Industry 3176 articles · 61 matching

近几年，自动驾驶技术路线加速变化，产业链上的公司也在重新寻找自己的位置。 其中，也包括长期为智驾行业提供仿真测试服务的五一视界。这家成立于2015年的数字孪生公司，通过仿真技术切入自动驾驶领域，随后延伸到更多领域。2025年12月，五一视界成功登陆港股。 8月18日，五一视界公布“物理AI宏伟蓝图2030”，将公司的发展路径划为三站：自动驾驶、具身智能和空天探索。 从汽车到机器人，再到卫星和火星探测器，业务跨度看起来很大，但背后延续的是五一视界过去十年积累的一套能力，将现实环境数字化，让机器在其中完成训练、测试和验证，再通过模型和数据与真实世界形成闭环。 汽车是这套能力最早实现规模商业化的载体，机器人正在成为第二个。 五一视界接下来的调整，就是把已经在自动驾驶中验证过的能力，继续推向更广泛的物理世界。 从自动驾驶，到更大的物理世界 2016年，五一视界开始构想“地球克隆”计划，创业前三年计划将人与建筑数字化，第二个三年扩展到车与道路，第三个三年继续进入水和城市，再连接不同的物理要素。 彼时，自动驾驶行业刚刚兴起，创业公司如雨后春笋冒出，针对道路环境复杂，大量危险和极端场景无法反复在现实…

Netflix" 开源了一个用于观察性因果推断（OCI）的智能代理工作流"。该工作流可以减轻因果分析过程中的繁琐工作。基于观察数据和人类用户的分析计划，该智能代理能够利用行为者-批评者循环来评估因果关系、撰写报告并提出后续步骤的建议。 该智能代理工作流基于 Netflix 现有的 OCI 工具构建。其目标是自动化易出错或重复性任务（如敏感性分析或跟踪多次迭代），将更高层次的任务（如问题构建和结果评估）留给人类用户。OCI 分析被定义为目标试验仿真"，即寻找能够解答该问题的最优 A/B 测试方案。Netflix 团队使用大西洋因果推断会议（ACIC）竞赛数据集"对该工作流进行了评估，发现其与各类基准系统相比非常“具有竞争力”。据 Netflix 说： 在因果推断中使用智能代理面临着一个挑战：在没有真实标注数据的情况下，我们如何评估智能代理在各项任务中的表现？为了应对这一挑战，我们的工作流程将流程审核与人工监督相结合。为了让其他人能够借鉴并评判这一工作流程，我们已经将一个轻量级的独立版本开源。我们希望，这项工作能激发更多在没有真实标注数据的情况下对智能代理进行评估的研究与开发。 人类分析师…

北京，2026年8月19日—23日，2026世界机器人大会（WRC）在北京举行。 本届大会，智身科技以"真正干活的机器人"为核心定位，通过五大展区集中展示从核心关节、整机本体、机械臂到运动控制（小脑）、智能决策（大脑）的全栈技术能力，回答行业之问：机器人什么时候能真正创造社会价值？ 01.展台即现场：从零件到整机，从表演到干活 走进智身科技展台，五大区域构成一条清晰的"干活能力"展示链路。 表演舞台区三大亮点集中呈现：智身银毅NE01人形机器人上演《功夫女足》多机协同舞蹈，单关节峰值扭矩达 180N・m；智身钢镚 L2 四足机器人展示镂空楼梯感控一体通行与 7:1 极限负载自重比。 台面操作演示区，智身科技带来首款行业级可组装教育四足机器人，从通识启蒙、专业实训到工程应用，支撑不同阶段的具身智能教育建设；MATRiX 2.0开放世界仿真训练平台提供手柄虚拟对战体验，打通从虚拟训练到真实部署的链路，现已开源。 地面操作演示区，智身铜锤M1搭载铁臂TA01机械臂展示VLA操作能力，即使智身铜锤M1腿部、本体持续运动时，智身铁臂 TA01 末端依然能保持稳定。系统能够根据动作与负载变化实时协…

Insider Brief Researchers led by the University of Hong Kong have developed a new benchmark for testing physical AI systems across both simulation and real-world robot tasks, finding that even the strongest models remain far behind human performance. The RoboDojo platform was developed by the Multimedia Laboratory at the University of Hong Kong, in collaboration […]

Current world models like Sora or Genie only simulate physics and ignore what people think, want, or feel. The new "Mental World Modeling" framework adds mental variables like beliefs and intentions. Even weaker language models using this approach outperform stronger models without mental modeling. The biggest bottleneck: predicting how physical and mental states change together. The article World models that ignore human beliefs predict the wrong actions, new research shows appeared first on Th…

Research suggests more than half of electronic product failures stem from thermal failures caused by rising chip junction temperature. As transistor density climbs, power density and heat grow, making package thermal design a key bottleneck. Huatian Technology has built a full-process thermal design and simulation capability in fan-out packaging to address heat at the source.

