Workflow
Artificial Intelligence
icon
Search documents
物理学家,危,Anthropic联创:AI觉醒,2-3年写出菲尔兹级论文
3 6 Ke· 2026-02-25 10:23
粒子物理十年无新发现,LHC成了「标准模型的坟场」。但Anthropic联创、哈佛物理博士Jared Kaplan却断言:再过2-3年,AI就能写出媲美顶尖物理学 家的论文,50%物理学家或将被彻底取代! 物理学界与科技圈地震! Anthropic联创、前物理学大牛Jared Kaplan放话:两到三年内,理论物理学家有50%概率被AI取代! 要知道,他博士毕业于哈佛大学物理学,同时是JHU的理论物理学教授,又是Anthropic的首席科学官。 对于AI和理论物理学,他都是行家,他的判断绝非无的放矢,白费口舌。 Kaplan引用内部研究与模型进展指出,未来2–3年内,AI在理论推导、数值模拟、公式发现和实验设计等核心科研环节中的能力,将逼近甚至超过大量人 类研究者。 他评估,至少有50%的物理学家工作内容,存在被AI替代或边缘化的明显风险。 替代50%理论物理学家,菲尔兹奖得主亦不例外 自2012年希格斯玻色子(即「上帝粒子」)被发现后,大型强子对撞机(LHC)的实验数据一直严格符合已有理论「标准模型」的预测,没有发现任何预 期之外的新粒子或新物理现象。 戏剧性并非源于希格斯粒子;当它在LHC现身时,其存在已 ...
“新春第一会”五年锁定同一主题,武汉凝聚科技创新开局之势
Core Insights - Wuhan is committed to becoming a nationally influential technology innovation center, leveraging technology as a driving force for the transformation and upgrading of traditional industries [1][2][9] - The city has established a comprehensive innovation ecosystem, focusing on integrating technological and industrial innovation to support high-quality economic development [3][5][6] Group 1: Technology Innovation Achievements - Wuhan has seen significant advancements in technology innovation, with 11 new academicians added in 2025, ranking second nationally, and a total of 91 academicians in the province [3] - The city ranks fifth among major cities in China for R&D expenditure intensity, reflecting a growing commitment to foundational research [3] - Major autonomous innovations have emerged, including the world's first high-precision Beidou chip and significant advancements in semiconductor technology [4][5] Group 2: Policy and Strategic Framework - A series of policy documents have been released to support the construction of the technology innovation center, including the "Regulations on Promoting the Construction of the Technology Innovation Center" and the "Three-Year Action Plan for Building an International R&D Center City" [2][6] - The city aims to implement over 100 major technology projects and achieve breakthroughs in more than 10 key technologies by 2030 [6] Group 3: Innovation Ecosystem Development - Wuhan is developing a multi-faceted innovation space structure, focusing on "One City, Three Corridors, and Multiple Belts" to enhance the integration of technology and industry [1][7][8] - The East Lake Science City is positioned as the core of innovation, with plans to produce over 20 significant outcomes in key areas by 2030 [6][8] - The "Three Corridors" initiative aims to create world-class industrial clusters in photonics, automotive, and digital economy sectors, enhancing regional collaboration [7][8] Group 4: Future Goals and Employment Impact - By 2030, the "Ring University Innovation Development Belt" is expected to create 1 million new jobs for graduates and facilitate over 10,000 technology transfer projects [8] - The city plans to increase the number of specialized and innovative enterprises by 300 and achieve significant revenue growth in high-value-added service industries [8]
Anthropic控告中国AI蒸馏,马斯克和整个互联网都笑了
Sou Hu Cai Jing· 2026-02-25 10:13
2026 年 2 月 23 日,美国 AI 公司 Anthropic 在官方博客发出了一篇义正言辞的声明,指控三家中国 AI 实验室 DeepSeek、月之暗面(Moonshot AI)和 MiniMax 对其旗下 Claude 模型发动了"工业级别的蒸馏攻击"。 据 Anthropic 描述,这三家公司通过约 24,000 个虚假账户,与 Claude 进行了超过 1,600 万次对话交互,意在提取 Claude 的核心能力来训练自家模型。声 明措辞严厉,将此事上升到"国家安全"的高度,声称被蒸馏出的模型可能被"威权政府"用于网络攻击、虚假信息传播、军事用途和大规模监控。 先解释一下"蒸馏"(Distillation)到底是什么。在机器学习领域,知识蒸馏(Knowledge Distillation)最早由 Hinton 等人在 2015 年提出,本意是让一个更 小的"学生模型"通过学习一个更大、更强的"教师模型"的输出来获得能力。简单理解:你问一个聪明的 AI 一堆问题,把它的回答记录下来,再拿这些回 答去训练另一个 AI。 图丨相关论文(来源:arXiv) 这个方法在行业里极其普遍,Anthropic ...
