本文来自微信公众号： 新智元 ，作者：新智元，编辑：好困、定慧，题图来自：AI生成 如果将时间拨回三年前的今天，也就是2022年12月1日，那是一个相对安静的周三。 位于旧金山的一家名为OpenAI的非营利实验室，悄无声息地发布了一个名为"ChatGPT"的研究预览 版。 没有盛大的发布会，没有乔布斯式的演讲，只有一个朴素的对话框。 当时的人们并不知道，这个对话框将彻底改变世界。 ChatGPT早已不是那个偶尔会算错数学题的聊天机器人，它和它的继承者、竞争者们已经成为了人类 在数字AI世界赖以生存的"氧气"。 然而，伴随着技术的指数级跃迁，一种难以名状的群体性焦虑正在全球蔓延，和每个人都息息相关。 ChatGPT三年前的样子 三年后的今天，2025年12月12日，当我们站在这个时间节点回望，世界已经被彻底重塑。 这三年里，围绕ChatGPT和生成式AI，我们见证了前所未有的狂热与恐慌交织在一起：硅谷高歌猛 进，华尔街亦疯狂逐利，但普通人和各行各业从业者却充满焦虑和不安。 正如《大西洋月刊》评论所言，我们正身处"ChatGPT建造的世界"！ 一个充满不稳定性的时代，大家都在战战兢兢地等待下一只靴子落地。 年轻人 ...

中证报中证网讯（龚梦泽 熊永红）12月1日，第七届全球智能驾驶大会在苏州相城区举办，大会以"智 联世界 驾驭未来"为主题，助力构建智能驾驶全球化发展新格局。 在此背景下，大会围绕"探索汽车智能化产品出海路径"和"构建汽车数字化与服务化出海生态"两大主题 展开专题交流，来自中国机电产品进出口商会、中国汽研（601965）、岚图汽车、轻舟智航、曹操出 行、奥托立夫等机构与企业的多位行业知名人士积极参与讨论。 此外，现场还对《江苏省无人驾驶装备商业示范应用工作指引（试行）》进行了解读。此次《工作指 引》的出台，为无人驾驶技术走向市场化、规模化应用提供了重要指导。 当前，自动驾驶领域正围绕端到端、VLA与世界模型等主流技术路线展开探索。其中，世界模型通过 对物理环境的高维认知建模，让智能体首次具备"理解世界、预测未来、自主决策"的能力，依托苏州丰 厚的车路云数据资源顶尖院所携手行业领袖，将全面启动世界模型联合研发，旨在攻克下一代智能驾驶 核心技术，驱动产业范式变革。 近年来，苏州奋力打造全球领先的"智驾之城"，已集聚相关企业超800家，苏州智能车联网产业规模达 1100亿元，成功获批国家5G车联网验证与应用项目国家首 ...

Core Insights - Runway's Gen-4.5 model achieved the highest ELO score of 1,247 in the Artificial Analysis leaderboard, surpassing all other AI video models globally [1][5][28] Company Overview - Runway is the first company to successfully commercialize text-to-video technology as a SaaS product, launching Gen-1 and Gen-2 in early 2023, while competitors like Google's ImagenVideo and Meta's Make-A-Video were still in experimental stages [7][30] - The company has established itself as a leader in the AI video generation space, creating a distinct commercial pathway for AI video generation ahead of OpenAI's Sora, which was released in early 2024 [8][31] Technology and Innovation - Gen-4.5 utilizes advanced technology to set new benchmarks in video generation, particularly in motion quality, adherence to prompts, and visual fidelity [3][26] - The model demonstrates significant improvements in pre-training data efficiency and post-training techniques, positioning itself as a foundational model for world modeling [5][28] - Gen-4.5 is capable of producing highly realistic movements and interactions, showcasing unprecedented physical accuracy and visual precision [31][32] Market Position and Competitive Edge - Runway's focus on efficiency and a dedicated team passionate about video generation has allowed it to compete effectively against larger companies with more resources [37][40] - The company emphasizes the importance of "taste" in model training, which refers to the intuitive understanding of how to train models effectively [40] Future Applications - The potential applications of video models extend beyond entertainment, including non-linear interactive experiences, embodied AI for robotics, and personalized learning [46] - Runway aims to create a new medium capable of simulating a wide range of scenarios, moving beyond just video editing tools [46]

