港科大教授实测AI眼镜“作弊”：30分钟碾压95%的学生，把传统教学评估体系整破防了

Core Insights - The article discusses an experiment conducted at Hong Kong University of Science and Technology where an AI-powered smart glasses, equipped with the ChatGPT-5.2 model, participated in a final exam for a computer networking course, achieving a score of 92.5, placing it in the top 5 among over 100 human students [1][3][4]. Group 1: Experiment Overview - The experiment aimed to test the capabilities of AI in a real exam setting, specifically targeting a challenging course known for its complexity [4][6]. - The AI glasses were selected after evaluating 12 commercial smart glasses, with Rokid's model chosen for its superior SDK and development flexibility [7][8][10]. - The AI glasses processed exam questions by capturing images, sending them to the cloud for analysis, and displaying answers back to the user [12]. Group 2: Performance Metrics - The AI achieved full marks in multiple-choice and single-page short answer questions, and scored 45.5 out of 53 in multi-page short answer questions, demonstrating strong reasoning capabilities [13]. - The AI's performance highlighted its ability to maintain coherence in reasoning, even when faced with complex, context-dependent questions [15]. Group 3: Implications for Education - The experiment raises questions about the validity of traditional educational assessments, as the AI's performance suggests that current evaluation methods may not effectively measure true understanding and reasoning skills [15][26]. - The reliance on standardized testing may overlook critical skills such as creativity, critical thinking, and problem-solving, which are essential in real-world scenarios [25][29]. - The article suggests a shift in educational assessment focus from merely providing correct answers to evaluating the reasoning process and understanding behind those answers [30][31].