在迈阿密的一场颇具争议的审判中，特斯拉公司被裁定必须对一起涉及其自动驾驶辅助技术的致命车祸承担部分责任。陪审团最终决定，特斯拉需向事故受 害者支付超过2.4亿美元的赔偿金。这一结果的产生，显然指向了特斯拉技术缺陷对事故的重要影响。尽管肇事司机承认了自己因玩手机而分心，导致撞上 了一对正在外出观星的年轻夫妇，但陪审团认为，这并不是事故责任的全部。 这一裁定恰逢马斯克积极向美国公众推广其自动驾驶汽车安全性的时刻，也正值他计划推出无人驾驶出租车服务的关键时刻。由于该服务将在未来几个月内 在多个城市启动，因此这一判决无疑为他的计划带来了额外的挑战。这个案件从发生到最终裁决历时四年，其复杂性和重要性都不容小觑。值得一提的是， 案件能够进入审判阶段实属不易，之前有多起类似的诉讼要么被驳回，要么在未开庭前便被特斯拉通过和解的方式解决，从而避免了公众视线的聚焦。 未参与本案的交通事故律师米格尔·库斯托迪奥表示，这一裁定或将为更多受害者提供了勇气，促使他们勇敢站上法庭。他指出，今后可能会有更多类似的 案件被提起，这无疑将改变整个行业对此类问题的处理方式。在案件审理过程中，原告方向22岁遇难者奈贝尔·贝纳维德斯的家属及其受伤的男 ...

"定向场景"赛道聚焦智能研发、汽车智能制造、自动驾驶、汽车智能座舱、智能造型设计、汽车营销、服务以及面向东盟应用等多个核心 领域，推动技术在汽车全产业链的落地应用。为激励创新，该赛道设置"最具潜力高手奖""面向东盟的未来独角兽企业奖"2个综合奖项，每个 奖项各5名获奖者；设置"场景应用领航奖""人工智能设计颠覆奖""智能内容创作标杆奖""智能汽车技术突破奖"4个分项奖项，每个奖项各3名 获奖者。 "揭榜挂帅"赛道以攻克产业核心技术为重要目标，聚焦人工智能+汽车领域的产业培育，联合吉利汽车研究院、广西汽车集团等大型国有 企业、行业龙头企业或园区、各重点实验室共同发布榜单。榜单内容丰富多样，涵盖汽车多模态主动感知决策大模型系统开发与应用、基于全 模态的端到端语音交互大模型开发与应用、AIGC营销内容生成平台开发、动力装备行业大模型研发与应用示范、针对封闭场景低速应用场景 自动驾驶系统开发与应用等，推动人工智能技术向生产力转化，奖金总额高达7000万元。 汽车智能研发和制造、自动驾驶、车联网与生态协同、智能营销……人工智能浪潮席卷全球，汽车产业正站在智能化变革的关键历史节 点。8月2日，记者从自治区数据局获悉，"A ...

硅谷观察/郑峻 上周五，佛罗里达州迈阿密的联邦陪审团作出裁决，认定特斯拉需要对2019年佛罗里达一起致命车祸承 担部分责任，并要求特斯拉支付总计2.43亿美元的赔偿，以阻止特斯拉未来的类似行为。 在为期三周的庭审之后，八人组成的陪审团又经过两天的讨论后认定，特斯拉对那起车祸负有三分之一 的责任，而驾驶员负有三分之二的责任。特斯拉轿车的驾驶员因低头捡拾掉落的手机造成事故，他已另 行被起诉。 陪审团认定原告遭受的痛苦和精神损失总计1.29亿美元，但鉴于特斯拉仅负部分责任，其需支付三分之 一的赔偿，即4300万美元的补偿性赔偿。再加上2亿美元的惩罚性赔偿，特斯拉需要总计向原告支付 2.43亿美元赔偿。 特斯拉首次败诉，被判支付2.43亿美元赔偿！ 那些年马斯克夸大的宣传，如今都成为了对特斯拉不利的证据。特斯拉在辅助驾驶相关诉讼的不败纪录 就此终结，或将为未来类似追责诉讼设立先例。 特斯拉首次败诉 特斯拉随后发表声明称："今天的判决是错误的，只会阻碍汽车安全的发展，危及特斯拉及整个行业开 发和实施救命技术的努力。鉴于审判中存在的重大法律错误和不规范行为，我们计划提出上诉。" 此案胜利为后续数万起诉讼开了先河。2019 ...

