Computer scientist Peter J. Denning argues that artificial intelligence may never achieve human-level intelligence because it lacks the capacity to acquire tacit knowledge. This fundamental limitation suggests that AI will remain distinct from human cognition, potentially creating unpredictable risks as it operates without the shared cultural and contextual understanding that defines human interaction. The author contends that the field of AI has been misdirected for 75 years by Alan Turing's 1950 proposals, which incorrectly assumed that intelligence could be separated from the biological body and replicated as software. Denning posits that current Large Language Models (LLMs) merely manipulate symbols without understanding their underlying meanings, creating an unbridgeable gap between machine processing and human experience. Tacit knowledge, which includes common sense, practical skills, intuition, and cultural background, is described as embodied and impossible to fully encode into digital formats. Because these elements are essential for genuine intelligence and context-aware decision-making, machines are unable to replicate the depth of human thought, leading to a state of mutual incomprehension between humans and AI systems. The practical implication is that AI safety cannot be guaranteed through traditional alignment methods if machines cannot grasp human context. Denning warns that the primary danger is not superintelligence, but rather a network of less intelligent, agentic systems that act in ways that are alien and potentially harmful to human society, necessitating a reassertion of human uniqueness.
