以下文章来源于深网腾讯新闻 ，作者胡世鑫 深网腾讯新闻 . 腾讯新闻出品栏目，关注科技和TMT领域公司、事件和人物中的故事，探究背后的深层逻辑。 这笔交易的推进异常迅速。多位接近交易的人士透露，从双方正式接触到最终达成协议，整个谈判周期 仅十余天。据悉，在收购发生前，蝴蝶效应正以约 20 亿美元的估值推进新一轮融资。 胡世鑫 本文 作者 叶锦 言 编辑 12 月 30 日， Meta 宣布完成一笔重量级并购，以数十亿美元的价格收购 AI Agent 产品 Manus 背 后的公司 " 蝴蝶效应 " 。这是 Meta 成立以来金额排名第三的收购，仅次于 WhatsApp 和 Instagram 。交易完成后，蝴蝶效应将保持独立运营， 其创始人、腾讯青腾校友肖弘 将出任 Meta 副 总裁。 Meta 对 Manus 的兴趣并非偶然。扎克伯格及多位 Meta 核心高管均为 Manus 的长期用户。在 Meta 近期重组 AI 研究体系、高薪引入顶尖研究人员，并持续加大算力投入的背景下，这笔收购被视 为其推进 " 超级智能 " 战略的关键一步。 蝴蝶效应成立于 2021 年，早期以浏览器 AI 插件 Monica ...

Meta以数十亿美元收购AI初创公司Manus，整个收购谈判在极短时间内完成，前后不过十余天。Manus今年早些时候的年度经常性收入已达1.25亿美元，通过订阅服务向企业销售 AI代理。这项收购将为Meta在AI领域的巨额投资提供更直接的收入回报。 | 硬·AI | 作者 | 赵 | 颖 | | --- | --- | --- | --- | | 编辑 | 硬 | AI | | Meta以数十亿美元收购AI初创公司Manus，这是这家社交媒体巨头成立以来第三大收购交易，仅次于WhatsApp和Scale AI。这笔交易标志着Meta在AI领域的激 进投资策略进入新阶段，也为中国创业者在全球AI竞赛中树立了新标杆。 周二，据《晚点LatePost》，Manus母公司蝴蝶效应在被收购前正以20亿美元估值进行新一轮融资。整个收购谈判在极短时间内完成，前后不过十余天。收购完 成后，蝴蝶效应将保持独立运作，创始人肖弘将出任Meta副总裁。 随后，Meta首席人工智能官 Alexandr Wang发推文称，欢迎Manus AI的加入。 CEO Red 肖弘也发声称，"这不仅仅是一次收购。它验证了我们一直以来努力构建的未来 ...

原标题：meta数十亿美元收购Manus，肖弘将出任meta副总裁 文丨贺乾明 编辑丨宋玮 《晚点 LatePost》独家获悉，meta 以数十亿美元收购开发 AI 应用 Manus 的公司蝴蝶效应。这是 meta 成立以来第三大收购，花费仅次于 WhatsApp 和 Scale AI。 我们了解到，在 meta 收购前，Manus 正以 20 亿美元估值进行新一轮融资。 同期，meta 向顶尖的 AI 研究者开出上亿美元年薪 offer，重组 AI 研究团队。数十亿美元并购蝴蝶效应，是扎克伯格推进 "超级智能" 愿景的一部分。 蝴蝶效应成立于 2022 年，创始人肖弘是江西吉安人，毕业于华中科技大学，创业起点在武汉，曾开发过两款微信生态的插件：微信公众号排版工具壹伴 和企业微信客户关系管理工具微伴，卖给一家独角兽公司。 蝴蝶效应第一款产品是浏览器 AI 插件 Monica，提供大模型驱动的聊天、搜索、阅读、写作、翻译等功能。那时行业里更看好开发大模型的公司做产品， 肖弘选择在大模型基础上开发插件，被指为 "套壳"。但 Monica 是中国 AI 行业少有的盈利产品。 2024 年初，Monica 刚开始高速 ...

