Core Insights - Artificial intelligence (AI) is a strategic technology leading a new wave of technological revolution, significantly transforming human production and lifestyle [1] - Beijing is positioning itself as the "AI capital" of China, with over 2,400 AI companies and a core industry scale nearing 350 billion yuan, accounting for half of the national total by 2024 [1] Group 1: AI Innovation and Research - Beijing is recognized as the city with the richest AI innovation resources in China, hosting 21 national key laboratories and over 40% of the nation's top talent [2] - The city has established four new research institutions focused on AI, producing globally leading original results, including the first native multimodal large model, Emu [2] - The Zhiyuan Institute has developed the "Wudao" series of large models, with Wudao 1.0 and Wudao 2.0 being significant milestones in China's AI model development [2][3] Group 2: AI Applications and Developments - Beijing has launched 132 large models, leading the nation in this area, and is focusing on disruptive technologies like optical computing and wafer-level chips [4] - The integration of AI with hardware is exemplified by companies like Mianbi Intelligent, which focuses on edge AI models that perform processing directly on user devices [4] - The education sector is set to benefit from AI with the introduction of MAIC (Massive AI-empowered Courses), which aims to enhance teaching efficiency and learning outcomes [5] Group 3: Future Directions and Infrastructure - Beijing plans to enhance its AI infrastructure, with an expected addition of 8,620 PetaFLOPS of computing power by 2024, bringing the total to over 33,000 PetaFLOPS [7] - The city aims to establish itself as a global hub for AI innovation and industry, focusing on interdisciplinary fields such as AI + life sciences and AI for science [7] - Efforts will be made to integrate data and applications, leveraging Beijing's rich data resources and comprehensive industrial system to promote the application of large models in the economy [7]

作者 | 华卫 今年的智源大会上，智源研究院推出全新的"悟界"系列大模型，其中包括原生多模态世界模型 Emu3、脑科学多模态通用基础模型见微 Brainμ、跨本体具身大小脑协作框架 RoboOS2.0 与具身大 脑 RoboBrain2.0 以及全原子微观生命模型 OpenComplex2。 据介绍，Emu3 作为原生多模态统一架构让大模型具备理解和推理世界的能力，Brainμ基于则 Emu3 架构，引入脑信号这一新的模态数据，实现了单一模型完成多种神经科学任务的大一统。在初代版本 的基础上，RoboOS2.0 与 RoboBrain2.0 的原有性能有大幅提升，并新增多机协作规划与物理常识 驱动的空间推理能力。OpenComplex2 可在原子分辨率层面捕捉分子相互作用及平衡构象，探索微 观构象波动与宏观生物功能的跨尺度关联。 "大模型技术还远没有到发展的尽头。"在大会前夕，智源研究院长王仲远向我们透露了这一系列新模 型背后的技术思考与智源当下的战略布局。 王仲远指出，去年智源就对大模型的技术路线进行了预判，会从大语言模型往多模态、尤其是原生多 模态世界模型的方向发展。当前，智源的工作布局都是围绕这一技术发 ...

以ChatGPT为代表的语言类大模型重塑内容生成方式时，多模态模型还在等待它的"iPhone时刻"。近日召开的2025智源大会上，智源研究院（以下简称"智 源"）正式发布了包括原生多模态世界模型Emu3等"悟界"大模型系列，Emu3实现了文本、图像、视频的任何组合理解与生成，通过单一模型就可以捕捉世 界的规律。 AI发展之快，每年都有新话题，2024年，价格战是大模型的关键词，2025感到风向变了，大模型应用百花齐放，反而有种大模型发展"变慢"了的体感。 事实上，市场上新旧产品同台竞技，呈现出立体、多维度的思考，多模态大模型更是如此。按照当前技术成熟度评估，视频生成等核心能力仍处于GPT-2到 GPT-3的过渡阶段，与产业预期存在显著差距。多模态模型将经历更长的技术沉淀期，这也意味着更大的想象力空间。 技术路线未收敛 大模型爆发至今，很多时候无外乎是选对了方向，又懂得流量密码，一个现象级产品就横空出世了。事实上，这种选择需要前期足够多的思考、实践和勇 气。 严格来说，Emu3是智源2024年10月发布的多模态模型，目前智源已在训练下一个版本。基于Emu3，智源还官宣了全球首个脑科学多模态通用基础模型见微 Br ...