当大语言模型化身为智能体走向真实世界，它们面临的困境，与十年前在十字路口指挥交通的强化学习模型如出一辙。 作者丨岑峰 幸丽娟 编辑丨岑峰 在通往 AGI 的浪潮中，基础模型智能体（Foundation Model Agents）正以前所未有的速度接管各种任务。业界曾普遍乐观地认为，凭借大模型海量的预训练知识，智能体可以无缝适应各种真实场景。 但在现实中，当我们将一个在英文语境下表现完美的智能体，切换到中文或低资源语言环境中时，其错误率往往会呈现断崖式上升。 这种环境切换带来的“智力降级”，真的是大模型时代特有的新问题吗？ 在正在进行的 2026 年的德国不莱梅 IJCAI大会的 Early Career Spotlight（早期职业焦点）环节，亚利桑那州立大学（ASU）助理教授魏华给出了一个极具启发性的“降维”视角：今天大模型智能体所面临的脆弱性，本质上与传统强化学习（RL）在物理世界中遭遇的“仿真到现实（Sim-to-Real）”差距如出一辙。 作为长期深耕强化学习与真实世界决策的学者，魏华教授早年曾在腾讯 AI Lab 担任研究员，后于新泽西理工学院和 ASU 任教。早在2017 …

An artificial intelligence startup called Astromech that’s developing models that can predict biological change has raised $20 million in funding, lifting its valuation to $3.8 billion. The round was led by biotechnology investor Bob Nelsen and saw the participation of Peak 6, NeoGenesis Capital, Builders VC and CAZ Investments, bringing the company’s total amount raised […] The post Astromech raises $20M to build a biological operating system that can forecast evolutionary change appeared first…

Google announced a broad set of new AI-powered study features across Search and Gemini, including interactive visuals, 3D simulations, customized practice quizzes, and a dedicated learning hub, as the company works to position Gemini as a primary AI tool for students amid growing competition from OpenAI and education-focused startups. Within Search, students can now generate […]

Turing Award winner Richard Sutton calls synthetic data a "big mistake" for scaling large language models. The world is infinitely complex, and any simulation of it is "microscopic," with human expertise acting as a bottleneck that blocks real scaling. Sutton's alternative is agents that learn continually from their own experience instead of relying on frozen models. The article KI-Pioneer Sutton calls synthetic data a "big mistake" in the face of an infinitely complex world appeared first on Th…

宇树的下一张牌，是搭起一套系统，让全球 AI 的持续进步，转化为机器人不断加快的进化速度。 作者 | Li Yuan 8 月 19 日，宇树科技正式登陆科创板。上市首日，公司股价一度较发行价涨超 6 倍，总市值突破 4400 亿元。 当市场盯着这家「机器人身体最强公司」的股价时，一个更重要的问题随之出现：获得更多资金和资源之后，宇树接下来要做什么？ 当天中午，王兴兴参加了宇树的上市答谢午宴。这几乎是公司上市后的标准动作：感谢一路支持的政府、股东和合作伙伴，再回望一遍创业十年的艰辛。 但在略显磕巴地完成这些常规致谢后，王兴兴没有就此结束，而是突然把话题转向技术，开始讲述宇树上市后真正准备投入的一套系统。 他的状态也随之变化。表达明显流畅起来，语速越来越快，从大模型讲到自动编程，从仿真训练讲到真机测试，再讲到机器人技能的自我进化。 显然，这部分内容他已经想了很久。 01 当 AI 进入机器人研发全过程 王兴兴讲的并不是宇树即将发布的某个模型或产品，而是他对机器人技术下一阶段的判断：只做一个 AI 模型或者视频模型，整体竞争力和迭代速度已经不够。下一阶段真正重要的，是能否把模型、工具、仿真、…