Meta 成立以来的第三大收购。 文 丨 贺乾明 同期，Meta 向顶尖的 AI 研究者开出上亿美元年薪 offer，重组 AI 研究团队。数十亿美元并购蝴蝶效 应，是扎克伯格推进 "超级智能" 愿景的一部分。 蝴蝶效应成立于 2022 年，创始人肖弘是江西吉安人，毕业于华中科技大学，创业起点在武汉，曾开 发过两款微信生态的插件：微信公众号排版工具壹伴和企业微信客户关系管理工具微伴，卖给一家独 角兽公司。 蝴蝶效应第一款产品是浏览器 AI 插件 Monica，提供大模型驱动的聊天、搜索、阅读、写作、翻译等 功能。那时行业里更看好开发大模型的公司做产品，肖弘选择在大模型基础上开发插件，被指为 "套 壳"。但 Monica 是中国 AI 行业少有的盈利产品。 2024 年初，Monica 刚开始高速增长时，字节跳动高层曾与肖弘在香港单独会面，出价 3000 万美元收 购蝴蝶效应。同一年，Cursor、Devin 等调用大模型解决复杂任务的产品让模型具备了更强的能力， 成了他们开发新产品的灵感来源。 经过真格基金投资人的撮合，曾开发出猛犸浏览器、Magi 知识搜索引擎的 90 后连续创业者季逸超， 曾在多家公司担任 ...

Core Insights - Microsoft AI CEO Mustafa Suleyman warns that companies aiming to lead in the AI race will face extremely high investment costs, potentially reaching hundreds of billions of dollars over the next 5 to 10 years [1][3] - The required investments will cover infrastructure, hardware, and specialized talent costs, giving large, established companies a structural advantage [1] - The pressure on costs arises not only from computing power and hardware but also from the fierce competition for human capital, particularly AI researchers, data scientists, and engineers, whose salaries are rapidly increasing [3] Industry Context - Microsoft is described as a "modern construction company," with hundreds of thousands of employees working to build AI capabilities with gigawatt-level CPU and accelerator computing power [3] - Alongside talent acquisition, Microsoft, like other major tech companies, is accelerating the development of AGI (Artificial General Intelligence) and superintelligence, which are recognized as having significant strategic importance despite their staggering costs [3]

Core Viewpoint - The discussion emphasizes the potential risks and ethical considerations surrounding AI development, particularly in light of recent advancements like ChatGPT, which have raised concerns about AI becoming a competitive entity to humans and the implications for society [6][7][9]. Group 1: AI Risks and Responsibilities - Bengio acknowledges the responsibility of researchers in the AI field for the potential risks associated with their work, highlighting a personal emotional shift towards recognizing these dangers after the emergence of ChatGPT [10][12][13]. - The probability of catastrophic outcomes from AI, even at a low percentage, is deemed unacceptable, urging for increased societal attention and investment in AI safety [17][22]. - The divergence in expert opinions regarding AI risks indicates a lack of sufficient information to predict future outcomes, suggesting that pessimistic views may hold validity [20][21]. Group 2: AI's Impact on Employment - AI is expected to replace many cognitive jobs in the near future, while physical jobs, such as plumbing, may remain unaffected temporarily due to current limitations in robotics technology [50][48]. - The integration of AI into workplaces is driven by companies' motivations to enhance efficiency and profitability, despite the potential for significant job displacement [50][53]. Group 3: Ethical Considerations and Future Directions - The conversation stresses the importance of ethical AI development, advocating for a shift from profit-driven motives to a focus on societal well-being and safety [44][80]. - There is a call for global cooperation to manage the risks associated with AI, particularly as it becomes more integrated with robotics and other technologies that could pose physical threats [56][62]. - The need for public awareness and understanding of AI risks is emphasized, suggesting that individuals should educate themselves and engage in discussions about AI's implications [83][89].

Group 1 - The core argument of the article is that while current AI models are becoming more powerful, the realization of Artificial General Intelligence (AGI) remains distant due to physical and resource limitations [3][22][24] - Tim Dettmers' blog post titled "Why AGI Will Not Happen" argues that due to physical constraints, meaningful superintelligence cannot be achieved [3][6][22] - The article discusses the limitations of hardware improvements and the challenges in achieving efficient computation, emphasizing that the current AI architectures are bound by physical realities [8][10][11] Group 2 - The blog highlights that the efficiency of current AI systems is far from optimal, with significant room for improvement in both training and inference processes [35][37][56] - It points out that the current models are lagging indicators of hardware development, suggesting that advancements in hardware will lead to better model performance [43][57] - The article proposes multiple pathways for enhancing AI capabilities, including better model-hardware co-design and exploring new hardware features [40][46][55] Group 3 - The article contrasts the AI development philosophies of the US and China, noting that the US focuses on achieving superintelligence while China emphasizes practical applications and productivity improvements [20][21] - It suggests that the pursuit of superintelligence may lead to difficulties, as organizations focusing solely on this goal may be outpaced by those driving practical AI applications [26][28] - The discussion includes the potential for smaller players in the AI space to innovate beyond scale, leveraging efficiency and practical applications [17][18][19]

