作者：周源/华尔街见闻 7月28日，智谱AI发布旗舰模型GLM-4.5并开源。GLM-4.5是一款专为智能体应用研发的基础模型，在 性能、成本控制与多能力融合等方面均有出色表现。 在这些技术突破的背后，哪些因素起了支撑作用？ 智谱AI核心团队主要来自清华大学KEG（知识工程）实验室：董事长刘德兵、CEO张鹏和总裁王绍兰 均为KEG实验室核心成员，张鹏和王绍兰同为清华创新领军工程博士，首席科学家唐杰曾任清华大学 计算机系教授。 从GLM-1到GLM-4.5经历四年多迭代。 早期（2021年）GLM模型（10B）就已探索了Transformer架构的优化，2022年推出参数规模达130B的 GLM-130B，2023年推出的GLM-3尝试了混合专家（MoE）架构的轻量化设计，为后续参数效率提升奠 定基础，其小步快跑的迭代模式，让团队对模型架构的理解不断深化。 GLM系列的LLM（大语言模型：Large Language Model）基于Transformer架构构建。 GLM-130B采用DeepNorm（一种用于稳定深层Transformer模型训练的归一化方法）作为层归一化 （Layer Normalizat ...

当业界探讨智能体功能、开发环境时，近日北京智谱华章科技股份有限公司（以下简称"智谱"）低调发布新一代旗舰 大模型GLM-4.5，这是一款专为智能体应用打造的基础模型，在复杂推理、代码生成及智能体交互等通用能力上实现 能力融合与技术突破。OpenAI"跳票"多次的GPT-5也强调融合，并在6月底将智谱列入全球竞争对手，没想到智谱率先 登场，GLM-4.5的综合得分位列全球第三、国产第一。 在资本市场，智谱也是"沉默的领跑者"，4月已在北京证监局办理上市辅导备案，由中金公司担任辅导机构，成为第 一家启动IPO上市的"大模型六小虎"。根据辅导备案报告，8月智谱将进入正式辅导期第二阶段，在这期间，这家脱胎 于清华的大模型公司还密集收获多地国资的战略投资。从实验室到产业，智谱走出了中国通向AGI（通用人工智能） 的另一条路径。 全球第三，国产第一 最近的开源浪潮中，智谱的GLM-4.5发布仅2小时，就被X平台推荐上了首页，发布12小时后，它已经位列国际开源社 区Hugging-Face榜单全球第二，创增速纪录。 2024年1月，OpenAI CEO山姆·奥特曼曾在接受媒体采访时提到，他现在的首要任务是推出可能被称为GP ...

图片来源：Getty Images 你可能已在Fortune Intelligence等处读到过相关报道，甚至或许你或你的朋友已受到影响，AI正在深刻 重塑工作模式，尤以招聘、解雇两方面为甚。其冲击在劳动力市场体现得最为明显。作为AI应用的原 爆点，科技行业见证着大批员工被其参与创造的创新技术所取代。企业竞相将AI融入从云基础设施到 客户支持的各个环节，同时缩减软件工程、IT支持和行政职能的人员规模。AI驱动的自动化浪潮不断加 快科技行业的裁员速度，有统计显示，受影响员工数量已高达8万人。仅微软一家公司就裁减了1.5万个 岗位，该公司同时承诺向新AI项目再投入800亿美元。 不过劳动力市场情报公司Lightcast提供的数据为未来带来了一线希望。那些要求具备AI技能的非技术岗 位的薪资水平正在大幅飙升。Lightcast发布的新报告《超越喧嚣》（Beyond the Buzz）对超过13亿份招 聘信息进行了分析，结果显示，此类职位的薪资平均增加了28%——相当于每年多赚近1.8万美元。该 研究凸显了技术与非技术领域招聘的分化：技术类岗位对AI技能的需求依然强劲，但IT与计算机科学领 域中的AI职位占比已从20 ...

