摘要: 摘要:[目的/意义]AI训练数据真实性是当前国内外法规政策和标准中的关注焦点,研究其概念体系和合规要求有助于提升AI技术的质量、合法性、透明度和社会接受度等。[方法/过程]使用内容分析法对目标文献中的AI训练数据真实性相关内容进行分析和抽取,使用ISO 704:2022概念分析方法从抽取的条目中识别AI训练数据真实性相关概念关系和特征,以构建AI训练数据真实性概念体系,基于概念体系构建分析框架,进一步识别对应的合规要求。[结果/结论]与AI训练数据真实性相关的概念有AI训练数据客观性、AI训练数据可靠性、AI训练数据确实性、AI训练数据完整性。AI训练数据真实性合规要求包括具有合法来源、与基本事实相符、数据准确、没有受到不当操作和能反映人工智能系统的背景或预期用途5类。
Abstract: Abstract: [Purpose/Significance] The authenticity of AI training data has become a key focus in regulations, policies, and standards both domestically and internationally. Researching the conceptual framework and compliance requirements for AI training data authenticity can help enhance the quality, legality, transparency, and social acceptance of AI technologies. [Method/Process] Content analysis was used to analyze and extract relevant information about AI training data authenticity from target literature. The concept analysis method, ISO 704:2022, was applied to identify concept relationships and characteristics related to AI training data authenticity from the extracted entries. A conceptual framework of AI training data authenticity was constructed, which was then used to develop an analysis framework for identifying corresponding compliance requirements.[Results/Conclusion] Key concepts related to AI training data authenticity include objectivity, reliability, credibility, and integrity of AI training data. AI training data authenticity compliance requirements include five categories: having a legitimate source, consistent with basic facts, accurate data, no improper manipulation, and reflecting the context or intended use of the AI system.
[V1] | 2024-10-12 10:54:57 | PSSXiv:202410.00829V1 | 下载全文 |
1. 农村基层情感治理视阀下村级档案管理的逻辑、困境与策略 | 2025-04-24 |
2. 信息哲学研究的四条进路 | 2025-04-24 |
3. 后摩尔时代半导体技术的演进路径与颠覆性技术预测 | 2025-04-24 |
4. 我国数据产权相关政策文本分析:演化脉络、主体协同与结构特征 | 2025-04-24 |
5. 智能体赋能科研知识服务的路径解析 | 2025-04-24 |