一水 发自 凹非寺 量子位 | 公众号 QbitAI 吵起来了。 图灵奖得主和诺贝尔奖得主，为了"智能的本质"——直接激情友好地交流上了。 AI三巨头之一、图灵奖得主Yann LeCun明确表示： 纯粹就是胡扯（complete BS）。 而诺贝尔奖得主、谷歌DeepMind CEO哈萨比斯也不留情面了，指名道姓回击： LeCun的说法简直是大错特错。 当然，马斯克的站队可能也有别的原因。毕竟他和LeCun素来不是很对付，跟哈萨比斯则亦师亦友——马斯克还是哈萨比斯DeepMind早期投 资人。 论战之激烈，关注度之高，已经让专门开辟了一个话题板块： 马斯克也跑来吃瓜了—— 没有任何多余的解释，但这波他站哈萨比斯——"Demis is right"。 事情还要从LeCun几天前接受的一场采访说起。 他在节目中言辞犀利地指出： 根本不存在所谓的"通用智能"，纯粹就是胡扯（complete BS） 。 这个概念毫无意义，因为它实际上是用来指代人类水平的智能，但人类智能其实是高度专业化的。我们在现实世界里确实干得不错，比 如认个路、导航blabla；也特别擅长跟人打交道，因为咱们进化了这么多年就是干这个的。 但在国际 ...

吵起来了。 图灵奖得主和诺贝尔奖得主，为了"智能的本质"——直接激情友好地交流上了。 AI三巨头之一、图灵奖得主Yann LeCun明确表示： 纯粹就是胡扯（complete BS）。 而诺贝尔奖得主、谷歌DeepMind CEO哈萨比斯也不留情面了，指名道姓回击： LeCun的说法简直是大错特错。 论战之激烈，关注度之高，已经让专门开辟了一个话题板块： 马斯克也跑来吃瓜了—— 没有任何多余的解释，但这波他站哈萨比斯——"Demis is right"。 当然，马斯克的站队可能也有别的原因。毕竟他和LeCun素来不是很对付，跟哈萨比斯则亦师亦友——马斯克还是哈萨比斯DeepMind早期投资人。 要科学吃瓜，可能还是要知道他们究竟在激辩什么？ 争论焦点：智能的本质是什么？ 事情还要从LeCun几天前接受的一场采访说起。 他在节目中言辞犀利地指出： 他这是把"general intelligence"和"universal intelligence"两个概念搞混了。 根本不存在所谓的"通用智能"，纯粹就是胡扯（complete BS）。 这个概念毫无意义，因为它实际上是用来指代人类水平的智能，但人类智能其实是高度 ...

一水 发自 凹非寺 量子位 | 公众号 QbitAI 吵起来了。 图灵奖得主和诺贝尔奖得主，为了"智能的本质"——直接激情友好地交流上了。 AI三巨头之一、图灵奖得主Yann LeCun明确表示： 纯粹就是胡扯（complete BS）。 而诺贝尔奖得主、谷歌DeepMind CEO哈萨比斯也不留情面了，指名道姓回击： 马斯克也跑来吃瓜了—— 没有任何多余的解释，但这波他站哈萨比斯——"Demis is right"。 LeCun的说法简直是大错特错。 论战之激烈，关注度之高，已经让专门开辟了一个话题板块： 当然，马斯克的站队可能也有别的原因。毕竟他和LeCun素来不是很对付，跟哈萨比斯则亦师亦友——马斯克还是哈萨比斯DeepMind早期投 资人。 要科学吃瓜，可能还是要知道他们究竟在激辩什么？ 争论焦点：智能的本质是什么？ 事情还要从LeCun几天前接受的一场采访说起。 他在节目中言辞犀利地指出： 根本不存在所谓的"通用智能"，纯粹就是胡扯（complete BS） 。 然而，这一观点很快遭到了哈萨比斯的直接回怼。哈萨比斯表示： LeCun的说法简直是大错特错。 他这是把"general intellige ...

鹭羽 发自 凹非寺 量子位 | 公众号 QbitAI 如何回顾2025年的AI进展？ 今年王者归来的谷歌，刚刚由 Jeff Dean 和 哈萨比斯 牵头，完成了年度总结和趋势展望报告—— 这是AI Agent、推理和科学发现的一年。 2025年创造性AI工具全面提升。 AI在科学和数学领域实现多项成果，尤其是数学和编程竞赛表现优异。 谷歌始终强调安全、责任与合作开放生态。 AI已广泛融入谷歌的主要产品中。 报告最后梳理出了 八大研究方向 ，系统性地回答了一个更重要的问题： 当大模型进入推理时代，AI 正在变成什么？ 从Gemini的推理能力、多模态理解，到Agent、机器人、科学研究和物理世界建模，谷歌正在描绘一个可以协作、可以行动、甚至开始参与科 学发现的智能系统。 模型的推理、多模态理解、生成能力及效率得到显著提升。 以下是量子位的原文总结，在不改变原意的基础上，进行了适当修改润色： 谷歌年度回顾全文 回顾2025年， 这是研究领域取得非凡进展的一年 。 在人工智能方面，可以清晰地看到其发展轨迹正从一种工具转变为一种实用手段：从人们使用的东西变成了 可以投入工作使用 的东西。 如果说2024年是为这个时代 ...