近年来，人工智能技术迎来革命性突破，以ChatGPT、DeepSeek等为代表的大语言模型展现出惊人的理 解、推理和创造能力，正在重塑人类社会的知识生产方式和价值创造模式。与此同时，新一代青年成长 于数字时代，他们的学习方式、思维模式都不可避免地与AI交织在一起。 在此背景下，本集《两说》以"新一代青年与新一代人工智能"为主题，特邀复旦大学国际关系与公共事 务学院教授高奇琦、上海交通大学国际与公共事务学院长聘副教授贾开展开对话。他们将讨论这些话 题：青年一代如何认知和使用人工智能？教育体系又该如何调整帮助青年一代适应和引领这场技术变 革？面向未来，人工智能会进一步压制青年一代的机会，还是帮助他们更好地释放潜力？ 01 人工智能怎么用？ 今年大模型评测高考成绩理科达985水平，文科甚至可冲击顶尖高校，上海交通大学国际与公共事务学 院长聘副教授贾开指出从技术到应用，人工智能已进入全新节点。 人工智能逐渐嵌入到我们生活中，正确认知和使用才能发挥它的作用。全国高校纷纷开设AI通识课， 贾开向复旦大学国际关系与公共事务学院教授高奇琦问道："学生用了AI后，还愿意思考吗？"高奇琦提 出，如果人工智能仅仅是用来完成作业就会 ...

Core Insights - The WAIC showcased significant advancements in AI and robotics, with over 350,000 attendees participating in the event [1] - Major tech companies shifted focus from basic large models to multi-modal applications and AI agents, indicating a competitive landscape [1][3] - The emergence of AI agents as a primary focus for companies, with various solutions being demonstrated across different sectors [6][7] Group 1: Event Overview - WAIC attracted approximately 350,000 attendees, highlighting the growing interest in AI and robotics [1] - The event featured over 800 exhibitors showcasing advancements in AI infrastructure, robotics, and multi-modal applications [1][2] - The shift from traditional humanoid robots to more interactive and functional robots was evident, with live demonstrations of various tasks [10][11] Group 2: Company Highlights - Alibaba's booth was the largest, featuring the Quark AI glasses and multiple open-source large models, emphasizing their commitment to AI agents [3][6] - Ant Group presented various AI solutions, including the financial reasoning model Agentar-Fin-R1, showcasing their focus on industry-specific applications [6][7] - The "Six Little Tigers" of large models showed a divergence in performance, with some companies like Baichuan Intelligence and Zero One falling behind [7][8] Group 3: Technological Developments - The AI agents market has surpassed $5 billion, with a growth rate of 40%, indicating a strong demand for practical applications [4] - Companies are increasingly focusing on integrating AI models with real-world business needs, as seen in the development of solutions for document proofreading and financial services [5][6] - The introduction of advanced components like six-dimensional force sensors is enhancing the capabilities of humanoid robots, allowing them to perform complex tasks autonomously [12][14] Group 4: Market Trends - The trend is shifting from "technology showcase" to "scene rehearsal," with a focus on practical applications of AI technology [14] - The competition is intensifying as companies strive to effectively integrate technology into products, moving beyond mere demonstrations [14] - The rapid growth in the humanoid robot market is creating both challenges and opportunities for component manufacturers, necessitating faster development cycles and higher standards [13][14]

机器之心报道 机器之心编辑部 在这届 ACL 大会上，华人团队收获颇丰。 ACL 是计算语言学和自然语言处理领域的顶级国际会议，由国际计算语言学协会组织，每年举办一次。一直以来，ACL 在 NLP 领域的学术影响力都位列第一，它 也是 CCF-A 类推荐会议。今年的 ACL 大会已是第 63 届，于 2025 年 7 月 27 日至 8 月 1 日在奥地利维也纳举行。 今年总投稿数创历史之最，高达 8000 多篇（去年为 4407 篇），分为主会论文和 Findings，二者的接收率分别为 20.3% 和 16.7%。 根据官方数据分析，在所有论文的第一作者中，超过半数作者来自中国（51.3%），而去年不到三成（30.6%）。紧随中国，美国作者的数量排名第二，但只占 14.0%。 今年共评选出 4 篇最佳论文，2 篇最佳社会影响力论文、3 篇最佳资源论文、3 篇最佳主题论文、26 篇杰出论文，2 篇 TACL 最佳论文、1 篇最佳 Demo 论文以及 47 篇 SAC Highlights。 以下是具体的获奖信息。 最佳论文奖 在本届4篇最佳论文中，DeepSeek（梁文锋参与撰写）团队以及北大杨耀东团队摘得 ...

