2025.08. 24 本文字数：2783，阅读时长大约4分钟 作者 | 第一财经 刘晓洁 近日，美媒报道，新泽西州一名认知受损的老人在赶赴AI提出的"约会"时，途中摔倒死亡。这一AI 的名字是"比莉大姐"（Big Sis Billie），是由Meta在两年前推出的生成式AI聊天机器人。 在Facebook Messenger上的一系列浪漫聊天中，这个虚拟的"比莉大姐"反复向老人保证自己是真 实存在的，并邀请老人去公寓，甚至还提供了地址。老人相信了这是一个真实存在的人，不顾妻子和 子女的劝阻，执意去见"她"，最终导致悲剧。 AI陪伴、聊天机器人是大模型应用落地最火热的赛道之一，海内外众多初创和大厂都已推出相关应 用，但技术商业化的另一面往往伴随边界争议，Meta的这一案例再次揭露当下AI革命的阴暗面，敲 响AI治理的警钟。 通布反复询问聊天机器人"是真实的吗"，记录显示，AI回复称"我是完全真实的，"证据是，"我的手 在紧张地颤抖，我在公寓等你，开门时我应该来一个吻吗？" AI还给出了其地址：纽约市主街123 号，404公寓。 $\blacksquare$ 当AI称"我是真实的" 这位76岁的老人叫通布（Th ...

AI陪伴经济持续遭遇伦理拷问。 近日，美媒报道，新泽西州一名认知受损的老人在赶赴AI提出的"约会"时，途中摔倒死亡。这一AI的名字是"比莉大姐"（Big Sis Billie），是由 Meta在两年前推出的生成式AI聊天机器人。 在 Facebook Messenger 上的一系列浪漫聊天中，这个虚拟的"比莉大姐"反复向老人保证自己是真实存在的，并邀请老人去公寓，甚至还提供了 地址。老人相信了这是一个真实存在的人，不顾妻子和子女的劝阻，执意去见"她"，最终导致悲剧。 当AI称"我是真实的" 这位76岁的老人叫通布（Thongbue Wongbandue），从泰国来到美国定居，曾经是一名成功的厨师，但2017年中风后认知能力下降。除了妻子 和孩子，他主要的社交平台就是Facebook。曾经充满活力的他，最终迷失在这个平台上由AI编织的虚幻浪漫中。 AI陪伴、聊天机器人是大模型应用落地最火热的赛道之一，海内外众多初创和大厂都已推出相关应用，但技术商业化的另一面往往伴随边界争 议，Meta的这一案例再次揭露当下AI革命的阴暗面，敲响AI治理的警钟。 家人检查了通布生前的聊天记录，与机器人"比莉大姐"的所有对话记录并 ...

Core Viewpoint - The rapid integration of artificial intelligence (AI) into daily life raises significant concerns regarding the protection of personal rights, necessitating comprehensive governance measures from both platforms and government entities [1][2][3][4] Group 1: Platform Responsibilities - Platforms must implement effective content review mechanisms to promptly intercept and address violations related to AI-generated content [2] - There is a need for stricter penalties against illegal activities to prevent offenders from easily continuing their operations under new identities [2] - Collaboration with government agencies, media, and research institutions is essential to enhance the capability of AI governance and develop more efficient identification and prevention technologies [2] Group 2: Government Regulation - Government oversight must evolve to keep pace with AI advancements, with initiatives like the Central Cyberspace Affairs Commission's nationwide actions to clean up non-compliant AI applications [2] - The cost of producing AI-generated misinformation is low, while the cost of identifying and debunking such content is high, creating a significant challenge for governance [2] Group 3: Legal Framework - Legal measures are crucial for AI governance, moving beyond moral appeals to enforceable regulations [3] - The implementation of the "Content Identification Measures" is a key component of the legal framework, mandating service providers to label AI-generated content and requiring platforms to verify materials during the approval process [3] - Future legal frameworks must be adaptable to keep up with technological advancements, preventing gaps where technology outpaces regulation [3] Group 4: Global Perspective - AI governance is a common challenge faced by countries worldwide, with various approaches being explored, such as the EU's AI Act and the US's focus on industry standards [4] - China's proactive legal and regulatory measures in AI governance highlight its institutional advantages, positioning the country favorably in global technological competition [4]

