Workflow
图形学与AI融合
icon
Search documents
英伟达为机器人推出懂推理的“大脑”!升级版Cosmos世界模型来了
量子位· 2025-08-12 09:35
Core Viewpoint - NVIDIA is making significant advancements in robotics by integrating AI and computer graphics, aiming to create a comprehensive virtual environment for robot training and development [16][19][21]. Group 1: Product and Technology Updates - NVIDIA introduced the upgraded Cosmos world model at the SIGGRAPH conference, which is designed to generate synthetic data that adheres to real-world physics [1][2]. - The upgrade focuses on planning capabilities and generation speed, with enhancements in both software and hardware, including the new Omniverse library and RTX PRO Blackwell servers [3][5]. - The new Cosmos Reason model features 70 billion parameters and reasoning capabilities, serving as a "brain" for robots to assist in task planning [6][9]. Group 2: Software Enhancements - The Cosmos Transfer-2 model accelerates the conversion of virtual scenes into training data, significantly reducing the time required for this process [11]. - A streamlined version of Cosmos Transfer-2 simplifies the data generation process from 70 steps to just 1 step, allowing for faster operation on RTX PRO servers [12]. - The Omniverse library now supports compatibility with various robot simulation formats, enabling easier simulation across different platforms [13]. Group 3: Hardware Developments - NVIDIA launched the RTX PRO Blackwell server to support various workloads related to training, synthetic data generation, and robot learning [15]. - The DGX Cloud platform allows developers to run Omniverse and Cosmos applications efficiently from the cloud, reducing hardware barriers and operational costs [15]. Group 4: Strategic Focus and Market Position - NVIDIA is heavily investing in the robotics sector, recognizing the fusion of computer graphics and AI as a transformative force in the industry [16][19]. - The company aims to create a realistic and scalable "virtual parallel universe" for robots to safely experiment and evolve, thereby lowering the costs and barriers associated with robot training [21][22]. - NVIDIA's strategy includes partnerships with various Chinese companies and cloud services, indicating a strong commitment to expanding its influence in the robotics market [25][26].