21世纪经济报道记者孔海丽 北京报道 2025年智源大会上，人形机器人不再是吉祥物，被"围堵"的人从杨植麟变成了王兴兴。 这一年，AI进展迅猛，迭代周期甚至少于3个月，且不再局限于大语言模型，而是转化为人形机器人训 练、落地的强辅助。 "人工智能正在加速从数字世界走向物理世界。"智源研究院院长王仲远在接受包括21世纪经济报道在内 的记者采访时直言："人工智能应该为世界做一些实实在在的事情，帮助人类摆脱繁琐的、重复的以及 简单的劳动。" AI技术路线转向世界模型 "大模型技术还远没有到发展的尽头，过往所说的'百模大战'更多是大语言模型的竞争，而大语言模型 受限于互联网数据的使用，基础模型性能虽然还在提升，但是提升速度不如以前。"在王仲远看来，大 语言模型性能提升瓶颈的解法主要包括三个方面，一是强化学习优化推理能力，二是合成高质量数据替 代人类标注，三是激活海量未充分利用的多模态数据，多模态数据的规模可达文本的"百倍乃至万倍"。 在智源研究院的判断中，大模型的技术路线会从大语言模型往多模态尤其是原生多模态世界模型的方向 发展。原生多模态世界模型本质上是为了让人工智能感知和理解物理世界，进而推进和物理世界的交 互。 ...

Group 1 - The core viewpoint of the articles emphasizes the rapid development of AI and its transition from the digital world to the physical world, highlighting the importance of world models in this evolution [1][3][4] - The 2023 Zhiyuan Conference marked a shift in focus from large language models to the cultivation of world models, indicating a new phase in AI development [1][3] - The introduction of the "Wujie" series of large models by Zhiyuan represents a strategic move towards integrating AI with physical reality, showcasing advancements in multi-modal capabilities [3][4] Group 2 - The Emu3 model is a significant upgrade in multi-modal technology, simplifying the process of handling various data types and enhancing the path towards AGI (Artificial General Intelligence) [4][5] - The development of large models is still ongoing, with potential breakthroughs expected from reinforcement learning, data synthesis, and the utilization of multi-modal data [5][6] - The current challenges in embodied intelligence include a paradox where limited capabilities hinder data collection, which in turn restricts model performance [6][8] Group 3 - The industry faces issues such as poor scene generalization and task adaptability in robots, which limits their operational flexibility [9][10] - Control technologies like Model Predictive Control (MPC) have advantages but also limitations, such as being suitable only for structured environments [10] - The development of embodied large models is still in its early stages, with a lack of consensus on technical routes and the need for collaborative efforts to address foundational challenges [10]

Core Insights - The Beijing Zhiyuan Conference showcased the latest developments in AI, including the release of the "Wujie" series of models by the Zhiyuan Research Institute, which aims to advance AI's understanding of the physical world [2][4] - The director of Zhiyuan, Wang Zhongyuan, emphasized that the next phase of AI development requires moving beyond language models to multi-modal world models that can perceive and interact with the physical environment [4][5] Model Releases - The "Wujie" series includes four models: Emu3, Brainμ, RoboOS 2.0, and RoboBrain 2.0, each designed to enhance AI's capabilities in understanding and interacting with the physical world [2][3] - Emu3 utilizes a new visual tokenizer technology to unify the representation of text, images, and videos, allowing AI to process them in a cohesive manner [3] - Brainμ aims to serve as a new engine for neuroscience research and clinical applications, integrating over one million neural signal data units [3] - RoboOS 2.0 improves performance by 30% compared to its predecessor, enabling faster integration of developer plugins and enhancing real-time response capabilities [3] - OpenComplex2 targets life sciences by simulating molecular movements at atomic resolution, potentially accelerating drug development and biological research [3] Strategic Partnerships and Goals - Zhiyuan has signed a strategic cooperation agreement with Hong Kong Investment Management Company to foster talent, technology, and capital collaboration [6] - The organization is committed to open-source and international collaboration, having already open-sourced 200 models with a total of 640 million downloads [7] - Wang Zhongyuan highlighted the importance of patience and sustained capital investment for long-term goals, despite short-term commercialization challenges [5][6]