Anthropic指控AI公司蒸馏剽窃,马斯克硬刚“贼喊抓贼”
Sou Hu Cai Jing· 2026-02-25 10:13
Anthropic发难,指控中国AI公司"模型蒸馏"侵权。 2026年2月24日,美国独角兽公司Anthropic发布声明,指控中国三家头部大模型型企业DeepSeek(深度 求索)、Moonshot(月之暗面)和MiniMax利用欺诈性账户和代理服务,窃取Claude大模型能力。该公 司声称,三家中国AI公司使用了一种叫做"模型蒸馏"的技术来提升自身模型能力,即利用一个能力较弱 的模型在更高级模型的输出结果上进行训练。 Anthropic认为,DeepSeek、Moonshot和MiniMax的行为违反其服务条款和区域访问限制,这几家公司通 过约24,000个欺诈性账户生成了超过1,600万次与Claude的交互。 核心指控,Claude模型能力被大规模窃取 Anthropic表示,中国AI公司的"蒸馏"技术都遵循了相似的路径,它们利用欺诈性账户和代理服务大规模 访问Claude,同时逃避检测,目标直指Claude的代理推理(agentic reasoning)、工具使用和编码能力。 DeepSeek(深度求索):超过15万次交互,重点在于提取Claude在多样化任务中的推理能力。这些活动 产生了同步的流量 ...
AI春节“红包大战”,砸出了什么?
Xin Jing Bao· 2026-02-25 10:11
比如字节跳动称,除夕当天,豆包AI互动总次数达到19亿次;阿里宣布,有近2亿用户使用"千问一句 话"下单,点奶茶、订机票酒店、买电影票;腾讯披露,元宝日活用户突破5000万,月活跃用户数达到 1.14亿。 从第三方视角看,参与这场混战的大厂似乎是都是赢家,却又都没有"获胜",行业竞争格局依旧。但如 果说最终是一个多赢格局也不为过,因为确实网民们得了实惠,电视台多了收入,厂商增了用户,中国 AI应用的渗透率,短时间内大幅提高。 字节跳动投入超过10亿元,豆包深度参与央视春晚互动,发出数亿红包和礼品;阿里冠名四家地方卫视 春晚,配合旗下AI助手千问App,耗资30亿元进行春节推广;腾讯投入10亿元、百度投入5亿元,为AI 应用导流。 马年春节刚过,有媒体粗略统计了几大科技巨头假期期间发起的AI红包大战,投入规模堪称空前。而 今大战过后,各家也公布了自己的阶段性成果。 首先,这场AI战役没有任何一家取得像2015年春节期间,微信借助春晚普及了微信红包那样巨大的战 绩,大力没有出奇迹,而更像"内卷式竞争"。 其次,短期流量爆发无法代表长期价值。大规模红包和补贴确实能迅速拉新,但最终能形成多少稳定留 存、付费转化和持续 ...
“砸”了数十亿元,春节AI红包大战落幕!贴得近方能走得远,AI应用的马拉松才刚刚开始
Mei Ri Jing Ji Xin Wen· 2026-02-25 09:51
每经编辑|段炼 多年以后,当我们回望中国AI(人工智能)发展史,2026年的春节一定会被反复提起。 随着阿里、腾讯和字节跳动相继交出成绩单,一场如火如荼的AI红包大战落下帷幕,AI巨头们也完成了史上最大规模的"拉新实验"。 数据显示,2026年春节假期,阿里千问帮大家"一句话下单"近2亿次,腾讯元宝日活跃用户(DAU)一度超5000万,字节跳动旗下的豆包则在除夕夜(2 月16日)创下AI总互动19亿次的纪录。三大巨头用数十亿元的真金白银与场景落地,试图把AI从科技概念变成国民级日常应用。 不过,这场全民实验的真正答卷并不在一时的流量与热度里,而应交给时间去沉淀。 红包能买来一时的下载,但买不来习惯;能换来DAU,但不一定换来心智。2月25日,一度登顶苹果商店免费App排行榜首位的腾讯元宝已经掉至了十名 开外,蚂蚁阿福位居第六,豆包仍在力保榜首位置。 对于整个行业而言,春节一战的意义不在于分出一时高下,而在于为中国AI的普及与落地写下了极具标志性的开篇。 笔者认为,相较于眼花缭乱的技术进步,技术向善更是行业行稳致远的根本底色与不可逾越的底线。科技的温度从来不在于算力有多强、场景有多酷,而 是在于能否守护好用户权 ...
