当一台四足机器人的四条腿被电锯全部锯断，它没有停止工作，而是继续用残存的机身向前蠕动。 这不是科幻电影，而是机器人创业公司Skild AI在X上发布的真实视频。屏幕上不断闪现 的"Adapting…"，宣告了一个机器人新时代的来临。 后排居中两位男性为Skild AI 创始人：左 迪帕克·帕塔克 ，右 阿比纳夫·古普塔 Skild AI 是卡内基梅隆大学的衍生公司，成立于2023年5月 ， 致力于构建能够安装到各种机器和机 器人设备上的人工智能系统，称为通用大脑。 创业邦 获悉， Skild AI在短短两年内已完成三轮融资，总额超4亿美元。 更令人惊叹的是其估值飙升速度——从2024年7月A轮的15亿美元，跃升至2025年6月的45亿美元， 一年内翻了三倍。 这家初创公司的投资方阵容堪称"全明星"：软银集团、贝佐斯、红杉资本、光速创投、Coatue等顶 级机构齐聚，甚至卡内基梅隆大学也以战略投资者身份入场。 "两位创始人是机器人技术的'催化剂'，过去几年该领域取得的一些进步，几乎都可以追溯到他们的学 术成果。"一位参与投资尽调的VC合伙人如此评价。 两位教授，缔造最快独角兽 Skild AI的背后，是两位来自 ...

Core Viewpoint - Sergey Brin, co-founder of Google, expresses skepticism about humanoid robots, emphasizing that while the hardware is impressive, the software often falls short. He believes that focusing too much on humanoid forms may underestimate AI's adaptability [1][2]. Group 1: Trends in Robotics - Brin notes that Google has acquired and sold several robotics companies, including Boston Dynamics, and acknowledges the current surge in robotics startups focusing on system development [1][2]. - He highlights that many people are drawn to humanoid robots because they mimic human activities, but he questions the necessity of humanoid forms for effective operation [2][3]. - Major players in the embodied intelligence sector include Tesla, known for its humanoid robot Optimus, and Nvidia, which aims to create a robust development ecosystem [3]. Group 2: Gemini Robotics Models - Google DeepMind has launched two models based on Gemini 2.0: the VLA model for understanding and executing new tasks without specific training, and the VL model for enhancing physical world understanding [5][12]. - The Gemini Robotics models are characterized by their ability to quickly adapt to new environments and tasks, natural language understanding, and the capability to perform complex, precise operations [12]. Group 3: Industry Challenges and Perspectives - Brin and other industry leaders, including Carolina Parada from DeepMind, emphasize that many impressive robot demonstrations are often pre-programmed and do not reflect true intelligence or adaptability [9][10]. - The industry consensus is that while humanoid robots have potential, significant breakthroughs in their "brains" and practical applications are still needed for meaningful commercialization [16][17].