点击下方 卡片 ，关注" 自动驾驶之心 "公众号 戳我-> 领取 自动驾驶近30个 方向 学习 路线 最近和业内专家jason老师讨论了很多，分享一个最近被问到很多的问题： 世界模型是不是端到端？ 答案是明确的：不是。 其实世界模型和端到端都不指某个具体的技术，而是一类具备某些特定能力的模型。 端到端自动驾驶可以这么定义：没有显示的信息处理与决策逻辑，一端接受信息输入，另一端输出决策结果的模型。 世界模型使用类似的定义：它接受信息输入，内在建立起对整个世界/环境的完整认知，能够重建、预测未来变化的模型。 所以世界模型是一种实现端到端自动驾驶的途径。 先前平台打造的《端到端与VLA自动驾驶小班课》备受大家好评，因此我们进一步推出这门世界模型小班课， 课程聚焦于通用世界模型、视频生成、OCC生成等 世界模型算法，涵盖特斯拉世界模型、李飞飞团队Marble等。欢迎大家加入学习~ 早鸟优惠！开课即止~ 讲师介绍 Jason：C9本科+QS50 PhD，已发表CCF-A论文2篇，CCF-B论文若干。现任国内TOP主机厂算法专家，目前从事端到端、大模型、世界模型等前沿算法的预研和量 产，并已主持和完成多项自动驾驶感知和端 ...

Industry Insights - The Robo-X initiative is expected to reach a milestone in 2026, driven by supportive policies, technological advancements, and cost reductions in L4 autonomous driving [3][4] - The global L4 market is projected to exceed trillions by 2030, with the domestic Robotaxi market estimated at 236 billion yuan annually, and Robovan and Robotruck markets also showing significant potential [4][12] - The competitive landscape includes key players such as Pony.ai and WeRide in the Robotaxi sector, with various companies emerging in Robovan, Robotruck, Robobus, and Robosweeper markets [4] Company Analysis - Pony.ai reported a 72% year-on-year revenue growth in Q3, with ongoing progress in the commercialization of Robotaxi services [1][2] - WeRide achieved a remarkable 144% year-on-year revenue growth in Q3, indicating accelerated commercialization of its L4 products [2][1] Policy Developments - Global policies are increasingly supportive of autonomous driving, with countries like the UAE and Singapore implementing frameworks to facilitate the testing and deployment of autonomous vehicles [12][14] - In China, the Ministry of Industry and Information Technology has initiated pilot programs for smart connected vehicles, involving major automotive companies [14][15] Investment Trends - In 2025, the L4 sector is expected to attract significant investment, with over 49 financing events reported, totaling nearly 21.8 billion yuan in funding [16]

新华财经上海12月17日电（记者杜康、龚雯）在日前举办的2025全球开发者先锋大会上，机器人在插 花、搬运、救灾等真实场景中"各显神通"。冷冰冰的技术参数，在这里化作了鲜活的技能比拼。当然， 大赛也暴露了具身智能"笨拙"的一面：在叠衣服、拧螺丝等精细操作背后，不少机器人仍连着"遥操 作"的手柄。 恰恰是在这"能"与"不能"的缝隙中，公众得以窥见这一火热领域的技术边界与未来方向。 从机器人的"能"里看技术进阶 回望过去一年，中国具身智能领域"快步疾行"：智元远征A2人形机器人完成无间断百公里跨省行走， 充分证明了机器人能够"走得稳"；行业商业化"大单"频现，机器人真正进入工厂，负责分拣、上下料； VLA（视觉-语言-动作）模型的进化，则让机器人大脑更聪明，能够听懂人的需求。 在2025全球开发者先锋大会上，观众再一次真切看到了机器人的"能"。 更棘手的是环境干扰。"光照变化、桌子周边物体的摆放、强光下周边物体在桌子上的倒影等，都有可 能让机器人'智商下线'，操作不准。这种难以将目标与'背景噪音'剥离的困境，折射出当下具身智能在 物理场景理解能力上的短板——泛化性不足。"参赛队员对记者表示。 ——拧螺丝等精细活儿 ...

