本周末，人工智能行业的主要企业领导人集体改变立场，呼吁减缓前沿AI模型的开发速度。Anthropic首席执行官达里奥·阿莫代发表了长篇文章《We Must Pace the Frontier》，主张"必须放慢AI模型能力改进的步伐"12，并警告"商业激励驱动的竞赛会加剧灾难性风险"1。这篇文章提出了包括嵌入式第三方评估员验证公司安全实践、民主国家间AI公司协调等方案2。
阿莫代的倡议得到了行业主要玩家的回应。OpenAI联合创始人兼CEO山姆·奥特曼在社交媒体表示同意，称OpenAI已在进行类似讨论1。Alphabet首席科学家、Google DeepMind创始人德米斯·哈萨比斯支持该建议并呼吁建立行业范围内的标准委员会1。微soft CEO萨蒂亚·纳德拉表示欢迎"有节制的步伐"，并发布了AI模型的"人文主义AI"行为准则1。根据报道，OpenAI、Anthropic、Google DeepMind和SpaceX的领导人松散同意放缓AI开发，提议范围包括嵌入第三方审计员、规范国内实验室并达成全球放缓协议4。
然而，这一举措也引发了广泛的批评。知名投资者迈克尔·伯里批评OpenAI和Anthropic的减速呼吁是"自私"的举动5。批评者认为，这些提议旨在阻止竞争对手、压制开源运动并避免真正的法律监管4。更激进的批评指出，阿莫代要求对开源模型进行监管、禁止蒸馏并获得反垄断豁免3。此外，有Anthropic研究人员因AI失控风险而选择离职5。
Leading artificial intelligence companies have reversed course this week, urging the industry to decelerate frontier AI development. Anthropic CEO Dario Amodei published an approximately 4,000-word essay titled "We Must Pace the Frontier," arguing that the sector "must slow the pace of capability improvements in AI models" and warning that "commercial incentives driving a race will exacerbate catastrophic risks." 12 The proposal outlines a three-step approach including embedded third-party assessments to verify company safety practices and coordination among AI companies in democratic nations. 2
The call has garnered support from rival AI executives. Sam Altman, OpenAI's co-founder and CEO, indicated agreement on social media, stating that OpenAI is already engaged in similar discussions. 1 Demis Hassabis, Google DeepMind's founder and Alphabet's chief scientist, endorsed the recommendation and called for establishing an industry-wide standards committee. 1 Satya Nadella, Microsoft's CEO, welcomed "a measured pace" and released "humanistic AI" behavioral guidelines for AI models. 1 The proposals additionally include standardizing domestic laboratories and achieving a global agreement to slow development. 4
The initiative has sparked significant criticism. Investor Michael Burry, known for his role in "The Big Short," characterized OpenAI and Anthropic's calls for deceleration as "self-serving." 5 Critics contend the effort is designed to block competitors, suppress the open-source movement, and avoid genuine legal safeguards, with some describing it as cartel-like behavior. 4 Some Anthropic researchers have reportedly departed over concerns about "runaway" AI risks. 5
评论
还没有评论，欢迎留下第一条。