具身智能产业关注的重点正从“能力涌现”进入“生产力兑现”阶段，拼的不只是单一模型能力。8 月 19 日下午，在北京举行的 2026 世界机器人大会上，百度集团副总裁袁佛玉表示，机器人从“会展示”到“能上岗，需要模型、数据与应用协同形成的系统能力。 百度集团副总裁 袁佛玉 模型从理解语言，走向理解和预测物理世界 机器人真正进入物理世界，仅理解语言并不够，还要理解物体关系、动作后果及环境变化。随着 VLA、世界模型和运动控制加速演进，计算工程系统也面临更高要求。 袁佛玉指出，依托百度百舸 AI 计算平台，百度智能云持续完善面向具身智能的 AI Infra，支持企业完成从数据处理、模型训练、仿真评测到真机推理部署的研发全流程，推动模型从研究原型走向工程系统；在部分合作中，单步推理延迟已进入百毫秒以内区间。 数据从训练材料，变成具身智能最核心的生产要素 具身数据采集成本高、标准不一，且与具体本体、任务和环境高度相关。 袁佛玉认为，未来的重要竞争要素，在于企业积累了多少有效任务小时、捕获了多少高价值失败样本，以及形成了多少能够回流训练的数据。相比重复完成同一动作，失败、纠正、人工接管和边界案例数…

给机器人造一个更接近真实的“第二世界”

In July, NVIDIA joined more than 200 companies and organizations in signing “Open Weights and American AI Leadership,” an open letter arguing that AI leadership will be measured not by any single frontier model but by whether an open ecosystem reaches every sector.

Image credit: Olivier Porchet, Biorobotics Laboratory, EPFL When a fish holds its position against a current in a river, its brain must figure out how fast to swim and how to steer to offset the water flow. Most fish use vision to register the world sliding past, detect optic flow speed and direction, and their […]

Unlike autonomous driving or industrial robotics, healthcare robotics can’t rely on internet-scale data collection or unlimited real-world experimentation.... Unlike autonomous driving or industrial robotics, healthcare robotics can’t rely on internet-scale data collection or unlimited real-world experimentation. Every demonstration requires specialized equipment, clinical expertise, and access to patients or laboratory environments. This creates three fundamental challenges for developers. Firs…

Google commits $40M in AI tokens and credits for the Genesis Mission

Developers building 3D, design, simulation, robotics, and industrial digital twin applications need ways to bring physical AI capabilities into the tools and... Developers building 3D, design, simulation, robotics, and industrial digital twin applications need ways to bring physical AI capabilities into the tools and services they already use. Many of these workflows already depend on OpenUSD scenes, simulation-ready (SimReady) assets, Blender-based workflows, CAD pipelines, or domain-specific a…

Imagine you want to teach a robot to push an object on a table. The standard recipe in robot learning is to collect hundreds of expert demonstrations on a real robot, train an imitation learning policy on that data, and then evaluate the policy by running it many times on the same real robot. Both […]

OpenUSD is an open, extensible framework that provides a common scene description language for physical AI. It enables teams to bring CAD data, simulation... OpenUSD is an open, extensible framework that provides a common scene description language for physical AI. It enables teams to bring CAD data, simulation assets, and real-world telemetry into a shared, physically accurate view of the world. Until now, building a USD implementation has typically required adapting a large existing codebase— …

As more teams move from humanoid robot bring-up to task-specific skill development, the need for repeatable development workflows is growing. Building humanoids... As more teams move from humanoid robot bring-up to task-specific skill development, the need for repeatable development workflows is growing. Building humanoids remains complex, and today’s development pipelines are still highly fragmented. As a result, developers spend significant time configuring robotics infrastructure before they …

NVIDIA Omniverse NuRec is a neural reconstruction pipeline for building high-fidelity 3D representations of real-world environments from multisensor data such... NVIDIA Omniverse NuRec is a neural reconstruction pipeline for building high-fidelity 3D representations of real-world environments from multisensor data such as cameras and lidar. It is used to reconstruct dynamic scenes captured by autonomous vehicle (AV) and robotics platforms into simulation-ready digital environments that can be re…

We’re expanding access to Google AI Ultra subscribers globally and introducing a new capability powered by Street View.