Workflow
GLM 4.6
icon
Search documents
AGI为什么不会到来?这位研究员把AI的“物理极限”讲透了
3 6 Ke· 2025-12-17 11:43
这意味着,智能的提升并不是"想象空间"问题,而是绕不开能量、带宽、存储、制造和成本的物理限 制。 AGI 会不会到来? 这是AI 行业里反复被讨论、却一直始终缺乏清晰论证的问题。 最近,西雅图艾伦人工智能研究所(AI2)的研究员蒂姆·德特默斯(Tim Dettmers)在一篇文章,题目很 直接——《为什么 AGI 不会实现?》。 蒂姆·德特默斯 在这篇文章中,他提出了一个被长期忽视、却至关重要的前提: 计算并不是抽象概念,而是一件彻底受物理规律约束的事情。 德特默斯认为,当下市场对AGI 的判断普遍偏乐观,一个关键原因在于: 很多讨论只停留在模型、参数和算法层面,却忽视了支撑这些能力的物理基础正在逼近极限。 在文章中,德特默斯第一次从物理约束的角度,系统性地解释了为什么AGI 面临一系列难以回避的现 实。这些判断,也有助于我们更好地理解当前的AI行业。 他在文章中总结了几条关键判断: 1)Transformer 的成功并非偶然,而是在当前物理约束下接近最优的工程选择,继续通过架构改进获得 的边际收益正在快速下降。 2)当下大量所谓"创新",本质仍是既有框架上的渐进改进,很难带来结构性跃迁。 3)AI 过去的 ...
Zai GLM 4.6: What We Learned From 100 Million Open Source Downloads — Yuxuan Zhang, Z.ai
AI Engineer· 2025-11-20 14:14
GLM 4.6 is the only open-source model currently tied for #1 on the LMSYS Chatbot Arena, standing shoulder-to-shoulder with GPT-4o and Claude 3.5 Sonnet. In this talk, Zhang Yuxuan from zAI breaks down the technical roadmap that led to over 100 million downloads across the GLM family. Zhang deep dives into the specific training recipes behind GLM 4.6, including their move to single-stage Reinforcement Learning (RL), the "SLIME" RL framework for handling complex agent trajectories, and how they structured 15 ...
计算机周报20251116:叙事的逆转:中美大模型差距是否在拉大?-20251116
Minsheng Securities· 2025-11-16 14:02
Investment Rating - The report maintains a "Recommended" investment rating for the industry [5]. Core Insights - The gap between domestic and overseas large models in AI is rapidly narrowing, with domestic AI ecosystems represented by Tencent and Alibaba showing significant development. This suggests a potential turning point for accelerated growth in domestic AI [3][22]. - The report emphasizes the importance of focusing on core targets in domestic computing power and AI agents, highlighting key companies in cloud computing, chip design, and AI applications [3][22]. Summary by Sections Market Review - During the week of November 10-14, the CSI 300 index fell by 1.08%, the SME index decreased by 1.71%, and the ChiNext index dropped by 3.01%. The computer sector (CITIC) saw a decline of 3.72% [30]. Industry News - AMD's CEO predicts that the AI data center market will exceed $1 trillion by 2030, growing from approximately $200 billion currently, with a compound annual growth rate (CAGR) of over 40% [23]. - The Ministry of Industry and Information Technology has issued a notice to accelerate the construction of pilot platforms in the manufacturing sector, aiming to enhance innovation and technology transfer [24]. Company News - Lingzhi Software plans to acquire 100% of Kaimiride (Suzhou) Information Technology Co., Ltd. through a share issuance and cash payment, with a share price set at 15.31 yuan [27]. - Zhengyuan Wisdom's board approved a share repurchase plan, intending to reduce up to 2,842,000 shares within six months [29]. Weekly Insights - Domestic large models like MiniMax and DeepSeek are now among the top global models, with MiniMax M2 achieving a daily token usage surpassing 50 billion, indicating strong market acceptance [9][12]. - The report highlights the competitive landscape in AI, with Tencent and Alibaba intensifying their efforts in AI applications, suggesting an imminent phase of heightened competition in the domestic AI market [20][22].
最新外国「自研」大模型,都是套壳国产?
3 6 Ke· 2025-11-01 05:02
Core Insights - The article discusses the emergence of Chinese open-source AI models as significant players in the global AI landscape, particularly in light of recent developments from American tech companies [4][21][26] Group 1: New Developments in AI Models - Cursor has released a major update, introducing its own code model, Composer, which utilizes reinforcement learning and is capable of processing code efficiently [4][7] - The Composer model reportedly generates code four times faster than similar models, indicating a significant advancement in performance [7] - Speculation arises regarding the underlying technology of these models, with suggestions that they may be based on Chinese AI models, particularly the GLM series [9][11][16] Group 2: Industry Reactions and Analysis - Industry experts suggest that many new models, including Cursor's Composer, are fine-tuned versions of existing Chinese models rather than entirely new creations, highlighting the high costs associated with developing foundational models from scratch [17][18] - The success of open-source models is emphasized, with Nvidia's CEO noting their role in accelerating AI applications and the need for developers to leverage these resources [21][23] - The article points out that the leading open-source models in the HuggingFace community predominantly originate from Chinese companies, showcasing their growing influence [23][26] Group 3: Implications for Global AI Competition - The advancements in Chinese open-source models are reshaping the competitive landscape of AI, with a shift in positions between leaders and followers in the technology race [26] - The article concludes that the capabilities of Chinese models are now sufficient to support the development of Western products, indicating a new era of multipolar competition in AI [20][26]
最新外国「自研」大模型,都是套壳国产?
机器之心· 2025-11-01 04:22
Core Insights - The article discusses the emergence of Chinese open-source AI models as significant players in the global AI landscape, suggesting that foreign developers may need to start learning Chinese due to the influence of these models [1][29]. Group 1: New Model Releases - Cursor has released a major update to its AI code tool, introducing its own code model called Composer, which utilizes a new interface for collaborative work among multiple intelligent agents [5]. - The Composer model, trained using reinforcement learning, is a large MoE model that excels in handling actual code and operates at a speed four times faster than similar models [6][8]. - Cognition has also launched its latest AI model, SWE-1.5, which boasts a parameter count in the hundreds of billions and significantly enhances speed, outperforming Haiku 4.5 by 6 times and Sonnet 4.5 by 13 times [9]. Group 2: Model Development and Origins - There are speculations that both Cursor's Composer and Cognition's SWE-1.5 models are based on Chinese AI models, with evidence suggesting that Cognition's model is customized from Zhiyu's GLM 4.6 model [14][21]. - The release of these models has sparked discussions about the reliance on Chinese open-source models, with industry experts indicating that many new models are fine-tuned rather than built from scratch due to the high costs associated with training foundational models [24][25]. Group 3: Market Trends and Implications - The article highlights the growing dominance of Chinese open-source models in the AI sector, with significant market share held by models like Alibaba's Qwen, which has been leading in downloads and usage since 2025 [30][32]. - The increasing capabilities of these models are not only aiding developers but are also becoming essential for startups, indicating a shift in the competitive landscape of global AI [32][35]. - The article concludes that the positions of followers and leaders in the AI model technology race are gradually changing, with Chinese models establishing a leading status [36].