Core Insights - The article posits that 2026 will be a pivotal year for multimodal technology, particularly in video generation and world models, with significant advancements expected in both research and practical applications [2][3]. Group 1: Definition and Importance of World Models - Various definitions of world models exist, including comparisons to human brain representations and neural networks that understand physical rules [4][5]. - World models are increasingly important due to three trends: limitations of language-based intelligence, rapid advancements in architecture and algorithms, and the demand for embodied intelligence [5]. Group 2: Key Improvements Needed for World Models - Long-term memory is crucial for generating coherent, continuous worlds, with current models limited to short video segments [6][7]. - Interactivity is essential, allowing users to influence world generation through real-time actions, which requires innovative training methods [8][11]. - Real-time feedback is critical for applications like gaming and VR, with current models struggling to meet low latency requirements [12][15]. - Physical realism is vital for high-stakes applications like autonomous driving, necessitating models that adhere to real-world physics [16][18]. Group 3: Two Development Paths for World Models - The first path focuses on real-time video world models for consumer applications, prioritizing interactivity and long-term memory over physical realism [19][20]. - The second path emphasizes structured 3D models for robotics and autonomous driving, prioritizing physical accuracy and reliability [21][22]. Group 4: Market Players and Their Positions - The market is categorized into four quadrants based on representation forms and target audiences, with players like Decart and Odyssey positioned in different segments [24][26]. - World Labs is highlighted as a leading startup focusing on spatial intelligence, emphasizing 3D consistency and persistence in its models [26][28]. - General Intuition leverages vast gaming data to train agents for spatial-temporal reasoning, positioning itself uniquely in the market [33][35]. - Decart aims for speed and efficiency with its interactive AI model Oasis, while Odyssey focuses on high-fidelity reconstruction for creative industries [39][45].

Core Viewpoint - The article discusses the advancements and potential of embodied intelligence, particularly focusing on the development of a "brain" for robots that can adapt and learn across various forms and tasks, highlighting the contributions of companies like Qianjue Technology and Liufeng Space [2][3][4]. Group 1: Embodied Intelligence Development - Embodied intelligence has emerged as a hot investment area, with significant advancements in creating "small brains" but challenges remain in developing a comprehensive "big brain" [3][4]. - Recent scientific research indicates substantial potential for embodied intelligence, although the foundational paradigms are still evolving [4]. - Qianjue Technology aims to create a "brain in a jar" that can be utilized by various robot forms, with plans to connect 100,000 devices to its system by next year [4][5]. Group 2: Technical Approaches - Qianjue Technology employs a decoupled approach to brain modeling, allowing for independent optimization and evolution of different brain regions, which enhances efficiency [5][14]. - Liufeng Space focuses on building world models that drive embodied brains, utilizing real-time interactive space generation technology [6][11]. - The two companies represent different paths in the development of embodied intelligence, with Qianjue emphasizing brain-like structures and Liufeng leveraging world models for practical applications [8][10]. Group 3: Data and Training - Data scarcity is a significant challenge in training embodied intelligence systems, with Qianjue Technology achieving multiple generations of pre-training, which is rare in the industry [14][17]. - Liufeng Space believes that good robot data should be treated as an asset, emphasizing the importance of diverse and abundant data for effective training [12][17]. - Both companies recognize the need for extensive data to achieve effective pre-training, with estimates suggesting that a billion clips may be necessary for comprehensive training [26][27]. Group 4: Future Outlook - The timeline for achieving a mature embodied brain technology is optimistic, with both companies suggesting that significant advancements could occur within two years [26][27]. - The potential for embodied intelligence to surpass language models is highlighted, with expectations for the emergence of numerous billion-dollar companies in this sector [27].

