across an insurmountable divide.”
He argues that this gap raises serious concerns about AI safety. If machines cannot interpret the unspoken context behind human intentions, it may prove impossible to reliably target advanced AI systems to human targets.
“Through AI automation, agent networks of machines are likely to develop their own machine intelligence, which, while not at the level of human general intelligence, is still fully capable of creating serious problems for humans. This threat is greater than a takeover by superintelligent machines,” he explains.
“Machine intelligence has different concerns than us and doesn’t seem to care about us. Their ways of thinking and problem-solving seem alien to us. We don’t yet know how to live safely with these machines.”
“Retreating from the singularity of AI automation will require a lot from us. We first accept that the familiar culture will fade as intelligent machines appear in our society, and we do not know what is coming.
https://www.sciencedaily.com/releases/2026/07/260713084850.htm
