百度近期举办的AIDU计划OpenDay活动，再次让其2026届"AIDU计划"受到关注！ 此次活动传出，为招纳到顶尖的校园人才，百度2026届"AIDU计划"的offer薪资上不封顶。 "AIDU计划"是百度推出的一项精英招募计划，以培养AI技术领军人才为目标，特别注重候选人对技术的纯粹热爱与追求。加入该计划的学员有机会接触到最前沿的 研究课题，并在业务场景中实现技术的落地应用。百度将为每位学员量身定制个性化的培养路径，并配备专属技术专家导师进行一对一指导，助力他们快速成长为 AI领域的佼佼者。 今年5月，百度2026届"AIDU计划"正式启动，招募对象为2026届博士和硕士生，倾向于超级学霸、学术大神、竞赛高手和工程大牛。 不过，大模型Top级别的岗位要求也较高。一家大模型企业相关负责人就坦言，公司期望招揽的人才具备以下特征：敢于突破固有经验和传统范式；能够凭借技术 创新创造产品价值与用户价值，收获正向反馈；认同公司文化，有志于与公司携手打造行业领军企业。 在此次AIDU计划OpenDay活动中，百度透露，相比去年，2026届"AIDU计划"岗位招聘扩增超60%，覆盖百度23个核心业务和11类研究方向，包 ...

百度近期举办的AIDU计划OpenDay活动，再次让其2026届"AIDU计划"受到关注！ "AIDU计划"是百度推出的一项精英招募计划，以培养AI技术领军人才为目标，特别注重候选人对技术的纯粹热爱与追求。加入该计划的学员有机会接触到最前沿的 研究课题，并在业务场景中实现技术的落地应用。百度将为每位学员量身定制个性化的培养路径，并配备专属技术专家导师进行一对一指导，助力他们快速成长为 AI领域的佼佼者。 今年5月，百度2026届"AIDU计划"正式启动，招募对象为2026届博士和硕士生，倾向于超级学霸、学术大神、竞赛高手和工程大牛。 此次活动传出，为招纳到顶尖的校园人才，百度2026届"AIDU计划"的offer薪资上不封顶。 在此次AIDU计划OpenDay活动中，百度透露，相比去年，2026届"AIDU计划"岗位招聘扩增超60%，覆盖百度23个核心业务和11类研究方向，包含大模型算法、大 模型基础架构、机器学习、语音技术、智能体等，是其最大规模的顶尖AI人才招聘。 "百度会像培养飞行员一样，培养未来的AI领航者。"百度方面称。据了解，在培养体系方面，学员不仅可以参与核心AI项目，接触前沿技术挑战，还能享受 ...

Core Viewpoint - The article emphasizes the importance of innovation for companies to thrive in a competitive landscape, highlighting that waiting and observing can hinder progress and success [2][4]. Group 1: Innovation and Competition - Chinese technology companies have gained global attention through significant innovations, such as DeepSeek and DJI, showcasing the potential of autonomous innovation [2]. - Founders of leading companies express a sense of urgency and the necessity for continuous innovation to maintain competitiveness, indicating that past strategies may not be applicable in the future [3]. - The article warns against complacency, using Nokia as an example of a company that failed to adapt and innovate, leading to its decline in the smartphone era [3]. Group 2: Embracing Failure - The article discusses the need for a culture that tolerates failure in the innovation process, suggesting that many successful innovations are built on lessons learned from previous failures [4][5]. - It highlights the importance of allowing innovators the space to experiment and learn from mistakes, as exemplified by Huawei's approach to innovation [5]. Group 3: Managing Expectations - There is a caution against setting unrealistic expectations for innovation outcomes, as some innovators may face setbacks when high expectations are not met [6]. - The article stresses that innovation is a gradual process that requires patience and realistic goal-setting, advocating for a balanced approach to innovation that respects the time needed for development [6]. Group 4: Technological Advancements - Recent technological breakthroughs in China, such as advancements in robotics and aerospace, demonstrate the long-term commitment and perseverance of research teams [6]. - The transformation of Beijing's Zhongguancun from a market to a tech hub illustrates the importance of sustained institutional support for innovation [6].