"通往超级智能的那条路——无非是不断训练大语言模型、喂更多合成数据、雇上几千人做后训练、再在强化学习上搞点新花样——在我看来完全是胡 扯，这条路根本行不通。" 近日，在一档名为《The Information Bottleneck》的访谈栏目中，主持人 Ravid Shwartz-Ziv 和 Allen Roush 与图灵奖得主、前 Meta 首席 AI 科学家 Yann LeCun 展开了一场近两小时的高质量对话，在访谈中，LeCun 解释了为什么会在 65 岁这个别人已经退休的年纪他还在创业，此外，他也对当前硅谷主流 的人工智能发展路径给出了罕见而尖锐的评价。 结束在 Meta 长达 12 年的职业生涯后，LeCun 正将个人学术声誉与职业"遗产"押注在一套截然不同的 AI 愿景之上。他直言，业界对大语言模型规模化的 执念，正在把人工智能引向一条看似高速、实则封闭的死胡同。 在 LeCun 看来，真正制约 AI 进步的关键，并不是如何更快地逼近"人类级智能"，而是如何跨越一个常被低估却极其困难的门槛——让机器具备"狗的智 能水平"。这一判断挑战了当前以语言能力和知识覆盖面为中心的评估体系。在他看来，现实世 ...

Government Initiatives - The Ministry of Industry and Information Technology (MIIT) has revised the "Management Measures for Public Service Platforms for Industrial Technology," effective from December 5, 2025, focusing on key industries such as equipment, petrochemicals, steel, and artificial intelligence [1][1] - The National Development and Reform Commission, along with other ministries, has issued opinions to strengthen the construction of data element disciplines and digital talent teams, aiming to support the development of a digital economy and society [1][1] - The Ministry of Ecology and Environment has released guidelines for the construction of a product carbon footprint factor database to support the establishment of a carbon footprint management system [1][1] - MIIT is seeking public opinions on the "Comprehensive Standardization System Construction Guide for the Metaverse Industry (2026 Edition)," aiming to establish over 50 national and industry standards by 2030 [1][1] Local Actions - Shandong Province is promoting the metaverse as a new economic growth point, supporting cities like Jinan and Qingdao in building future industry pilot zones [1][1] - Jiangsu Province has established a Metaverse Standardization Technical Committee in Nanjing to fill the gap in the standardization system within the province [1][1] Industry Developments - The GPU leader, Moore Threads, has officially listed on the STAR Market, becoming the first domestic GPU stock, with a market capitalization of 305.5 billion yuan and an opening surge of 468.78% [3][3] - Google has integrated AI simultaneous translation into all its headphones and launched an experimental browser named "Disco," aiming to redefine web browsing experiences [3][3] Academic Insights - Academician Zhang Yaqin predicts that the future of large models will not exceed ten, emphasizing the integration of information, physical, and biological intelligence [4][4] - Academician Tan Jianrong stresses the importance of small models as the foundation for large models, advocating for a shift towards precision small models and industry-specific intelligent agents [4][4] Technology and Applications - The Ministry of Industry and Information Technology has granted approval for China's first batch of L3-level conditional autonomous driving vehicles, marking a significant step towards commercialization [6][6] - Mathematician Terence Tao and his team have solved the 50-year-old Erdős 1026 problem in just 48 hours using AI tools, showcasing the potential of AI in solving complex mathematical challenges [6][6]

