“物理AI的ChatGPT时刻”！英伟达最新发布 黄仁勋发声

Core Insights - Nvidia has taken a significant step in the autonomous driving sector by open-sourcing its first inference VLA (Vision-Language-Action) model, Alpamayo, aimed at accelerating the development of safe autonomous driving technology [2][4] - The Alpamayo model processes complex driving scenarios using human-like reasoning, providing new pathways to address the long-tail problem in autonomous driving [2][4] Group 1: Model Features and Capabilities - The Alpamayo model is designed to enable vehicles to "think" in unexpected situations, such as traffic light failures, by analyzing inputs from cameras and sensors to propose solutions [4][5] - It integrates three foundational pillars: an open-source model, a simulation framework, and datasets, creating a comprehensive open ecosystem for automotive developers and research teams [4][5] - Alpamayo 1 features a 10 billion parameter architecture that generates trajectories and reasoning paths from video inputs, showcasing the logic behind each decision [4][6] Group 2: Future Developments and Applications - Nvidia emphasizes that the Alpamayo model will not run directly in vehicles but will serve as a large-scale teacher model for developers to fine-tune and integrate into their complete autonomous driving technology stack [5][6] - Future models in the Alpamayo family are expected to have larger parameter sizes, enhanced reasoning capabilities, and more flexible input-output options for commercial use [5][6] - The reasoning VLA model can break down complex tasks into manageable sub-problems, providing a more accurate problem-solving approach and a degree of introspection on its operations [6][7] Group 3: Strategic Initiatives and Market Position - Nvidia plans to test a self-driving taxi service by 2027, indicating its commitment to advancing autonomous vehicle technology [7] - The company is also set to release new Rubin data center products that will significantly enhance AI training and inference performance, with improvements of 3.5 times and 5 times, respectively, compared to the previous Blackwell architecture [8] - Major cloud service providers, including Microsoft, are expected to be among the first to deploy the new hardware based on the Rubin architecture [8]