Group 1 - The Beijing Zhiyuan Conference showcased cutting-edge AI achievements, gathering hundreds of global young scientists, top scholars, and industry experts to outline the future of the AI industry [1] - AI is rapidly transitioning from the digital world to the physical world, with the release of the original multimodal world model Emu3, which enhances understanding and reasoning in physical contexts [3][4] - The original multimodal model integrates various data types from the beginning of training, allowing for a more comprehensive understanding of the world, unlike traditional models that may lose capabilities when learning additional modalities [4] Group 2 - Beijing has over 2,400 core AI enterprises, contributing to a core industry scale of nearly 350 billion yuan, accounting for half of the national total [5][9] - The conference featured advanced humanoid robots demonstrating their capabilities, with companies like Galaxy General planning to open 100 unmanned pharmacies in major cities [6][8] - Discussions at the conference included topics such as multimodal AI, deep reasoning, and the future paths of AI, emphasizing the need for global cooperation and safety measures in the face of rapid AI advancements [10][13]

北京智源大会6月6日开幕。全球最强的开源具身大脑大模型、助力新型治疗方案研发的全原子微观生命 模型……作为北京市人工智能领域的新型研发机构，智源研究院在开幕式上发布"悟界"系列大模型，推 动人工智能从数字世界迈向物理世界。 从"悟道"到"悟界"，人工智能迈入现实物理世界 智源研究院院长王仲远表示，大模型技术还远没有到发展的尽头，过往所说的"百模大战"更多的是大语 言模型的竞争，而大语言模型受限于互联网数据的使用，基础模型性能虽然还在提升，但是提升速度不 如以前。 "大语言模型性能提升的解法有很多。"他说，一是通过强化学习，在后训练和推理上提升，例如 DeepSeek R1等，这是过去一年大模型产业界最大的进展之一。二是数据合成，目前学术界仍在突破。 互联网数据都是人类创造的，如果人工智能合成的数据、生成的数据质量能够达到人类创造的数据质 量，那意味着人工智能有可能实现自我学习和进步。三是使用多模态数据，在全世界范围内，多模态数 据是文字数据的千万倍甚至更多，这些数据远没有被有效利用。 大模型正在从大语言模型向原生多模态大模型、世界模型的方向演进。原生多模态世界模型本质上是为 了让人工智能感知和理解物理世界，进 ...

华泰证券认为这或将持续提升车载的芯片算力以及传感器的精度，对算法公司和主机厂技术研发能力也 提出了新的要求。亿欧智库的报告则称，世界模型通过云端训练+车端蒸馅提升泛化能力，但其规模化 落地仍受限于算力成本与数据质量。 截至6月6日14:47，中证大数据产业指数(930902)盘中震荡。成分股方面涨跌互现，石基信息涨停，科华 数据上涨2.43%，神州泰岳上涨1.91%；神州信息领跌3.04%，拓维信息下跌2.51%，税友股份下跌 1.99%。数据ETF(516000)多空胶着，最新报价0.92元。拉长时间看，截至2025年6月5日，数据ETF近1 周累计上涨1.89%，涨幅排名可比基金第一。流动性方面，数据ETF盘中交易活跃，换手6.44%，成交 2853.13万元。 消息方面， 6月6日上午，在2025北京智源大会上，北京智源人工智能研究院发布了"悟界"系列大模 型，宣布围绕物理AGI（通用人工智能）所做的大模型最新科研成果和布局。"悟界"系列大模型目前包 含：全球首个原生多模态世界模型"悟界·Emu3"、全球首个脑科学多模态通用基础模型"悟界·见微 Brainμ"、具身大脑RoboBrain 2.0、全原子 ...

6月6日，第七届"北京智源大会"在北京开幕。在大会上，智源研究院推出"悟界"系列大模型，包括原生 多模态世界模型Emu3、脑科学多模态通用基础模型见微Brainμ、跨本体具身大小脑协作框架RoboOS 2.0与具身大脑RoboBrain 2.0以及全原子微观生命模型OpenComplex2。 （36氪） 智源发布"悟界"系列大模型 ...