在下午的科技产业投资专场中，圆桌对话《穿越周期的早期投资：从"赛道思维"到"认知红利"》探讨了 在共识廉价、市场极度内卷的当下，投资人如何穿越周期，从"赛道思维"转向"认知红利"。 英诺天使基金合伙人、北京前沿国际人工智能研究院理事长王晟作为嘉宾主持人，对话红杉中国合伙人 张涵、元禾原点合伙人乐金鑫、峰瑞资本合伙人马睿、心资本合伙人吴炳见等多位嘉宾。 面对AI、具身智能等赛道的迅速拥挤，嘉宾们指出，单纯赌赛道的时代已经结束，真正的决胜点在于 对人、对周期以及对非共识的深刻理解。 在"红海"共识中寻找认知的非共识。 2025年12月3日，「甲子光年」在北京万达文华酒店圆满举办"轰然成势，万象归一"2025甲子引力年终 盛典。 红杉中国合伙人张涵 乐金鑫：我是来自元禾原点的乐金鑫，元禾大本营是在苏州，既不靠北也不靠南。元禾原点一直是元禾 旗下早期的投资平台，到今年也12年的时间了。 从红杉中国的全链条布局，到峰瑞资本的内容影响力构建，再到新兴机构的个人IP打造，投资人们正在 通过不同的方式建立自己的"认知模型"和项目雷达。 大家普遍认为，保持"手感"、建立正向反馈循环以及在行业低谷期的坚持，是"捕捉下一个珍珠"的 ...

作者丨 许华哲 编辑丨具身智能之心 本文已经得到许华哲博士的授权，未经允许，不得二次转载。 点击下方 卡片 ，关注" 具身智能之心 "公众号 >> 点击进入→ 具身 智能之心 技术交流群 昨天看到了许华哲老师在社交媒体上的分享，关于数据、量产、本体和场景。类似的观点，今年IROS圆桌期间，许博也站在智能第一性原理上，将具身的未来发展 方向划分为欲望、先验和经验三个模块。 欲望。 在做智能体的时候，无论是物理的还是虚拟的，总觉得现在机器学习没有自己的学习欲望。我们可以设想一下，能不能给机器人一种自己的欲望？ 经验。 经验是完成世界最终闭环的一种手段。有一天，在家里面看到一位维修师傅就是帮我们修煤气灶，他踩在一个梯子上拧一个东西，整个身体造型极为扭曲， 但他仍可以完美控制重心保持平衡，并且手上还可以做非常精细的操作。 ★ 这种思想也贯穿在后续的研发和学术探索上。 回想起几年前，我们还在讨论机器人什么时候能全地形走路，后来发现这个话题变成了"跑酷"、"跳舞"、"篮球"。这个变化速率让我知道这个事儿已经成了，如果 明年可以攀岩我并不吃惊。 但这极快的变化速率又显得格外不协调，因为我没在任何地方看到人形机器人真正服务人 ...

点击下方 卡片 ，关注" 自动驾驶之心 "公众号 戳我-> 领取 自动驾驶近30个 方向 学习 路线 世界模型，近一年自动驾驶学术界和工业界的热词。很多小伙伴咨询柱哥，有没有一门系统讲解世界模型和自动驾驶的精品课程，筹备了很久终于和大家见面！ 我们联合 工业界大佬 共同开展，先前的《端到端与VLA自动驾驶小班课》备受大家好评，因此我们进一步推出这门世界模型小班课， 课程聚焦于通用世界模型、 视频生成、OCC生成等世界模型算法，涵盖特斯拉世界模型、李飞飞团队Marble等。欢迎大家加入学习~ 早鸟优惠！开课即止~ 讲师介绍 Jason：C9本科+QS50 PhD，已发表CCF-A论文2篇，CCF-B论文若干。现任国内TOP主机厂算法专家，目前从事端到端、大模型、世界模型等前沿算法的预研和量 产，并已主持和完成多项自动驾驶感知和端到端算法的产品量产交付，拥有丰富的端到端算法研发和实战经验。 课程大纲 这门课程讲如何展开 第一章：世界模型介绍 第一章主要针对自动驾驶世界模型概括性的内容讲解。 这一章老师会先复盘世界模型和端到端自动驾驶的联系，接着讲解世界模型的发展历史以及当下的应用案 例。然后介绍世界模型有哪些流派 ...