VLA(视觉语言动作模型)

Search documents
对话华为靳玉志:世界上根本没有免费的东西
Di Yi Cai Jing Zi Xun· 2025-08-26 09:48
Core Viewpoint - Huawei's automotive business is rapidly expanding, particularly in the field of assisted driving, with various collaboration models with car manufacturers being explored [1][2]. Group 1: Huawei's Automotive Business Development - Huawei's automotive business unit (BU) has seen key figures, such as CEO Jin Yuzhi and Huawei Qian Kun, publicly engaging with the media, indicating a shift from behind-the-scenes to a more prominent role [1]. - The assisted driving sector is experiencing significant growth, with Huawei's Qian Kun technology being integrated into 1 million vehicles and the upcoming large-scale deployment of Huawei ADS 4 [1]. - The collaboration models between Huawei and car manufacturers include component supply, single intelligence (either smart cockpit or assisted driving), dual intelligence (both smart cockpit and assisted driving), and full-stack solutions [1][7]. Group 2: Technical Perspectives and Strategies - Huawei Qian Kun does not align with the VLA (Vision-Language-Action) approach favored by some manufacturers, instead prioritizing WA (World and Action) for direct vehicle control through sensory inputs [2][9]. - The company emphasizes that there is no such thing as a free lunch in the automotive industry, suggesting that costs are often hidden in other areas, such as vehicle pricing [2][12]. Group 3: Collaboration and Empowerment - Huawei's strategy involves a comprehensive support system for car manufacturers, guiding them through the entire process from product definition to marketing, thereby enhancing their capabilities [7][8]. - The company aims to empower manufacturers, particularly state-owned enterprises, by sharing its successful experiences in product development and market strategies [7][8]. Group 4: Future Outlook and Market Dynamics - The number of players capable of achieving L3 and L4 autonomous driving is expected to decrease as the industry consolidates, driven by the need for data, computational power, and algorithms [10]. - The differentiation in assisted driving technology is minimal, with the primary goal being zero accidents and fatalities, which complicates pricing strategies [11].
拆解特斯拉机器人供应链:30 多位从业者看到的泡沫和希望
阿尔法工场研究院· 2025-06-08 13:36
Core Viewpoint - The article discusses the current state and challenges of humanoid robot development, particularly focusing on Tesla's efforts in this field, highlighting the significant investments and technological hurdles involved in creating robots that can effectively replace human labor in various tasks [1][3][30]. Group 1: Market Dynamics and Investment - Since Tesla's first public demonstration of humanoid robots in October 2022, the A-share robot concept sector has surged by 93%, while the Shanghai and Shenzhen 300 Index has only increased by about 1% [1]. - Global investors and suppliers have invested over 100 billion yuan in humanoid robots since Tesla's announcement, indicating strong market interest despite the current production challenges [2][3]. - Tesla's supply chain for humanoid robots has been built over three years, involving hundreds of companies, with significant contributions from both established and emerging suppliers [29][30]. Group 2: Technical Challenges - Humanoid robots currently lack the efficiency to replace blue-collar workers, with production still heavily reliant on manual assembly, and the expected output for the global humanoid robot industry this year is only around 20,000 units [2][3]. - The complexity of human-like movement requires advanced engineering solutions, such as the use of planetary roller screws and high-precision sensors, which are costly and difficult to manufacture [10][11][36]. - The current design of Tesla's humanoid robots includes a height of 1.72 meters, weight of 55 kg, and a cost of approximately $60,000, which is significantly higher than the target of $20,000 to $30,000 [30][41]. Group 3: Production and Supply Chain - Tesla's production of humanoid robots involves a complete redesign of components, with around 7,000 parts needed, which complicates the manufacturing process compared to electric vehicles [60][61]. - The supply chain for humanoid robots is characterized by rapid iteration and high precision requirements, with many suppliers struggling to meet Tesla's standards [25][29]. - The reliance on advanced materials and components, such as aluminum alloy structures and specialized sensors, adds to the complexity and cost of production [31][32][36]. Group 4: Future Prospects and Market Applications - Despite the challenges, there is optimism in the market, with some investors viewing humanoid robots as a potential source of emotional value rather than direct labor replacement [58][60]. - Tesla has adjusted its production targets for humanoid robots, pushing the timeline for mass production from 2027 to 2029, reflecting the ongoing difficulties in achieving the desired technological advancements [60][61]. - The current applications of humanoid robots are limited, with many being used for promotional purposes rather than practical labor, indicating a need for further development before widespread adoption [60].