这场由黄大年茶思屋总编主持，聚集了中科院自动化所、南京大学、北京通用人工智能研究院、极佳科 技等机构专家的大讨论，直指目前 AI 领域最热门的方向——世界模型。最近一段时间，从谷歌 Genie 3 的发布到李飞飞的长文论述，世界模型、空间智能等概念正成为新的焦点。 机器之心报道 机器之心编辑部 李飞飞等顶尖学者投身的创业方向——世界模型是 AI 的下一站吗？ 「AI 是人类自诞生以来，唯一担得起『日新月异』这个词的技术领域，」在机器之心近日举办的 NeurIPS 2025 论文分享会圆桌讨论上，茶思屋科技网站总编张群英的开场感叹引发了在场专家们的共 鸣。 四十多分钟的对话里，专家们围绕世界模型的定义、数据与架构方向、技术路径分歧，以及商业化前景 展开了讨论。在一些议题上，大家的观点一致，不过在很多重要方向上有着明显不同的思考。看得出， 面对这个正在快速发展的新兴领域，不论是技术还是评判标准，我们还有很多需要去探索、验证的。 首先，世界模型究竟是什么？ 几位嘉宾从不同角度给出了自己的定义。 极佳科技联合创始人、首席科学家朱政认为，世界模型本质上是预测模型：「给定当前状态及动作序 列，预测下一个状态。」他指出了世 ...

Core Insights - The article discusses the advancements in Tesla's Full Self-Driving (FSD) technology, particularly focusing on the integration of end-to-end models and world models, which are crucial for the evolution of autonomous driving technology [1][3][17]. Group 1: Tesla's FSD Developments - Tesla's AI VP Ashok Elluswamy shared significant updates on FSD, highlighting the use of a multi-modal input system that combines video, navigation maps, and audio signals into a single end-to-end neural network [1][3]. - The end-to-end architecture allows for direct output of control signals, enhancing the system's performance and reducing latency [3][4]. - The challenges faced in building an effective end-to-end system include the "curse of dimensionality," where the input data volume can explode, making real-time processing difficult [4][5]. Group 2: World Model Concept - The world model is described as a generative spatiotemporal neural system that compresses multi-modal inputs into latent states, enabling future environment predictions [18][20]. - It allows for action-conditioned future predictions, providing insights into how different actions will affect the environment, thus enhancing decision-making capabilities [21][22]. - The integration of world models with planning and control systems enables a closed-loop feedback mechanism, allowing for real-time evaluation of actions and risk assessment [22][24]. Group 3: Comparison of Approaches - The article contrasts world models with Visual-Language-Action (VLA) models, noting that world models focus on physical simulation and long-term evaluations, while VLA models leverage language processing for decision-making [46][49]. - World models are seen as more aligned with the physical nature of autonomous driving, while VLA models offer advantages in handling rare scenarios through language-based reasoning [49][50]. - The ongoing debate between these two approaches suggests that the future of autonomous driving may involve a combination of both methodologies [49]. Group 4: Developments in China - Chinese companies like NIO and Huawei are actively developing their own world models, with NIO's NWM (Nio World Model) being a notable example that integrates multi-modal information for future scene predictions [28][30]. - Huawei's WEWA architecture emphasizes direct perception-to-action pathways, avoiding language abstraction to enhance real-time decision-making capabilities [36][40]. - SenseTime's "KAIWU" world model focuses on generating high-fidelity simulation data, showcasing the growing importance of world models in the Chinese autonomous driving landscape [41][45].