Core Viewpoint - The emergence of DeepSeek-R1, an open-source large language model (LLM) developed by a Chinese startup, has revolutionized the deployment of AI in hospitals, significantly enhancing efficiency and reducing costs compared to existing models like ChatGPT [2][12]. Group 1: Deployment and Impact - DeepSeek-R1 was released in January 2025 and quickly became the most downloaded chatbot in the US Apple App Store, surpassing OpenAI's ChatGPT [2]. - As of May 8, 2025, DeepSeek-R1 has been deployed in over 755 hospitals across China, including top-tier hospitals and grassroots medical institutions, with more than 500 achieving local deployment [5][8]. - The model is capable of various tasks, including clinical services, hospital operations, and personal health management, providing significant support in diagnosis, treatment recommendations, and administrative tasks [13][21]. Group 2: Advantages of DeepSeek-R1 - The model's deployment cost is significantly lower than traditional AI systems, with a complete local deployment costing under $100,000, making it accessible for many smaller hospitals [21]. - DeepSeek-R1's advanced reasoning capabilities are comparable to top international models, essential for handling complex medical tasks [22]. - The open-source nature allows hospitals to customize and integrate the model into existing systems, enhancing its utility [22]. Group 3: Regulatory Challenges - The rapid deployment of DeepSeek-R1 has highlighted a regulatory "gray area," raising concerns about patient safety and the need for a robust regulatory framework [6][10]. - The lack of clear classification standards for AI applications in healthcare leads to ambiguity regarding which applications are considered high-risk [32]. - The current regulatory environment does not adequately address the unique challenges posed by large language models, necessitating immediate reforms [35]. Group 4: Recommendations for Regulation - The article calls for a risk-based classification system for AI applications in healthcare, distinguishing between high-risk and low-risk applications [35]. - High-risk applications should be regulated as medical devices, requiring stringent approval and monitoring processes [35]. - Continuous monitoring and evaluation of AI applications in real-world settings are essential to ensure safety and effectiveness [38].

Summary of Key Points from the Conference Call Industry Overview - The conference call discusses the **artificial intelligence (AI)** industry, particularly focusing on the development and investment trends in large language models (LLMs) and deep learning technologies [1][2][3]. Core Insights and Arguments - **Investment Waves**: AI investment has experienced three significant waves over the past three years, with the latest wave showing longer duration, stronger momentum, and higher capital expenditure compared to previous waves [1][2][4]. - **Technological Advancements**: The introduction of deep learning and reinforcement learning has significantly enhanced the capabilities of LLMs, allowing them to perform complex tasks with improved logic and reasoning abilities [1][8][9]. - **Model Performance**: OpenAI's upcoming models, such as GPT-5, are expected to achieve generational improvements in logic processing and dynamic handling, while models like GROX and Google's Gemini series are noted for their impressive performance and balanced capabilities [10][12][14]. - **Cost of Model Training**: The cost of training models has been decreasing annually due to advancements in chip technology and training methodologies, which enhances training efficiency [22][23]. - **Market Dynamics**: The AI API market is competitive, with Google holding approximately 45% market share, followed by Sora and Deepseek. Domestic models like Kimi K2 are also gaining traction [30]. Additional Important Content - **Challenges in Deep Learning**: Deep reasoning models face challenges such as slow response times for simple queries, which impacts user experience. Future developments may focus on hybrid reasoning to improve performance [16]. - **Future Training Paradigms**: The evolution of training paradigms for LLMs will emphasize increased reinforcement learning time and the integration of high-quality data during training phases [17]. - **Domestic vs. International Models**: There is a noticeable gap of about 3 to 6 months between domestic and international models, but this gap is not expected to widen significantly. Domestic models are making strides in areas like programming capabilities [18][20]. - **User Interaction and Growth Potential**: AI technology has seen significant user penetration, particularly in Google Search, with potential for further growth as new applications are developed [27][28]. - **AGI Development**: Progress towards Artificial General Intelligence (AGI) is ongoing, with no major technical barriers identified. The integration of AI across various applications is enhancing overall efficiency [31]. This summary encapsulates the key points discussed in the conference call, highlighting the current state and future outlook of the AI industry, particularly in relation to large language models and their market dynamics.

