互联网上，越来越多的句子有一股「AI味」。风景永远惊艳，事件永远关键，语言被磨成光滑的模板。于是，Wikipedia在成千上万次巡查里 最先察觉，并悄悄发起一场清理行动。 现在互联网上的句子，好像都是一个师傅教的。 编辑们越看越觉得奇怪。等他们把这些可疑段落攒在一起，真相也浮出水面： 来源模糊、论据空洞、表达统一得异常。这一切都指向同一个可能——越来越多的内容，是AI生成的。 为了避免百科被这种空心句子掏空，一群志愿者在2023年发起了WikiProject AI Cleanup。 风景总是breathtaking，建筑永远modern，随便一个事件都「标志着一个关键时刻」。 于是，一本奇怪又实用的《识别AI文风手册》，就在无数次删删改改里被写出来。 这些话听着完美，却轻得像飘在空中。读多了，人会感到恍惚。 同一个模板写出来的互联网 Wikipedia的编辑们也受够了这种「广告腔」，开始了一场清理行动。 互联网的语言越来越整齐划一，Wikipedia最早察觉到「味道不对」。 他们校对大量新增内容后发现，最爱说场面话的，竟然不是人，而是AI。 每天都有成千上万的新内容被塞进条目里，编辑们一边巡查一边发现：不少新 ...

据悉，阜博集团AI音乐检测服务依托先进的音频、旋律与人声识别技术，通过提升检测准确率与真阳 性率，构建音乐内容生态的 "版权防火墙"。随着NanoBanana、Gemini 、Sora2等多模态模型的密集落 地，AI 音乐因算力成本更低、成品成熟度更高已出现可规模化变现的作品，行业对 AI 生成内容的识别 与确权需求持续激增，阜博集团此次布局正顺应这一核心趋势。 值得注意的是，该服务的快速落地得益于公司此前对音频内容识别技术提供商 PEX的收购，形成技术 协同优势，助力公司顺利跟上多模态音频端的关键发展节奏，为串流平台、音乐发行商及版权协会提供 关键支持。业绩层面，阜博集团已从多模态需求扩张中持续受益，25年三季度收入增长超27%，其中主 要来自Youtube平台的美国业务收入增长达33%。收入增长核心动力之一来自AI生成视频数量的快速攀 升。当前公司管理的活跃资产中已有9.1%为AI相关内容，增速远超整体活跃资产增长，对公司收入形 成实质性贡献。 智通财经APP获悉，11月21日，阜博集团(03738)宣布推出AI音乐检测服务。该服务面向数字平台、音 乐发行商及版权管理组织，旨在精准区分人类创作与 AI 生 ...

Investment Rating - The report maintains an "Outperform the Market" rating for the media industry [5][35][40]. Core Views - The media industry experienced a decline of 2.31%, underperforming both the CSI 300 (-0.66%) and the ChiNext Index (-1.68%) during the week of November 10 to November 16 [1][12]. - Key companies showing positive performance include Xiangyuan Cultural Tourism, Sanwei Communication, Xinhua Dou, and Tianxia Show, while companies like Yue Media, Kunlun Wanwei, Rongxin Culture, and Kaiying Network faced significant declines [1][12]. - The report emphasizes the importance of monitoring the economic bottom and potential policy shifts, particularly in AI applications and content creation [4][35]. Summary by Sections Industry Performance - The media sector ranked 23rd in terms of weekly performance among all sectors, with a notable decline of 2.31% [1][12][14]. - The top three films for the week generated a total box office of 581 million yuan, with "The Life of Langlang" leading at 338 million yuan, accounting for 58.2% of the total [3][19]. Key Developments - The launch of Marble 3D by World Labs allows for the creation of navigable 3D virtual worlds from various input formats, marking a significant advancement in AI-generated content [2][16]. - The introduction of the new "head and tail frame" feature in the Keling 2.5 Turbo model enhances the generation quality of AI videos, improving controllability and consistency [2][17]. - The open-source AI framework Kosong by Moonlight provides developers with efficient tools for building intelligent applications [2][17]. Investment Recommendations - The report suggests focusing on the gaming sector and IP trends, recommending companies like Giant Network, Kaiying Network, and Jibite for their potential in the current market cycle [4][35]. - It highlights the importance of content policy shifts and AI application opportunities, recommending platforms like Mango Super Media and Bilibili, as well as content producers like Light Media and Huace Film [4][35].