允中 发自 凹非寺 量子位 | 公众号 QbitAI VLA模型性能暴涨300%，背后训练数据还 首次实现90%由世界模型生成 。 具身智能迈向开放世界落地的 最大瓶颈 ， 长期以来并非算法本身，而是高质量、大规模真实机器人交互数据的极度稀缺 。 真机数据采集成本高昂、周期漫长，且难以覆盖多样化的开放场景，严重限制了VLA大模型的规模化训练与泛化能力。而传统仿真虽能快速生 成数据，却受限于显著的Sim-to-Real gap，难以支撑真实世界的鲁棒部署。 世界模型（World Model）被认为是破解这一困境的关键 ：通过学习真实世界的规律，世界模型可以生成高保真、可控、多样化的具身交互 数据，突破真机数据不足的限制。 在此背景下，刚刚获得华为投资的国产世界模型公司 极佳视界 发布并开源具身世界模型 GigaWorld-0，成功将世界模型生成数据在VLA训 练中的占比提升至90% 。 所训练的VLA模型在新纹理（训练中未见材质表面）、新视角（训练中未见的观测角度）、新物体位置（训练中未见的空间布局） 三大泛化 维度上均实现近300%的性能提升 ， 标志着具身智能正式迈入"数据高效、高泛化、低成本"的新阶段 。 ...

在如今的人工智能竞赛里，扎克伯格和他的Meta可能是最"激进"的玩家，没有之一。 在过去一年时间里，扎克伯格豪掷千金、四处摇人，试图组建世界上最强大的AI产品团队，动辄就为那些有过OpenAI、Anthropic等头部公司工作经 历的人才开出1亿美元的"跳槽奖金"。其中最大一笔开支用在了汪涛身上——为了让这位天才少年顺利地加入Meta，带队人工智能团队，扎克伯格豪 掷148亿美元直接收购了汪涛创办的Scale AI，直接整体打包带走。 如果谈得再务实一点，大语言模型虽然在文本推理与知识处理上取得突破，但在理解真实物理空间、进行连续动作规划以及与环境实时交互方面仍 然存在根本性缺陷。这类缺陷不仅让AGI的实现遥遥无期，更直接限制了人工智能技术向具身智能等更实际应用场景的拓展。 除此而外，扎克伯格SSI的首席执行官、前Y Combinator合伙人丹尼尔·格罗斯（Daniel Gross）旗下的风险投资基金NFDG，并顺势邀请NFDG的两位 合伙人——丹尼尔·格罗斯与前GitHub首席执行官、著名科技播客"Hacker Medley"的主理人纳特·弗里德曼（Nat Friedman）加入Meta，准备组建Meta ...

长城将在全新蓝山智能进阶版上搭载 VLA 辅助驾驶方案。 文 丨 沈行 辅助驾驶正处在技术的交界点，"规则" 方案的上限已经触顶，去年以特斯拉为首的 "端到端" 路线开 始快速普及，即传感器原始信息直接输入模型、模型直接输出车辆动作。端到端的意义在于，它让辅 助驾驶研发进入了 AI 时代：能力不依赖人工编程，而是随着数据规模的扩大持续进化。 但随着应用场景变得更复杂，行业开始意识到仅靠模仿难以覆盖真实世界的多样性——端到端无法处 理未曾 "见过" 的极端场景，为了突破这个瓶颈，辅助驾驶从端到端继续演进，向能够理解语义、推理 场景、解释决策的 VLA（Vision-Language-Action）过渡，能力的重心从 "会开" 转向 "会想、会判 断"。 对于辅助驾驶发展趋势的探索，长城是行业内反应更快、判断更明确的那一个。长城是最早完成 VLA 量产落地的车企之一，并率先在即将发布的全新车型——魏牌全新蓝山智能进阶版上搭载。 今年前三季度，长城汽车累计投入研发费用约 66.36 亿元，同比增长 6.9%。智能化是重要的研发方向 之一，长城汽车 CTO 吴会肖曾表示，智能化相关研发费用目前约占长城总研发费用的一半 ...