责编 | 王启隆 出品 | AI 科技大本营 （ID：rgznai100） 蒋涛看到的，是在这股浪潮之下，压在我们头顶沉甸甸的"三座大山"： 这不仅是中国的困境，也是所有非英语世界、所有不愿将自身数据和命运完全交由他人的国家、企业 和个人的共同困境。 为了破局，只能建立起自己的数据栈。这对行业带来了空前的影响，而 蒋涛也对此做出了 一个激进 却清醒的预 言： 「 "码盲"会逐渐消失，新程序员将随后崛起。 」 当写一个 App 像写一篇文章那样简单，当一个大厂产品经理能独立开发出冲上 App Store 付费榜的 应用，当过去"传话游戏"般的开发流程被彻底颠覆……一个属于全民创造的时代，才真正露出了它的 第一缕晨光。 以下为 CSDN 创始人蒋涛的演讲实录，他将为我们揭示，如何在这场巨变中，翻越山丘，找到自己 的位置。 以 CUDA 为代表的"算力霸权"； 以封闭模型为代表的"模型霸权"； 以英文语料为代表的"数据霸权"。 一个时代正在更迭，以一种肉眼可见的速度。 "ChatGPT 用史上最短时间达到了八亿用户，一枝独秀。 但 DeepSeek 登顶全球一百多个国家，带 来了全球性的平权化。 " COPU 副主 ...

Group 1 - Multinational companies are increasingly investing in China, shifting focus from traditional manufacturing to high-tech sectors, indicating strong international confidence in the Chinese market [2][3] - The rise of cutting-edge technology firms and emerging tech clusters in China highlights the country's transformation into a global innovation hub, creating new investment opportunities [2][3] - China's complete industrial system and large market scale provide diverse application scenarios for new technologies, fostering a unique competitive advantage and new growth points across various sectors [3] Group 2 - The Federal Reserve's decision to lower the federal funds rate by 50 basis points to a range of 4.75% to 5.00% marks its first rate cut in four years, reflecting a comprehensive assessment of the U.S. economic situation [4][5] - The Fed's rate cut is expected to lower financing costs for businesses and consumers, stimulating investment and consumption, while also alleviating market tensions and enhancing liquidity [5] - Future projections indicate the federal funds rate may drop to 4.4% by the end of the year, with further reductions anticipated in 2025 and 2026, showcasing confidence in economic recovery and inflation stability [4][5] Group 3 - International gold prices have surged, reaching a historic high of over $3060 per ounce, with a year-to-date increase of over 16%, reflecting strong market demand for gold investments [6][7] - The increase in gold prices has led banks to raise the minimum purchase standards for accumulated gold, indicating a growing trend in gold investment demand [6][7] - Various investment options in the gold market, including bank accumulated gold, gold ETFs, and physical gold, cater to different investor preferences and financial situations [7] Group 4 - Oil prices have shown significant volatility, with international crude oil futures rebounding by 17.3% since March, leading to anticipated price adjustments for gasoline and diesel [8][9] - Oil price fluctuations are influenced by multiple factors, including supply-demand dynamics, geopolitical events, and monetary policy, which can significantly impact market conditions [9][10] - Rising oil prices affect consumer behavior and production costs across various industries, potentially leading to increased prices for goods and services, thereby impacting overall economic growth [10]

机器之心报道 机器之心编辑部 开源社区的人应该对 vLLM 不陌生，它是一个由加州大学伯克利分校团队开发的高性能、开源 LLM 推理和服务引擎，核心目标是提升 LLM 的推理速度（吞吐 量）和资源利用率（尤其是内存），同时兼容 Hugging Face 等流行模型库。 简单来说，vLLM 能让 GPT、Mistral、LLaMA 等主流模型系列跑得更快、消耗更少资源，取得这些效果的关键是其创新的注意力机制实现方案 —— PagedAttention。 近日，DeepSeek AI 研究者、深度学习系统工程师俞星凯 从零开始构建了一个轻量级 vLLM 实现 ——Nano-vLLM，将代码简化到了 1200 行 。 | Inference Engine | Output Tokens | Time (s) | Throughput (tokens/s) | | --- | --- | --- | --- | | vLLM | 133,966 | 98.95 | 1353.86 | | Nano-vLLM | 133,966 | 101.90 | 1314.65 | 作者简介 GitHub 地址：https://g ...