几十亿烧完了,春节AI大战到底谁赢了?
Di Yi Cai Jing· 2026-02-25 09:26
下载榜排名能否转化为超级入口仍是未知数。 伴随春节假期的结束,投入数十亿资金鏖战的"国产大模型春节战"告一段落。 大战期间,字节豆包、阿里千问、蚂蚁阿福、腾讯元宝轮流登顶App Store免费下载榜。截至2月25日,豆包维持第一位,千问第二,蚂蚁阿福跌至第六,元 宝跌至第十八。 从目前的排名情况来看,豆包与千问,字节跳动与阿里巴巴,守住了大模型时代的入口级位置。蚂蚁阿福与元宝跌出核心阵营,标志其该轮冲刺未能成功晋 级。 大洋彼岸的硅谷,大模型厂商在侧重模型性能提升的同时,开始追求务实,全面布局算力生态网络。中美厂商当下不同路径的选择,核心基于商业化发展成 熟度不同。中国SaaS付费习惯尚未全面养成,企业服务市场体量相对有限,巨头们看到的"最大蛋糕"依然是C端流量,即超级App的入口价值。 但需警惕的是,投入数十亿元烧出的下载榜排名能否转化为超级入口,积累的用户数据能否持续推动模型训练性能提升,仍是未知数。 表现在下载端,点点数据显示,2026年1月,大陆市场原生AI APP苹果端上,豆包、蚂蚁阿福、元宝、千问和DeepSeek位列TOP 5,占据了大盘的71%。其 中,豆包以23%的下载量占比位列第一梯队,蚂 ...
Theranos whistleblower on ‘metaphorical black box’ of AI
Yahoo Finance· 2026-02-25 09:00
This story was originally published on CFO.com. To receive daily news and insights, subscribe to our free daily CFO.com newsletter. ARLINGTON, Texas — Though the jury is still out on the long-term implications of the latest artificial intelligence tools, it’s hard not to see some parallels to the Theranos scandal. The medical device company run by convicted fraudster Elizabeth Holmes was full of promise, claiming to have developed a finger-prick blood test that could screen for a range of diseases and d ...
千问3.5登顶全球开源大模型榜首
Xin Lang Cai Jing· 2026-02-25 08:42
Core Insights - Alibaba's Qianwen 3.5 has topped the global open-source large model rankings, as reported by Hugging Face, the largest AI open-source community [1] - The model, which was open-sourced on New Year's Eve, has a total of 397 billion parameters, with only 17 billion activated, and its performance is comparable to Gemini 3, while its token cost is only 5% of Gemini 3's [1] - Notably, 8 out of the top 10 models in the rankings are from Chinese teams, including Alibaba, Zhipu, MiniMax, and Yuezhianmian [1]
Anthropic被曝放弃AI安全核心承诺
Xin Lang Cai Jing· 2026-02-25 08:41
Core Viewpoint - Anthropic, an AI startup founded by former OpenAI members, is revising its risk mitigation policy, raising concerns about AI safety governance in the industry [1][3] Group 1: Policy Changes - Anthropic has implemented a "Responsible Scaling Policy" (RSP) since 2023, which included a commitment to not train or release any AI models without sufficient risk mitigation measures in place [3] - The company has recently decided to overhaul the RSP, removing the key commitment that previously garnered praise for its focus on safety [3] Group 2: Industry Context - The shift in Anthropic's policy comes as major competitors like OpenAI and Google accelerate their development of large AI models, putting Anthropic at risk of being marginalized [3] - Jared Kaplan, the Chief Scientist of Anthropic, stated that halting AI model training is not beneficial, especially in a rapidly evolving technological landscape where competitors may gain an advantage [3]