Group 1 - The WISE 2025 conference in Beijing focuses on immersive experiences and the impact of AI on various industries, highlighting trends such as hardware transformation and brand globalization [1] - The concept of embodied intelligence is identified as a hot topic for 2025, emphasizing its transition from a mere execution tool to an intelligent partner capable of perception and autonomous decision-making [4][5] - The conference features discussions on significant advancements in AI and embodied intelligence, with industry leaders sharing their insights and product developments [12][13][14] Group 2 - Companies like DiGua Robotics, Yuanli Unlimited, and Kuawei Intelligent are actively developing products in the embodied intelligence space, showcasing innovations such as the RDK Agent and humanoid robots [12][13][14] - The advancements in AI are enabling developers to leverage large models, enhancing their capabilities and allowing for more complex applications in robotics [15][16] - The industry is witnessing a shift towards using synthetic data for training models, which is crucial for overcoming challenges in real-world applications [20][32] Group 3 - The panelists discuss the potential for AI to create scalable value in industries that are currently limited by slow productivity and knowledge transfer [21] - The importance of industry acceptance and the maturity of AI technology are highlighted as critical factors for achieving large-scale impact [22][23] - The future of AI and embodied intelligence is seen as a long-term journey, with expectations for significant changes in daily life and work processes over the next five years [33][34]

Core Insights - CES 2026 is set to showcase significant advancements in AI technology, with major companies like Siemens, Caterpillar, AMD, and Lenovo focusing on AI in their presentations [5][8][19] - The event will highlight a variety of AI hardware products, including AI glasses, AI PCs, AI smartphones, and humanoid robots, indicating a strong trend towards AI integration in consumer electronics [18][19] - Chinese brands are expected to dominate CES, showcasing their technological innovations across various categories, reflecting their growing influence in the global market [40][41] AI as the Central Theme - AI will be the overarching theme of CES 2026, with confirmed keynote speeches from industry leaders emphasizing its importance [5][19] - Companies like Siemens will demonstrate how AI and digital twin technology can transform manufacturing and infrastructure [8] - Lenovo plans to unveil innovations related to AI-driven experiences, including applications in sports and personalized user interactions [11] PC and Gaming Innovations - Intel, AMD, and NVIDIA are anticipated to launch new products, including Intel's Panther Lake mobile processors and AMD's R9 9950X3D processor with enhanced cache capabilities [19][21] - The introduction of new gaming processors and graphics cards is expected to attract significant attention from the gaming community [21][22] Display Technology Competition - Major TV manufacturers, including TCL and Hisense, are expected to showcase advancements in RGB display technology, competing with international brands like LG and Samsung [25][26] - The CES 2026 will feature a variety of display technologies, including Micro RGB LCD and Mini LED, highlighting the competitive landscape in the display sector [25][26] Smart Cleaning Devices - Chinese smart cleaning brands are set to unveil new products, including robotic vacuums and lawn mowers, reinforcing their leadership in the global smart cleaning market [27][30] - The focus will be on comprehensive cleaning solutions that leverage AI and advanced navigation technologies [30] Accessory and Audio Innovations - Accessory brands like Baseus and Ugreen are expected to expand their product lines beyond traditional charging devices, venturing into audio and smart home solutions [31][34] - The introduction of high-end audio products and smart home security devices will be a key focus for these brands at CES 2026 [36] AI Glasses and New Hardware - AI glasses are anticipated to be a major highlight, with various brands competing in this emerging category [38] - The presence of established players and new entrants in the AI hardware space will create a dynamic showcase of innovative products [39] Chinese Brands' Dominance - Chinese companies are projected to play a pivotal role at CES, with a significant share of exhibitors and a focus on technological innovation rather than just cost competitiveness [40][41] - The event serves as a platform for Chinese brands to demonstrate their rapid product development and engineering capabilities across multiple tech sectors [40][41]