《滕王阁序》是AI生成的？近日，AI检测工具屡屡爆出"翻车"的新闻，如《荷塘月色》被标 注"62.88%AI率"，《三体》片段被标红警示，引发公众对AI检测工具科学性的热议。 为探究AI检测工具的识别能力与技术原理，南方都市报、南都大数据研究院选取了国内10款热门的文 本、图片AIGC检测工具进行了抽样测评。结果显示：文本检测工具中，检测标准参差不齐，明显误 判、漏检、乱检的情况均有存在。而图片检测工具中，均对PS后的摄影图片难以识别。 当AI检测被用于高校毕业的"门槛"、期刊评审的"硬指标"，一度引发了困惑质疑。在专家看来，当前AI 检测技术尚处于探索阶段，误判或是技术演进的必经过程，强行将不稳定的技术跟学术诚信关联不可 取。但长远来看，技术迭代与合规框架构建仍需"双轨并行"。 AI文本检测 存误判、漏检、乱检"难题" AI检测工具靠谱吗？南都大数据研究院本次测评样本为知网、PaperPass、万方、维普、朱雀大模型检 测、挖错网、大雅、PaperYY、团象、茅茅虫共10款国内热门的文本、图片AIGC检测工具。 首先是文本类检测，尝试使用四类文章来测试对真实文章、以及不同程度AI生成内容的识别率。四篇 文章 ...

Group 1 - The core viewpoint of the articles revolves around the challenges and implications of using "AI rate" as a criterion for evaluating academic papers, highlighting the potential for misjudgment and the limitations of AI detection systems [1][2][3] - Several universities have implemented regulations linking the proportion of AI-generated content in graduation theses to their acceptance, aiming to prevent academic misconduct [1] - The detection systems analyze text features such as vocabulary frequency, sentence structure, and logical expression to determine similarity with AI-generated content, but this approach is flawed due to the inherent similarities between AI-generated and human-written academic texts [1][2] Group 2 - Controversies surrounding "AI rate" detection reflect the broader challenges faced by education in the era of technological transformation, where reliance on AI detection may lead to superficial modifications in original works [2] - Academic committees are deemed the ultimate authority in evaluating the authenticity of student work, as educators possess a deeper understanding of students' capabilities and research processes [2] - The focus should be on fostering independent thinking and innovative ideas in students rather than merely producing texts that pass AI detection [3]