技术内容平台The Consensus开发了一项AI特征检测工具，用于识别可能由人工智能生成的文章1。该检测器采用静态规则而非机器学习模型构建，通过比较2023年前后的写作特征来判断文章特性1。平台将在99百分位数的AI特征文章从日常推荐、社交媒体和周刊中过滤，但用户仍可通过"sparkle filter"或在搜索中设置"aitells<90"来手动调整过滤标准1。
平台创始人指出，读者普遍更加偏好人类撰写的内容1。根据调查，人们压倒性地反感AI撰写的文章，即使质量较低的人类写作也优于高质量的AI生成内容1。值得注意的是，该检测器可能出现误判情况，编辑本人的文章也曾被系统标记为具有AI特征1。平台未完全禁止具有AI特征的文章，而是限制对这类内容的主动推广范围1。
The Consensus, a technology content platform, has launched an artificial intelligence detection system designed to identify articles likely generated by AI and limit their distribution through algorithmic recommendations 1. The detector operates using static rules established by comparing writing patterns before and after 2023, rather than employing machine learning models 1.
Articles scoring at the 99th percentile for AI characteristics will be filtered from the platform's daily recommendations and social media promotion, though users retain the ability to adjust these filtering thresholds 1. The platform has not outright banned AI-generated content but instead restricts its active promotion 1. Users can modify filtering preferences through a "sparkle filter" option or by setting "aitells<90" in search parameters 1.
The decision reflects broader user sentiment about AI-authored content. Research by Cynthia Dunlop found that people overwhelmingly prefer human-written articles over AI-generated ones, even when the human-written pieces are of lower quality 1. The detector is not without limitations—it has mistakenly flagged editorial content as AI-generated in some cases 1.
评论
还没有评论，欢迎留下第一条。