Core Insights - PolitiFact's evaluation of xAI's Grokipedia reveals serious issues including content plagiarism and lack of factual accuracy, damaging its credibility as a reliable information source [2][3][4] - Grokipedia, launched in October 2023, aimed to compete with Wikipedia but has faced widespread criticism for its AI-generated content that often replicates Wikipedia without proper attribution [2][3] Content Plagiarism - PolitiFact analyzed 885,279 articles on Grokipedia, finding that many directly copied text from Wikipedia, with a similarity rate as high as 96% in some cases [2][4] - An example includes the "Monday" entry, which omitted 22 reference sources from Wikipedia while claiming to adapt content under a Creative Commons license [4] Lack of Factual Accuracy - Grokipedia's AI-generated content has been found to introduce errors, such as a misleading statement regarding the Nobel Prize, which lacked citation and was factually incorrect [4][6] - The platform often simplifies narratives by removing context from Wikipedia, sacrificing accuracy in the process [4][5] Transparency Issues - Grokipedia's editing mechanism lacks transparency, with no visible edit history or clear error correction processes, making its "fact-checking" label difficult to trust [6] - In contrast, Wikipedia maintains a transparent editing process, allowing anyone to edit and providing a complete history of article changes [5][6] Industry Context - The launch of Grokipedia coincides with xAI's legal disputes with Apple and OpenAI, which have accused these companies of suppressing AI competition [6] - The Wikimedia Foundation has raised concerns about Grokipedia's plagiarism and the implications for educational ecosystems, emphasizing the need for responsible use of open-source content [6]

"AI假象"带不来流量 近日，关于外卖档口利用AI技术包装出"精品堂食店"的消息引发关注。有人特意去探店，发现实地场景 与宣传图片大相径庭。真相曝光后，相关商家受到广泛批评，很多消费者表示，再也不会点这家饭店的 餐食。可见，用AI技术营造"网红门面"的假象，既成不了网红，也带不来流量。 随着技术发展，AI生成内容日益逼真，如果被误用、滥用或恶意使用，会带来很大安全风险。有网友 吐槽"现在看张图还要琢磨真伪"，也有网友认为"恶意使用AI会成为虚假宣传帮凶"。这些声音传递出公 众对技术滥用的担忧。 在商业推广中，利用技术手段包装产品和服务无可厚非，但前提是合法合规。一是内容应真实、合法， 不得含有虚假、欺骗和误导消费者的信息，必须如实反映商品或服务的实际情况；二是商家应清楚、明 白地对商品或服务的质量、价格等作出详细标识。对于平台店铺使用AI生成内容，必须严格遵循《人 工智能生成合成内容标识办法》的规定，对AI生成的文本、音频、图片、视频等内容需添加显式和隐 式标识。这些标识还不能随意添加，其字号大小、节奏快慢、时长等均须符合国家强制性标准。 如果商业推广越过合规红线，滥用AI生成内容，就可能侵犯消费者知情权，甚 ...

当下，亟需进一步规范AI生成内容在商业活动中的使用。一方面，商家要诚信经营，切实满足消费者 的知情权，用技术提升产品或服务的质量。若使用AI生成内容装点门面，也应主动按规定作出标识， 保障消费者的知情权。另一方面，平台应加强监管，督促撤换不真实或不符合要求的图片，善用客户评 价机制，履行好"把关人"责任。有关部门需完善与技术发展相适应的监管方式，制定符合实际的监管规 则或指引，加强消费者权益保护，倡导诚信经营，共同营造良好商业环境。 下次，再看到平台上光鲜亮丽的"精品堂食店""网红门店"时，建议您不妨多留一个心眼，仔细辨别、理 性判断，别被"AI假象"轻易蒙蔽双眼。 （文章来源：经济日报） 如果商业推广越过合规红线，滥用AI生成内容，就可能侵犯消费者知情权，甚至构成消费欺诈。消费 者权益保护法规定，消费者享有知悉他们购买、使用商品或接受服务真实情况的权利。从某种程度上 说，商家使用AI生成内容如果与实际情况不符，即便一时吸引来了顾客，也难以长久发展。 技术应当向善，不能成为非法牟利的工具。近年来，关于如何规范人工智能技术应用，已有相关司法裁 判明晰规则、厘清边界。例如，在全国首例"AI陪伴案"中，北京互联网法 ...

"流浪汉闯家中"式AI生成带来的危害是多方面的，并且可能导致实质性的风险。从社会层面来看， 它严重浪费了公共安全资源。警方为了应对这些虚假报警，不得不投入大量的人力、物力进行调查和处 理，这无疑会影响到真正紧急警情的处理效率。从个人层面来说，对于那些受到惊吓的业主和家人，这 种心理上的冲击可能会持续很长时间，甚至影响到他们的日常生活和心理状态。对于参与这类行为的人 来说，他们可能没有意识到自己的行为已经触犯了法律和道德的底线，一旦造成严重后果，将面临法律 的制裁。 让人担心的是，类似的AI生成不仅容易引发诸多效仿者，而且形式也日益多样。在短视频平台 上，不仅有"流浪汉闯家中"的AI整蛊视频，还出现了各种其他类型的AI恶搞内容。从伪造名人言论到制 造虚假新闻，从生成恐怖场景到编造灾难信息，这些行为都挑战了社会的公序良俗和法律的底线。而 且，随着AI技术的不断发展和普及，生成这类虚假内容的门槛越来越低，几乎任何人都可以轻松上 手。这就意味着，未来可能会有更多人参与到这种不良行为中来，给社会带来的潜在风险也将越来越 大。 "流浪汉闯家中"的AI生成事件，不仅是个人道德和法律意识淡薄的问题，更反映出当前技术监管方 面 ...

