能折纸，还会灌篮！谷歌发布机器人基座大模型，大幅强化机器人通用性

Core Viewpoint - The release of Google's DeepMind's new AI model, Gemini Robotics, marks a significant milestone in the development of general-purpose robots, enhancing their ability to adapt to complex environments and perform challenging tasks [1][9]. Group 1: Technological Advancements - The new AI model allows robots to perform tasks such as folding paper, organizing desks, and even dunking a mini basketball, showcasing its advanced capabilities [3][4][6]. - The Gemini Robotics model is reported to have double the versatility of previous models, representing a major leap towards general-purpose robotics [9]. - The model is trained using Google's Gemini 2.0 language model, endowing robots with three key abilities: environmental adaptability, instruction comprehension, and operational flexibility [10]. Group 2: Market Potential - Analysts predict a significant market expansion for humanoid robots, with an estimated annual sales of 1 million units by 2030 and a total ownership of 3 billion units by 2060, equating to 0.3 robots per person [13]. - Major tech companies, including Tesla and OpenAI, are racing to develop AI capabilities for robots, indicating a competitive landscape in the robotics sector [13]. - NVIDIA's CEO has stated that this technology could create a market worth trillions of dollars, potentially leading to the largest tech industry in history [13].