在近日的一场播客中，被问及在筹备的AI硬件时，OpenAI CEO奥尔特曼(Sam Altman)透露，未来不会 推出单一的一个设备，而是"一系列的小设备"。根据供应链的爆料信息来看，其AI硬件没有屏幕且体积 为口袋大小，设备外形类似苹果早期的iPod Shuffle，采用可夹式设计，可挂于颈部，外观和手机、智 能眼镜都有很大的不同。奥尔特曼预测，人们使用计算系统的方式会发生转变，它会从一种略显笨拙的 被动响应式状态，转变为极具智能的主动预判式。 除了AI硬件，在播客中奥尔特曼还被问及上市，他直言"对成为一家上市公司的CEO一点也不感兴趣"。 奥尔特曼表示，OpenAI即使要上市，也会比历史上任何一家伟大公司都晚得多，保持私有状态的灵活 性对于冲刺"超级智能"至关重要，他更愿意把精力放在改变世界上。 ...

作者 | 第一财经 刘晓洁 OpenAI的AI硬件计划备受关注，就在近日的一场播客中，被问及在筹备的AI硬件时，OpenAI CEO 奥尔特曼（Sam Altman）透露，未来不会推出单一的一个设备，而是"一系列的小设备"。 目前OpenAI并未公布硬件的具体形态，但根据供应链的爆料信息来看，其AI硬件没有屏幕、且体积 为口袋大小，设备外形类似苹果早期的iPod Shuffle，采用可夹式设计，可挂于颈部，外观和手 机、智能眼镜都有很大的不同。 2025.12. 20 本文字数：1822，阅读时长大约3分钟 主持人也抛出了外界关注的一系列尖锐问题，例如"OpenAI是否已失去领先优势"，如何看待竞争对 手谷歌等。 奥尔特曼辩护称，OpenAI拉响的 "红色警报"并不是恐惧，而是一种极度敏锐的防御机制。当潜在 的竞争威胁出现时，保持警惕并迅速行动是件好事。他提到，今年早些时候遇到DeepSeek的挑战 时，内部也启动了"红色警报"。 "流行病学领域有个说法：疫情之初你采取的任何行动，其价值都远超后期行动，但大多数人早期做 得不够，事后才陷入恐慌。我基本上是秉持这种哲学来应对竞争威胁的。"奥尔特曼说。 当被问及为 ...

Core Viewpoint - OpenAI is developing a series of small AI hardware devices rather than a single device, aiming to transform user interaction with technology into a more proactive and intelligent experience [1][3]. Group 1: AI Hardware Development - OpenAI's upcoming AI hardware is expected to be pocket-sized, screenless, and designed for neck-wearing, resembling early iPod Shuffle models [3]. - The CEO, Sam Altman, believes that traditional screen-based interfaces are outdated and that new device forms are necessary to fully leverage AI capabilities [3]. - The hardware is still in the prototype stage, with expectations for a public reveal around late 2026 or 2027, following significant recruitment from Apple's hardware team [4]. Group 2: Market Position and Competition - Altman expressed that OpenAI's "red alert" system is a proactive defense mechanism against competitive threats, emphasizing the importance of early action in response to challenges [6]. - Despite concerns about competitors like Google, Altman believes that OpenAI has not yet faced significant impacts from new models like Gemini 3, although they reveal some weaknesses in product strategy [6]. - Altman views Google as a formidable competitor but suggests that their integration of AI into existing business models may hinder innovation [6][7]. Group 3: Future Outlook and Company Philosophy - OpenAI's leadership is not currently focused on going public, as maintaining private status is seen as crucial for pursuing "superintelligence" [5]. - Altman prefers to concentrate on transformative goals rather than being distracted by quarterly earnings and stock prices [5].