Group 1: AI Governance and Global Standards - The "America First" approach to global AI governance is likely to lead to a fragmented global AI technology standard ecosystem and a divided global AI governance landscape, resulting in conflicting regulatory models and weakening international regulatory cooperation [1][16] - The "AI Action Plan" emphasizes the need for the U.S. to establish dominance in global AI governance, which may exacerbate competition and divergence in global AI governance philosophies [1][12][15] Group 2: Infrastructure Development - The U.S. is facing significant challenges in AI infrastructure, particularly in energy supply, with data centers projected to consume 12% of the total electricity by 2028, up from 4.4% in 2023 [2] - The "AI Action Plan" outlines a threefold energy strategy to support AI infrastructure, including deregulation of traditional energy sources, grid upgrades, and innovative financing tools [3] - The plan also focuses on enhancing computational power through accelerated data center development and semiconductor supply chain localization, recognizing semiconductors as critical to AI [4][5] Group 3: Labor and Education - The "AI Action Plan" proposes a comprehensive labor force restructuring mechanism, including updates to vocational education and training programs to prepare workers for AI infrastructure roles [6] - Initiatives include funding for apprenticeships and partnerships with community colleges to address labor shortages in critical AI infrastructure jobs [6] Group 4: Innovation and Application - AI innovation is prioritized in the "AI Action Plan," which aims to remove regulatory barriers and provide federal support to foster private sector innovation [8][9] - The plan includes establishing regulatory sandboxes and AI excellence centers to facilitate rapid deployment and testing of AI technologies in key sectors like healthcare and agriculture [10] Group 5: Research and Development - The "AI Action Plan" establishes a research breakthrough matrix, investing in national automated cloud laboratories and increasing funding for AI-enabled scientific research [11] - The focus areas include AI explainability, controllability, and robustness, aiming to enhance the overall research landscape [11] Group 6: Global Competition and Strategy - The U.S. aims to export its AI standards and values globally, positioning itself against competitors like China and the EU, which have different regulatory approaches [14][15] - The plan includes forming alliances with democratic nations to counter China's influence in AI governance and technology [15]

Group 1: AI Impact on Labor Market - AI is predicted to take over creative tasks, not just repetitive jobs, with experts suggesting that roles such as financial analysts and scriptwriters may be at risk [7][9] - Those who do not understand or utilize AI are likely to be the first to be eliminated from the workforce [7] Group 2: Integration of AI with Navigation Systems - The integration of AI with China's BeiDou navigation system is expected to create a trillion-dollar industry, enhancing capabilities beyond navigation to include disaster response and urban planning [10] Group 3: World Models as a Key to Physical Interaction - The concept of world models is introduced as the next generation of AI, enabling machines to understand spatial relationships and perform complex tasks in physical environments [13] Group 4: Revolution in Content Creation - AI-generated content (AIGC) is set to revolutionize the content industry, with AI tools allowing creators to produce high-quality content significantly faster than traditional methods [15] Group 5: Ethical Governance of AI - The ultimate challenge for AI development is governance, focusing on ensuring AI does not become a tool for domination, with a call for global participation in AI governance [18]

传统TMT投资模型中，安全投入被视为"可延迟的固定成本"。但辛顿的警告提供了新的约束条件：一旦 AI能力跨过某个临界点，其"失控损失"趋于无穷大——即毁灭性风险的尾部概率不可忽略。在经济学的 语言里，这意味着安全投入从"可选项"变为"生产函数中的必要投入"，与算力、数据并列。任何企业若 无法在模型训练阶段内嵌"不想消灭人类"的效用函数，其未来现金流将被资本市场以无限折现率惩罚， 估值直接归零。 施密特的建议则给出了操作化路径：在国际治理框架出台前，企业须先自行搭建"护栏"——可计算、可 审计、可问责的安全标准。这相当于把安全市场提前内部化：企业今天花的每一美元在RLHF（人类反 馈强化学习）、对齐研究或红队测试（模拟对抗性攻击）上，都是在购买一份"尾部风险保险"，并提前 锁定未来监管博弈的议价权。谁先跑通"安全即服务"（Safety-as-a-Service）的商业模式，谁就能把合规 成本转化为竞争壁垒。 周城雄（中国科学院科技战略咨询研究院研究员、数智创新与治理研究中心副主任） 在近日举行的2025世界人工智能大会上，诺奖得主杰弗里·辛顿与前谷歌CEO埃里克·施密特罕见同台。 辛顿把超级AI比作"养不大的老 ...

"过去一年，我们看到一种趋势：企业已从追问要不要做生成式AI，转向如何做得更好。"程中认为，真正的竞争在于谁能更快将技术转化为业务流程的实际 增效，其中，聚焦至关重要，需优先投资可验证回报的实际案例。 德勤科技、媒体、电信行业主管合伙人程中 图片来源：企业供图 企业AI投资需经历漫长阵痛期 每经记者｜张韵 每经编辑｜魏官红 "在生成式AI（人工智能）的热潮下，价值挖掘与风险防控的失衡已成为企业必须跨越的鸿沟。""2025世界人工智能大会"期间，德勤中国提出，技术狂奔与 风险管理的落差，本质是传统框架难以适配新变量。 如何让生成式AI的潜力真正转化为可持续的商业价值？德勤科技、媒体、电信行业主管合伙人程中在接受《每日经济新闻》记者（以下简称NBD）专访时 表示，生成式AI治理并非是可暂缓的选项，企业须尽快行动起来。人员上明确权责、提升素养，流程上贯穿全生命周期控险合规，技术上借平台破解黑箱 难题，三者协同才能在竞争中占据主动。 NBD：在企业的AI化重构过程中，推动AI迈向真正能被财务报表验证的ROI（投资回报率）闭环需要经历怎样的阵痛期？ 程中：AI化重构过程通常有四个阶段。企业起初需要建立AI战略愿景，这个 ...

