Recently, an experiment led by Professor Zhang Jun and Professor Meng Zili from the Hong Kong University of Science and Technology has garnered significant attention. In this experiment, a pair of Leqi Rokid AI glasses, powered by the ChatGPT - 5.2 model, took on a final - exam paper for the course 'Principles of Computer Networks'. Remarkably, it completed the paper within 30 minutes and scored an impressive 92.5, outperforming 95% of the human test - takers.

The experiment meticulously replicated a real - life exam setting. The AI glasses utilized a built - in camera to capture the exam questions. These questions were then transmitted to a large - scale language model. Within a mere 0.8 seconds, the model generated answers through logical reasoning. Subsequently, the answers were displayed on the lenses of the glasses, allowing for easy copying.

This experiment has brought to light a fundamental conflict within traditional teaching assessment methods. Written exams that primarily focus on rote memorization of knowledge points and standard step - by - step derivations are precisely the areas where AI demonstrates its greatest strengths. In light of this impact, universities have started to explore various reform strategies. These include introducing AI - based oral exam systems, incorporating assignments that require presentations and on - site defenses, and adopting 'process portfolios' as part of the assessment process.