Where can the dust alight何处惹尘埃 "Research on the Associative Functions of AI"AI 联想功能研究

Explainable AI is vital as certain key algorithms act as "black boxes." Our project highlights this issue through a design exploring Chinese characters’ visual association mechanisms, revealing AI’s unpredictable behaviors. Association is a conscious human activity, but over-association can lead to "delusion." While AI performs associations efficiently, it often generates unexpected outcomes like "loss of focus." We designed two approaches: (1) a dictionary-based method with predefined, stable algorithms, and (2) an autonomous method where AI independently generates associations, producing novel yet unpredictable results. We presented these methods through a webpage and an interactive installation resembling a mirror. Users input Chinese characters via the webpage, and the mirror displays AI’s associative process in real time. This immersive experience reveals AI’s "non-human" logic and emergent patterns. Our project doesn’t manage AI’s unpredictability but fosters understanding. By engaging users with AI’s processes, the installation prompts reflection on its limitations, unique logic, and implications for its role in society, encouraging us to reconsider expectations of AI-driven logic and meaning. 可解释的人工智能至关重要，因为某些关键算法就像 "黑盒子 "一样。我们的项目通过探索汉字的视觉联想机制，揭示了人工智能不可预测的行为，从而突出了这一问题。联想是人类有意识的活动，但过度联想会导致 "错觉"。人工智能在高效完成联想的同时，往往会产生 "失焦 "等意想不到的结果。我们设计了两种方法：(1) 一种基于词典的方法，采用预定义的稳定算法；(2) 一种自主方法，由人工智能独立产生联想，产生新颖但不可预测的结果。我们通过一个网页和一个类似镜子的互动装置来展示这些方法。用户通过网页输入汉字，镜子则实时显示人工智能的联想过程。这种身临其境的体验揭示了人工智能的 "非人类 "逻辑和突发模式。我们的项目不是管理人工智能的不可预测性，而是促进理解。通过让用户参与人工智能的过程，该装置促使人们反思人工智能的局限性、独特逻辑及其在社会中的作用，鼓励我们重新考虑对人工智能驱动的逻辑和意义的期望。