《欧盟委员会：2024生成式人工智能透明度：机器生成内容的识别研究报告（英文版）（6页）.pdf》由会员分享，可在线阅读，更多相关《欧盟委员会：2024生成式人工智能透明度：机器生成内容的识别研究报告（英文版）（6页）.pdf（6页珍藏版）》请在三个皮匠报告上搜索。

1、 Generative AI Transparency:Identification of Machine-Generated content HIGHLIGHTS Generative AI is a cutting-edge field of AI that can create realistic human-like con-tent.Concerns over its misuse and societal implications highlight the importance of identifying machine-generated content.Solutions

2、to identify AI generated content should possess the following four proper-ties:efficiency,integrity of data,robustness to content alteration,and protection against manipulation.Current technical solutions based on metadata,watermarking,fingerprinting or detection do not fulfil these requirements suf

3、ficiently for text,audio,image and video content.INTRODUCTION Generative AI(GenAI)1 is a cutting-edge field of artificial intelligence(AI)that has recently gathered considerable attention with the recent advancements of state-of-the-art techniques and the emergence of consumer-facing products such a

4、s ChatGPT,Midjourney or Sora.GenAI refers to machine learning models used to generate media content such as text,audio,image or video that mirrors human-made content.The potential of applications spans across various industries.The creativity and reasoning abilities of this technology may affect all

5、 intellectual and artistic professions.Despite the benefits that GenAI offers across various domains,its growing capabilities have sparked concerns about unique AI-specific risks to safety and fundamental rights.GenAI has the potential to support misinformation campaigns and amplify opinion manipula

6、tion 2,and to increase the efficiency of fraud by making plagiarism or impersonation more difficult to detect and more efficient 3.These risks can impact significantly democratic processes 4.GenAI also blurs the line between human-and machine-created content,triggering new paradigms for jobs,creativ