Core Points - Tesla's shareholders have approved a new compensation plan potentially worth up to $1 trillion, which CEO Elon Musk seems to celebrate on social media [1] - Musk released videos generated by his company xAI's new tool Grok Imagine, showcasing AI-generated content that has drawn mixed reactions from users [3][5] - Criticism of Musk's social media presence has intensified, with notable figures like Joyce Carol Oates questioning his understanding of life and culture [5][7] Group 1 - Tesla's new compensation plan could reach a value of $1 trillion, reflecting significant shareholder support [1] - Musk's use of AI-generated videos has sparked both interest and ridicule among users on the X platform [3][5] - The backlash against Musk's social media activity highlights a growing concern about his public persona and cultural awareness [5][7]

Group 1 - China Southern Airlines reported the unfortunate death of an employee during a 3-kilometer fitness test, prompting the company to form a special task force to assist the family [2] Group 2 - OpenAI has lifted the invitation-only restriction for its Sora 2 video generation tool, now available to users in the US, Canada, Japan, and South Korea, indicating a significant expansion into the Asian market [3][4] - The company is facing substantial cost pressures, with the cost of generating a 10-second video estimated between $0.5 and $2, leading to the introduction of a $4 "point package" for additional video generation [4] - OpenAI plans to create a "Sora economy" where copyright holders can charge for appearances in videos, addressing previous concerns over default content collection [4] Group 3 - The Russian government has terminated parts of an energy agreement with Finland due to Finland's refusal to purchase Russian electricity since April 2022, impacting cross-border energy cooperation [5] Group 4 - A near-collision incident involving a Southwest Airlines passenger plane and a medical helicopter near Cleveland Hopkins International Airport is under investigation by the National Transportation Safety Board [6] Group 5 - Microsoft’s recent financial disclosures revealed that OpenAI incurred a staggering loss of over $11.5 billion in a single quarter, significantly exceeding market expectations and highlighting the financial strain in the AI sector [7][8] - This loss is particularly notable when compared to OpenAI's revenue of $4.3 billion in the first half of the year, indicating a challenging financial landscape for AI companies [7] Group 6 - The Chinese Ministry of Commerce has responded to issues surrounding Anshi Semiconductor, emphasizing the need for stability in global supply chains and offering support to companies facing difficulties [9] Group 7 - Huang Renxun has completed his stock sell-off plan, cashing out over $1 billion since June, while still holding approximately 3.5% of NVIDIA's shares [12][13] Group 8 - Google and Amazon's investments in AI startup Anthropic have yielded significant profits, with Alphabet reporting a $10.7 billion gain and Amazon seeing a $9.5 billion increase in pre-tax earnings from this investment [16][17][18] Group 9 - The recent donation of 10,000 shares of Kweichow Moutai stock by investor Duan Yongping to a university highlights the trend of philanthropic contributions from wealthy individuals in the tech industry [23][24][25] Group 10 - Elon Musk predicts that in five to six years, traditional smartphones will evolve into AI-driven devices, eliminating the need for operating systems and apps, as AI will generate content in real-time [14]

AI生成的短视频爆火之后，所有人仿佛都被迫兼职当起"AI判官"，刷到一段视频，就要猜是不是AI生成的。有人甚至在电影节上都在质疑演 员是否被偷偷换脸。 人们越来越像是坐在一个无形的评审团里，时刻准备揭穿某个镜头的"造假"。这种怀疑越来越像是一种心理负担。 媒介技术越发达，我们彼此信任的成本也就越高。不过，人类就是这么矛盾，既想保持警惕，又无法抗拒好奇；既担心被欺骗，又享受被惊 艳。 十月的好莱坞有点紧张。哦，不对，不只是好莱坞，可能身在横店的影视行业从业者也有点紧张。 前有AI演员蒂利·诺伍德（Tilly Norwood）亮相苏黎世电影节，后有OpenAI的 CEO萨姆·奥特曼（Sam Altman）满世界上演《碟中谍》和《霸道总裁爱上更 年期的我》。这位CEO一会儿在大厂被训，一会儿操着一口流利的普通话在街采，还要抽空去超市偷显卡。演技之高，戏路之宽，唯有青霞曼玉能与之一 较高下。 （图/《武林外传》） 隔着屏幕的看客感觉已经进入了一个混乱次元——而这一次，混乱的源头就是OpenAI新推出的Sora APP。这款应用不仅能通过文字直接生成视频，连页 面设计都跟TikTok相差无几，竖屏、瀑布流、算法推荐一个 ...