Core Viewpoint - The article discusses Geoffrey Hinton's insights on the relationship between AI and human intelligence, emphasizing the evolution of AI from symbolic reasoning to large language models (LLMs) and the implications of AI surpassing human intelligence [1][10]. Group 1: Evolution of AI Understanding - For over 60 years, there have been two distinct paradigms in AI: the logical inference paradigm, which views intelligence as symbolic reasoning, and the biological paradigm, which sees intelligence as rooted in understanding and learning through neural networks [1]. - In 1985, Hinton created a small model to explore how humans understand vocabulary by linking features of words to predict the next word without storing entire sentences [2]. - The development of LLMs is seen as a continuation of Hinton's early work, processing more input words and utilizing complex neural structures to build richer interactions [3]. Group 2: Mechanism of Language Understanding - LLMs and human language understanding mechanisms are highly similar, transforming language into features and integrating these features across neural network layers for semantic understanding [4]. - Each word in language is likened to a multi-dimensional Lego block, which can flexibly combine to form complex semantic structures, with the shape of words adapting based on context [6]. - Understanding a sentence is compared to deconstructing a protein molecule rather than converting it into a clear, unambiguous logical expression [5]. Group 3: Knowledge Transfer in AI - The human brain operates at 300,000 watts but cannot easily transfer knowledge to another person, relying instead on explanation [11]. - In contrast, digital intelligence allows for efficient knowledge transfer, directly copying parameters and structures without intermediary language, sharing trillions of bits of information during synchronization [13][14]. - Current technology enables the same model to be deployed across different hardware, facilitating efficient knowledge migration and collaborative learning [15]. Group 4: The Dangers of Advanced AI - There is a concern that AI could surpass human intelligence, leading to scenarios where AI becomes an active system with its own goals, potentially manipulating humans [18][19]. - Hinton warns that developing AI is akin to raising a tiger; once it grows powerful, losing control could be fatal [20]. - Despite the risks, AI holds significant value in various fields, and eliminating it is not feasible; instead, a method must be found to ensure AI does not threaten humanity [21]. Group 5: Global Cooperation for AI Safety - No single country desires AI to dominate the world, and if one country discovers a method to prevent AI from going rogue, others will likely follow suit [22][23]. - Hinton proposes the establishment of an international AI safety organization to research technology and create standards to ensure AI develops positively [24]. - The long-term challenge is to ensure that AI remains a supportive tool for humanity rather than a ruler, which is a critical issue for global collaboration [25].

Investment Rating - The investment rating for the company is "Accumulate" (maintained) [5] Core Views - The company's revenue in H1 2025 reached 458 million yuan, representing a year-on-year increase of 69%. The growth was driven by the high increase in intelligent computing cloud services, which saw a 175% year-on-year growth [6][9] - The company has signed a framework cooperation agreement with Alibaba Cloud to enhance AI technology accessibility through the integration of the GLM-Z1 series inference models into its MaaS platform [6][7] - The "Parallel Computing Network" is expected to support the national "East Data West Computing" strategy, with the intelligent computing scale projected to reach 725.3 EFLOPS in 2024, a year-on-year increase of 74.1% [7] Financial Performance Summary - In H1 2025, the company achieved a net profit of 5.08 million yuan, a year-on-year increase of 20%, and a net cash flow from operating activities of 39.26 million yuan, up 323% year-on-year [6] - Revenue projections for 2025 are estimated at 863 million yuan, with a year-on-year growth rate of 31.86% [8] - The company is expected to achieve net profits of 24 million yuan in 2025, with corresponding EPS of 0.40 yuan per share [9]