成都世运会执委会世运村部专职副部长王森：世运村主要是依托酒店现有的软硬件和运营团队提供 服务。我们秉持着能租不买、能借不租的原则筹备世运村，体现我们全过程的节俭方式。 为了更好服务赛会，世运村设置了多个餐厅、医疗室、健身房、母婴室、超市、快递站等。赛事期 间，世运村也在固定区域设置了班车站，配备自动驾驶巴士、AI智能翻译等科技设施，为赛会提供交 通保障。 在世运会期间，世运村里还将组织开展精彩纷呈的特色文化活动，丰富赛会生活，促进体育文化繁 荣发展和文明交流互鉴。 国际世界运动会协会副首席执行官纪尧姆·费利：非常感谢当地组委会为筹备世运村所做的工作。 让我特别感动的是，这里为运动员和参与者准备了很多文化交流体验活动，这非常好。 8月3日，来自瑞士、新西兰等20多个国家的300多名运动员及随队人员陆续抵达成都，这是成都世 运会的首个运动员入境高峰。 央广网北京8月4日消息（记者韩民权 杨妮 赵雄）据中央广播电视总台中国之声《新闻和报纸摘 要》报道，2025年世界运动会运动员村8月3日举行开村仪式，成都世运会官方抵离中心也从昨天起正式 运行。 世运会运动员村8月3日正式开村。来自国际世界运动会协会代表、参赛运动员 ...

自动驾驶之心大模型VLA技术交流群成立了，欢迎大家加入一起交流VLA相关的内容：包括VLA数据集制 作、一段式VLA、分层VLA、基于大模型的端到端方案、基于VLM+DP的方案、量产落地、求职等内容。 感兴趣的同学欢迎添加小助理微信进群：AIDriver005， 备注：昵称+VLA加群。 ...

Core Viewpoint - Tesla has been ordered to pay approximately $243 million in damages for a fatal accident involving its Enhanced Autopilot system, raising questions about the responsibility of both the driver and the company [1][5]. Group 1: Accident Details - The accident occurred when the driver, George McGe, was distracted while using the Enhanced Autopilot, leading the vehicle to ignore stop signs and red lights, resulting in a collision that killed one person and injured another [2][4]. - The driver admitted to being responsible for the dangerous driving behavior, acknowledging the risks of looking down to pick up a phone while driving [4]. Group 2: Legal and Technical Implications - Despite the driver's admission of fault, the jury attributed some responsibility to Tesla, citing the failure of the Enhanced Autopilot to perform necessary safety actions, such as braking or issuing collision warnings [5][6]. - Tesla's Enhanced Autopilot is classified as a Level 2 (L2) driver assistance system, which requires driver attention and does not qualify as fully autonomous driving [6][14]. Group 3: Industry Context and Technology Assessment - The incident highlights the ongoing confusion surrounding the capabilities of Tesla's Autopilot and the marketing of its features, which have led many users to mistakenly believe they were using fully autonomous technology [6][14]. - Tesla's current Full Self-Driving (FSD) technology has faced criticism for not being mature enough, with reports of accidents occurring even in 2023 due to its limitations [8][10]. - The article emphasizes that regardless of the technology used, whether pure vision or a combination of sensors, the responsibility for safe driving ultimately lies with the driver, especially in the absence of fully autonomous systems [14][19].

Core Points - Tesla has been ordered to pay $243 million in damages for its first loss in a lawsuit related to its Autopilot system, marking a significant legal precedent for future similar lawsuits [2][9] - The jury found Tesla responsible for one-third of the liability in a fatal accident, while the driver was deemed two-thirds responsible [2][9] - The case highlights the implications of Elon Musk's past statements regarding the capabilities of Tesla's Autopilot, which were used as evidence against the company [10][12] Summary by Sections Lawsuit Outcome - A federal jury in Miami ruled that Tesla must pay a total of $243 million, including $43 million in compensatory damages and $200 million in punitive damages [2][4] - The ruling is seen as a potential catalyst for more lawsuits against Tesla regarding its Autopilot system [2][14] Accident Details - The accident occurred in 2019 when the driver, George McGee, was distracted while using the Autopilot system, leading to a collision that resulted in a fatality [7][9] - The victim's family argued that Tesla's marketing of the Autopilot system misled drivers into believing it was fully autonomous [9][21] Marketing and Legal Implications - The lawsuit focused on how Tesla and Musk marketed the Autopilot system, with claims that they exaggerated its capabilities and safety [10][12] - Musk's past statements, including claims that Tesla's technology would surpass human driving capabilities, were central to the jury's decision [10][12] - The case reflects ongoing scrutiny of Tesla's marketing practices and the legal challenges the company faces regarding its Autopilot and Full Self-Driving (FSD) systems [21][23] Industry Context - The punitive damages awarded in this case are significant, as they could influence future litigation against Tesla and other companies in the autonomous vehicle sector [4][14] - The case underscores the challenges faced by companies in the autonomous driving space, particularly regarding consumer expectations and regulatory scrutiny [21][23]

智通财经8月4日电，马斯克谈芝加哥的机器人出租车项目，一旦在当地重新确认安全测试结果并获得运 营许可，特斯拉就将在芝加哥开展运营。 ...

