Scholarlink

<p>This paper revisits McDermott’s seminal 1976 critique, Artificial Intelligence Meets Natural Stupidity, in light of the rapid rise of generative AI. McDermott’s warnings about conceptual imprecision, rhetorical inflation, and anthropomorphic metaphors remain strikingly relevant today, particularly in domains where linguistic fluency is mistaken for genuine understanding. Using computational and urban science as a case study, the paper highlights the limitations of large language models (LLMs) in spatial reasoning, including their hallucination of map features, distortion of topological relationships, and failure to grasp metric consistency. These shortcomings exemplify the epistemological risks McDermott identified nearly five decades ago. The paper argues for a renewed commitment to epistemic discipline and humility in the development and communication of generative AI, especially in high-stakes domains like geospatial applications. It proposes concrete guidelines for integrating spatial theory, documenting failures, and avoiding misleading terminology, advocating for hybrid approaches that combine LLMs with GIS frameworks and human oversight. By operationalizing McDermott’s insights, the paper calls for transparent, grounded, and responsible AI systems that can genuinely support, rather than distort, urban decision-making.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Computational and urban science amid generative AI: the enduring relevance of McDermott’s critique (opinion paper)

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月20日

查看全文
  • Christophe Claramunt,
  • Lars De Sloover,
  • Nico Van de Weghe

摘要

This paper revisits McDermott’s seminal 1976 critique, Artificial Intelligence Meets Natural Stupidity, in light of the rapid rise of generative AI. McDermott’s warnings about conceptual imprecision, rhetorical inflation, and anthropomorphic metaphors remain strikingly relevant today, particularly in domains where linguistic fluency is mistaken for genuine understanding. Using computational and urban science as a case study, the paper highlights the limitations of large language models (LLMs) in spatial reasoning, including their hallucination of map features, distortion of topological relationships, and failure to grasp metric consistency. These shortcomings exemplify the epistemological risks McDermott identified nearly five decades ago. The paper argues for a renewed commitment to epistemic discipline and humility in the development and communication of generative AI, especially in high-stakes domains like geospatial applications. It proposes concrete guidelines for integrating spatial theory, documenting failures, and avoiding misleading terminology, advocating for hybrid approaches that combine LLMs with GIS frameworks and human oversight. By operationalizing McDermott’s insights, the paper calls for transparent, grounded, and responsible AI systems that can genuinely support, rather than distort, urban decision-making.