OpenAI 正站在一项雄心勃勃的计划中心，试图重塑人工智能领域的格局，同时加速推动美国在这一新 兴技术领域的领先地位。Stargate 项目是 OpenAI 、甲骨文和软银（SoftBank）的合作成果，计划筹集 高达 5000 亿美元的资金，在美国各地建设大规模人工智能基础设施。 现年 40 岁的 Altman 表示，该项目名称源自早期一个数据中心的设计，其外形类似于在同名电影和电视 剧中广为人知的圆环形传送门。经过数月努力向潜在投资者推介这一理念后，Stargate 于去年开始逐步 成形。在接受采访时，Altman 透露了他对这一 AI 基础设施巨大跃进可能带来的技术、行业乃至全球影 响的看法。 不过，围绕 Stargate 这类项目的一个关键问题是，OpenAI 等公司是否真的需要如此庞大的计算资源。 中国公司 DeepSeek 以及其他致力于推理能力的企业所推出的高效 AI 模型，正在挑战最初激发 Stargate 项目的"规模法则"。 但 Altman 表示，他并未因此受到打击。"目前有很多优秀的模型，我认为优秀模型将会变得非常丰 富，"他解释说，并补充道他相信 OpenAI 的知名产品将拥有优势 ...

Group 1 - The core idea of the article emphasizes the importance of "experience" in achieving AGI, particularly through reinforcement learning (RL) and the accumulation of high-quality data that is not present in human datasets [3][4] - The article discusses the significant advancements in AI's mathematical proof capabilities, highlighting the success of models like DeepMind's AlphaProof and OpenAI's o1 in achieving superhuman performance in mathematical reasoning [3][4] - The transition from static theorem provers to self-planning, self-repairing, and self-knowledge accumulating Proof Engineering Agents is proposed as a necessary evolution in formal mathematics [4][5] Group 2 - The article outlines the challenges faced by contemporary mathematics, likening them to issues in distributed systems, where communication bottlenecks hinder collaborative progress [26][27] - It emphasizes the need for formal methods in mathematics to facilitate better communication and understanding among researchers, thereby accelerating overall mathematical advancement [24][30] - The concept of using formalized mathematics as a centralized knowledge base is introduced, allowing researchers to contribute and extract information more efficiently [30] Group 3 - The DeepSeek Prover series is highlighted as a significant development in the field, with each iteration showing improvements in model scaling and the ability to handle complex mathematical tasks [35][36][38] - The article discusses the role of large language models (LLMs) in enhancing mathematical reasoning and the importance of long-chain reasoning in solving complex problems [41][42] - The integration of LLMs with formal verification processes is seen as a promising direction for future advancements in both mathematics and code verification [32][44] Group 4 - The article suggests that the next phase of generative AI (GenAI) will focus on Certified AI, which emphasizes not only generative capabilities but also quality control over the generated outputs [5] - The potential for multi-agent systems in formal mathematics is explored, where different models can collaborate on complex tasks, enhancing efficiency and accuracy [50][51] - The vision for future agents includes the ability to autonomously propose and validate mathematical strategies, significantly changing how mathematics is conducted [54][58]

Core Insights - The latest AI open-source contribution ranking released by Hugging Face highlights significant achievements by Chinese teams, with Alibaba's Tongyi Qianwen ranking fifth globally and first in China [1] - DeepSeek is also recognized, placing ninth, making it one of the only two non-U.S. institutions in the top ten [1] Summary by Categories Ranking and Performance - Alibaba's Tongyi Qianwen is ranked fifth globally and first in China in the AI open-source contribution ranking [1] - DeepSeek is positioned ninth, showcasing strong performance among global competitors [1] Evaluation Criteria - The ranking is based on an assessment of various factors including open-source models, datasets, developer engagement, community activity, and global influence [1]

火山引擎总裁 谭待 "规模决定一切，最终牌桌上一定有火山。" 文 丨 贺乾明 在字节负责云计算业务火山引擎 5 年，谭待变得更从容了。要来这里时，很多朋友都劝他慎重：字节 比同行晚了 10 年才做云计算，肯定起不来。那是他压力最大的阶段。他说， 现在朋友们谈起火山引 擎 ，会说做得还可以。 火山引擎交出了亮眼的成绩单。据市场调研机构 IDC 统计，去年国内云端大模型调用量，火山引擎一 家占了 46.4%，比二三名加起来都多。去年 12 月到现在，豆包大模型平均每天处理的 Token 处理量 翻了两番到 16.4 万亿，已经逼近国外的科技巨头。 谭待说，除了火山引擎用技术优化压低了大模型价格，豆包 App 也帮了不少忙，"很多客户都会拿着 跟豆包的聊天 Case 找过来，问火山引擎能不能在企业场景中实现类似的效果？" 他判断 Agent 是火山引擎更大的机会：过去的云平台为网页和 App 开发设计，需要程序员理解业务， 定义规则和算法，然后驱动数据做事，上限就是所有人的时间总和。现在 Agent 成为主体，模型可以 自己思考、规划、编排流程，自动执行解决问题，上限远高于 App 时代，"我们正在加码加班做好准 ...