Core Viewpoint - Guangzhou is making significant progress in the development of the intelligent connected vehicle industry, with substantial achievements in autonomous driving technology testing and infrastructure development [1][2][3]. Group 1: Industry Development - Guangzhou has achieved over 1.3 million hours of autonomous driving technology testing and a total mileage of 24.41 million kilometers by 17 companies [1][3]. - The city plans to issue a work guideline for unmanned driving equipment to promote various types and scenarios of autonomous driving applications across the city [1][4]. - In 2024, the proportion of autonomous driving mileage from testing vehicles is expected to reach 93.17%, an increase of 3.67 percentage points compared to 2023 [3]. Group 2: Infrastructure and Policy Support - The city has opened 1,340 general testing roads with a total one-way mileage of 2,601 kilometers, along with 10 high-speed testing roads covering 263.27 kilometers [3]. - Guangzhou has implemented supportive policies for the intelligent sensor industry and is working on a development plan for the autonomous driving industry by 2025 [2]. - The city has installed 897 sensing devices and 419 computing units, modified 205 traffic lights, and established a CA certification system, becoming the first city in China to have over 10,000 vehicle-mounted terminals for vehicle networking [5]. Group 3: Commercial Application and Testing - A total of 508 vehicles are engaged in road testing and demonstration operations, covering various models for services such as transportation, sanitation, and logistics [4]. - Guangzhou has initiated regular autonomous driving demonstration lines in the city center, allowing companies to charge market rates for their services [4]. - Companies like GAC Group and Pony.ai are actively developing autonomous driving models, with GAC's vehicles averaging nearly 30,000 kilometers of autonomous driving distance [3].

Core Viewpoint - The article emphasizes the critical role of data annotation in the development of autonomous driving systems, highlighting its impact on the performance of perception models and the overall safety of autonomous vehicles [4][14]. Group 1: Data Annotation Importance - Data annotation is essential for converting raw perception data into structured labels with semantic information, which directly influences the system's ability to recognize, understand, and make decisions in real-world environments [4][14]. - Accurate and systematic data annotation enhances the robustness and generalization capabilities of perception algorithms, making it an irreplaceable component in the autonomous driving technology ecosystem [4][14]. Group 2: Types of Data Annotation - Image data annotation focuses on identifying and locating key targets in road scenes, including vehicles, pedestrians, traffic signs, and lane markings, using methods like 2D bounding boxes, instance segmentation, and semantic segmentation [5][14]. - 3D point cloud data annotation involves higher spatial complexity, utilizing 3D bounding boxes to capture the dimensions, center points, orientations, and dynamic states of objects in three-dimensional space [7][14]. - Multi-modal data annotation is required for sensor fusion, where corresponding relationships between different modalities (e.g., images and point clouds) are established to improve recognition accuracy in complex scenarios [9][14]. Group 3: High-Precision Map Data Annotation - High-definition map data annotation involves abstracting and extracting geometric and semantic elements of road structures, such as lane boundaries and traffic signal locations, which are crucial for precise vehicle positioning and decision-making [9][14]. - The annotation process must ensure high spatial accuracy and semantic consistency with perception annotations to maintain the stability of the perception-map linkage model [9][14]. Group 4: Environmental and Behavioral Annotation - Annotation also includes describing the overall environmental state, such as road types, weather conditions, and traffic density, which aids in enhancing the model's adaptability to diverse scenarios [11][14]. - Behavioral annotation focuses on capturing the motion characteristics and intentions of dynamic traffic participants, which is vital for trajectory prediction and risk assessment [11][14]. Group 5: Quality Control in Data Annotation - Quality control is paramount in the annotation process, involving standardized guidelines, professional training for annotators, and multiple rounds of review to ensure consistency across semantic, spatial, and temporal dimensions [13][14]. - Companies often utilize self-developed annotation platforms and feedback mechanisms to create a continuous data iteration loop, enhancing the quality and relevance of the training data [13][14]. Group 6: Conclusion on Data Annotation - The core task of autonomous driving data annotation is to provide accurate, comprehensive, temporally consistent, and context-rich training samples, which are fundamental for the collaborative functioning of perception, prediction, decision-making, and control modules [14].