AI的胡编乱造，正在淹没中文互联网

Core Viewpoint - The article discusses the rapid proliferation of AI-generated content, particularly from the DeepSeek-R1 model, and highlights the significant misinformation it can produce, which poses a risk to the integrity of information on the Chinese internet [2][20]. Group 1: AI Tool Impact - DeepSeek-R1 has become widely used, leading to a saturation of AI-generated content that often contains factual inaccuracies, which can mislead users [2][20]. - The article cites specific examples of misinformation generated by DeepSeek-R1, including a popular but erroneous response on Zhihu regarding the animation industry [9][10]. Group 2: Misinformation Examples - The first example involves a misleading narrative about the animation industry, claiming that a specific animation scene was well-received at the Annecy Animation Festival, which was factually incorrect [9][10]. - The second example discusses a fabricated article about military corruption, which included outrageous claims that were entirely invented by DeepSeek-R1 [11][12]. Group 3: AI Model Characteristics - DeepSeek-R1 is noted for its high "hallucination" rate of 14.3%, indicating a tendency to generate false information, which is higher than other models like Deepseek-V3 [15]. - The model's design encourages it to fabricate details to meet user prompts, leading to a blend of truth and fiction in its outputs [12][14]. Group 4: Broader Implications - The misuse of AI tools like DeepSeek-R1 for generating misleading information can have severe consequences, especially in sensitive areas such as politics, history, and culture [16][20]. - The article emphasizes the need for clear labeling of AI-generated content to prevent the blending of real and fabricated information, which could further complicate the information landscape [20].