Alan Turing’s foundational assumptions about artificial intelligence may have led the field astray for decades, argues prominent computer scientist Peter J. Denning in a new book. In "Turing's Mistake: Escaping the Yoke of Unintelligent Machines," Denning challenges two ideas from Turing’s 1950 paper: that intelligence can exist without a physical body and that a machine that convincingly imitates a human in conversation has demonstrated real intelligence. Denning says these twin premises have fueled a frenzy around large language models and artificial general intelligence that is unlikely to pay off.

The core problem, Denning writes, is what he calls "tacit knowledge" — the vast, unspoken understanding humans rely on every day. Common sense, intuition, cultural norms, emotional perception, and practical skills like playing a violin cannot be encoded into data or algorithms. He points to the Cyc project, a four-decade effort to catalog common sense facts, which amassed 25 million entries but still failed to make expert systems truly smart. "We do not know how to encode the embodied knowledge for skillful performance," Denning says. Even the most advanced LLMs manipulate words without grasping meaning. "Behind every word is a deep well of tacit knowledge that gives it meaning," he notes.

Denning warns that the gap between human and machine intelligence is not closing — it may be unbridgeable. As AI systems develop their own opaque forms of tacit knowledge, they could become dangerously misaligned with human intentions. "Machine intelligence has different concerns from us and does not appear to care about us," he writes. Rather than chasing AGI, Denning urges society to recognize the limits of computation and reassert what makes humans distinct: embodied experience, culture, and the ability to understand context. The real risk, he argues, is not superintelligent overlords but low-intelligence machines that cause severe problems because they cannot read the room.

Source: Science Daily