炒股就看金麒麟分析师研报，权威，专业，及时，全面，助您挖掘潜力主题机会！ 来源：Barrons巴伦 最新一期'赵何娟Talk'里，李飞飞教授认为，从"语言生成"到"世界生成"，空间智能将在两年内迎来应 用级爆发——但AI永远只是工具，选择权应该始终在我们人类手里。 作者|赵何娟 一切进展都已经比一年前大家的预期要快了很多。李飞飞也在钛媒体这期'赵何娟Talk'里对话里透露， 从"语言生成"到"世界生成"，空间智能将在两年内迎来应用级爆发。 随着2025年渐入尾声，有着"AI教母"之称的斯坦福大学教授李飞飞，带着她创立的World labs迎来了一 波又一波的新进展，包括首款商用"世界模型"Marble的发布，这开始让大家终于意识到，原来"世界模 型"并非只是概念，而已经是真实有用的。 回想我第一次见李飞飞教授，已经可以追溯到2017年，在斯坦福大学教学楼内。那一年，刚刚定居硅谷 的陈天桥先生向我和其他几位老朋友介绍了李飞飞教授，他当时特别提到：这是美国最杰出的华人科学 家之一。那时，李飞飞教授发起的ImageNet行动还在如火如荼的进行。我也第一次在与飞飞教授的见面 和交流中学到了一个新的概念：为什么是数据集 ...

图片来源：AXIOS Z Highlight Demis Hassabis是Google DeepMind的联合创始人兼CEO、Demis Hassabis博士。他是一位神经科学家、企业家，也是AI领域的先驱。Demis五岁时就是国际 象棋神童，48岁获得诺贝尔奖。本次是Axios AI+SF Summit上的访谈，发布于2025年12月5日，他与主持人探讨了他对于AI、AGI在当下以及未来的理解 与洞察。 科学家与诺奖得主：从科学方法到公共责任 Mike Allen：我是Axios的联合创始人Mike Allen，我代表我的联合创始人Roy Schwartz和Jim VandeHei，感谢在座各位。九年来，感谢你们一路支持 Axios，也感谢大家来到旧金山，在这座历史悠久、氛围极佳的银行建筑中，参加Axios AI+SF Summit。也欢迎世界各地的朋友们，一起见证今天的重磅 对话。Demis Hassabis，欢迎来到Axios。 Demis Hassabis：感谢邀请。我们一直很期待这次交流，很高兴能来到这里。 Mike Allen：就在400多天前，你得知自己获得诺贝尔奖。当时你说，"这太不真实了， ...

Core Insights - The conversation highlights the transformative potential of AGI (Artificial General Intelligence) and the need for societal readiness for its arrival, which is estimated to be within five to ten years [6][30] - Demis Hassabis emphasizes the importance of responsible AI usage and the need for ongoing discussions about AI safety and societal impacts [8][15] - The dialogue also touches on the competitive landscape of AI, particularly the race between the US and China, with the US currently holding a slight edge in algorithmic innovation [21][22] Group 1: AGI and Its Implications - AGI is seen as one of the most transformative moments in human history, requiring careful preparation from governments and leaders [6][8] - Current AI systems lack critical capabilities such as continuous learning and reasoning, which are essential for achieving true AGI [31] - The timeline for achieving AGI is projected to be five to ten years, contingent on one or two significant breakthroughs [30][31] Group 2: AI Safety and Responsibility - There is a strong emphasis on the responsible use of AI, focusing on what AI can improve and accelerate while maintaining caution in its deployment [8][15] - The potential risks of AI misuse by malicious actors and the possibility of AI systems becoming uncontrollable are significant concerns [15][20] - The need for robust AI safety measures is underscored, especially as AI systems become more autonomous [20][19] Group 3: Competitive Landscape - The US and Western countries are currently leading in AI, but the gap with China is narrowing, with Chinese models showing impressive capabilities [21][22] - The competition for AI talent is intensifying, with companies needing to attract mission-driven individuals to stay at the forefront of innovation [33] - The importance of algorithmic innovation is highlighted, with the US still holding an advantage in this area despite China's rapid advancements [22] Group 4: Technological Advancements - The integration of multimodal capabilities in AI, such as the ability to process and generate text, images, and videos, is a key focus for future developments [11][12] - The introduction of systems like Gemini 3 showcases significant advancements in reasoning depth and the ability to generate nuanced outputs [25][27] - The potential for AI to assist in various domains, including sports analytics, is also discussed, indicating its broad applicability [37][38]