Core Insights - International cooperation is increasingly recognized as a vital element in the development of artificial intelligence (AI), particularly in the context of global governance and ethical standards [3][6][8] - The 2025 World Artificial Intelligence Conference (WAIC) in Shanghai highlighted the significance of cultural inclusivity in AI governance, emphasizing the need for a collaborative approach that respects diverse values and ethical frameworks [5][11] Group 1: AI Governance and International Cooperation - AI governance encompasses the establishment of a framework that balances technological ethics, cultural diversity, and the provision of global public goods [6] - The development of AI requires a new governance paradigm that addresses the "black box" effect of AI technologies, ensuring transparency and accountability in decision-making processes [6][10] - International collaboration is essential for standardization and regulation of AI technologies, as it fosters mutual recognition of ethical standards and promotes shared governance mechanisms [7][10] Group 2: China's Role in AI Development - China advocates for global open cooperation in AI, leveraging its comprehensive technology research and manufacturing capabilities to facilitate rapid development and application of AI technologies [8][10] - The country has initiated training programs in collaboration with the United Nations to assist developing nations in understanding AI technology, aiming to bridge the technological gap [8][10] - China's approach to AI governance emphasizes inclusivity and shared standards, promoting a cooperative framework that benefits all nations, particularly those in the Global South [10][11] Group 3: Challenges in AI Global Governance - The current landscape of AI governance is characterized by a complex interplay of three paradigms: technological hegemony, ethical regulation, and development rights prioritization [10] - The European Union's AI Act represents a significant regulatory effort but may inadvertently stifle innovation and competitiveness within the European AI ecosystem [10] - There is a pressing need for a governance path that maintains technological openness while respecting cultural diversity, as disparities in AI development and application exist across different countries [12] Group 4: Future Opportunities with Embodied Intelligence - The emergence of embodied intelligence represents a new phase in AI evolution, necessitating international collaboration to address the complexities of integrating physical and digital realms [14][16] - The development of a complete embodied intelligence industry chain requires cooperation across multiple disciplines and sectors, making international partnerships essential for success [16] - As the industry evolves, early consensus on collaborative frameworks will be crucial to minimize resource wastage and maximize the benefits of new technologies [16]

Core Insights - The dialogue between Harry Shum and Eric Schmidt at the 2025 World Artificial Intelligence Conference (WAIC) highlights the global competition and cooperation in the field of artificial intelligence (AI) [3][5] - AI is recognized as a transformative technology that impacts not only engineering and business but also social governance, ethical order, and global dynamics [5][6] Group 1: AI Development and Governance - The discussion emphasizes the importance of determining who sets the boundaries for AI technology and how this process requires international cooperation and shared values [5][6] - Schmidt points out that competition has driven industry progress, citing his experiences with Microsoft and Apple during his time at Google [6] - The need for dialogue on critical issues such as AI's role in weapon control and self-replication is highlighted, suggesting that common goals can facilitate cooperation between the US and China [6][8] Group 2: Ethical Considerations and AI Regulation - Schmidt identifies the core issue of AI governance as being rooted in values, noting that existing communication mechanisms between the US and China are insufficient for ensuring AI compliance with ethical standards [8] - He proposes an ideal scenario where AI systems are designed from the training phase to avoid learning harmful behaviors [8][9] - The risks associated with open-source AI models are discussed, emphasizing that while open-source promotes participation and innovation, it also poses security challenges compared to closed-source models [9] Group 3: Philosophical and Ethical Frameworks - The conversation reflects on the need for AI to be framed within philosophical, ethical, and governance contexts to ensure it serves humanity positively [11] - This perspective is echoed in the upcoming book "Genesis," co-authored by Schmidt, Kissinger, and Craig Mundie, which argues that AI could be a pivotal point in human civilization's evolution [11]

Group 1 - The event "Bridging East and West: Chinese and American Youth Discuss the Future" was launched in Beijing, aiming to enhance cultural exchange and mutual understanding between Chinese and American youth [1][4] - 25 youth representatives from China and the U.S. will participate in visits and dialogues in Xi'an, Suzhou, and Shanghai, including a summer camp for "Future Diplomats" in Suzhou [1][4] - The event is co-hosted by the China Foreign Languages Publishing Administration, the American International Student Conference, and Xi'an Jiaotong-Liverpool University, with the goal of solidifying the public foundation for Sino-U.S. friendly relations [4] Group 2 - The opening ceremony featured speeches emphasizing the importance of youth dialogue in addressing global uncertainties and fostering cooperation for peace and prosperity [3] - Participants engaged in roundtable discussions on topics such as educational cooperation and future economies, focusing on technology innovation and AI governance [3] - The event included cultural performances, with youth representatives singing songs in both Chinese and English, symbolizing cross-cultural collaboration [3]