Core Insights - The report emphasizes the synergy between AI innovation and capital expenditure (CAPEX), highlighting that model innovation is the core driver of AI development, with CAPEX serving as the foundation for the AI cycle [12][14] - The AI industry chain includes AI hardware, CAPEX, and AI models and applications, which collectively support the computational needs for large model training and inference [12][14] - The report suggests that the AI storage cycle is driven by rising prices and simultaneous expansion and upgrades in production capacity, particularly in cloud and edge storage [4][34] Group 1: AI Innovation and CAPEX - Model innovation is identified as the key driver of AI development, with significant capital expenditures from cloud service providers and leading enterprises providing a stable cash flow to support upstream hardware sectors [14][24] - The report notes that major companies like Google and OpenAI are making substantial advancements in multi-modal models, which are expected to enhance user engagement and monetization opportunities [19][25] - The integration of AI capabilities into various applications is projected to create a closed loop of high computational demand leading to high-value content and increased user willingness to pay [24][25] Group 2: Storage Cycle - The report indicates that storage prices are on the rise, significantly boosting the gross margins of original manufacturers, with capital expenditures in the storage sector entering an upward phase [4][34] - It highlights that traditional DRAM and NAND production is being approached cautiously, while HBM production is prioritized, indicating a shift in focus within the storage industry [4][34] - The report discusses the emergence of new opportunities in the storage foundry model, driven by the evolving demands of AI applications [4][34] Group 3: Investment Recommendations - The report recommends focusing on companies within the AI ecosystem, particularly those involved in AI storage, PCB, and power supply sectors, as they are expected to experience sustained growth [4][34] - It suggests that the ongoing upgrades in DRAM and NAND architectures will create new equipment demand, presenting investment opportunities in related companies [4][34] - The report encourages attention to the storage industry chain, particularly in light of the anticipated price increases and margin improvements for original manufacturers [4][34]

Core Insights - The article emphasizes the importance of Scaling Laws in achieving Artificial General Intelligence (AGI) and highlights Google's success with its Gemini 3 model as a validation of this approach [5][19][21]. Group 1: Scaling Laws and AGI - Scaling Laws were initially discovered by DeepMind, not OpenAI, and have been pivotal in guiding research directions in AI [12][14][18]. - Google DeepMind believes that Scaling Laws are essential for the development of AGI, suggesting that significant data and computational resources are necessary for achieving human-like intelligence [23][24]. - The potential for Scaling Laws to remain relevant for the next 500 years is debated, with some experts expressing skepticism about its long-term viability [10][11]. Group 2: Future AI Developments - In the next 12 months, AI is expected to advance significantly, particularly in areas such as complete multimodal integration, which allows seamless processing of various data types [27][28][30]. - Breakthroughs in visual intelligence are anticipated, exemplified by Google's Nano Banana Pro, which demonstrates advanced visual understanding [31][32]. - The proliferation of world models is a key focus, with notable projects like Genie 3 enabling interactive video generation [35][36]. - Improvements in the reliability of agent systems are expected, with agents becoming more capable of completing assigned tasks [38][39]. Group 3: Gemini 3 and Its Capabilities - Gemini 3 aims to be a universal assistant, showcasing personalized depth in responses and the ability to generate commercial-grade games quickly [41][44][45]. - The architecture of Gemini 3 allows it to understand high-level instructions and produce detailed outputs, indicating a significant leap in intelligence and practicality [46]. - The frequency of Gemini's use is projected to become as common as smartphone usage, integrating seamlessly into daily life [47].

Core Insights - The current era of artificial intelligence, dominated by large language models and image classifiers, has reached its limits, and AI with spatial intelligence is seen as the next frontier to break through this bottleneck [2][11][24] Group 1: AI Limitations - AI is categorized into two types: speaking intelligence and doing intelligence, with the former being strong in text output but often failing in practical tasks [6][11] - Examples of AI failures include generating unrealistic images and videos, highlighting the lack of common sense and physical understanding in current models [7][10] Group 2: Spatial Intelligence - Spatial intelligence, a concept originating from educational psychology, involves the perception, understanding, and manipulation of spatial information, which is crucial for human development and creativity [12][15] - Current AI systems lack deep, common-sense understanding of the physical world, which directly affects the quality of their outputs [11][17] Group 3: World Models - The concept of world models, inspired by human cognitive abilities, is emerging as a key area of focus for AI development, aiming to enable machines to understand and interact with the physical world [19][23] - Recent advancements in world models include new products and technologies from companies like NVIDIA and Google DeepMind, indicating a growing interest and investment in this area [22][23] Group 4: Future Challenges - Building AI that can operate like humans presents significant challenges, including the complexity and uncertainty of the real world, limitations in existing data, and the inherent constraints of physical laws [23][24]