Core Insights - DaXiao Robotics has launched its first open-source and commercially applicable Kairos 3.0 world model, along with the embodiment super brain module A1, which aims to significantly reduce data collection costs in the industry [1][2] Group 1: Company Overview - DaXiao Robotics is co-founded by Wang Xiaogang, a notable figure from SenseTime, and Tao Dacheng, former chief scientist at JD Exploration Institute, attracting considerable attention from the industry [1] - The company addresses the critical shortage of training data in the field of embodied intelligence, which has traditionally relied on costly and inefficient data collection methods [1] Group 2: Innovation in Data Collection - The newly proposed "environmental data collection" approach utilizes a human-centered data collection system, integrating multi-modal data from various perspectives, including visual, tactile, and mechanical properties [1][2] - This method can achieve data collection of tens of millions of hours annually, significantly lowering costs and enhancing efficiency [2] Group 3: Technological Advancements - The Kairos 3.0 world model represents a shift from imitation to reasoning, seen as a key evolutionary direction for embodied intelligence, enabling machines to better understand physical world dynamics and human behavior [2] - The model incorporates multi-dimensional inputs, allowing for a deep understanding of physical laws and the underlying logic of human actions, thus providing a high-fidelity virtual training environment [2] Group 4: Product Features and Ecosystem - The Kairos platform supports 11 major categories and 54 subcategories, with a total of 328 tags, covering 115 different embodied scenarios, enabling developers to quickly generate visual task simulations [3] - DaXiao Robotics has opened its API to the industry, promoting the rapid emergence of lightweight and customized embodied intelligence products, thereby fostering ecosystem development [3]
突破大语言模型局限 大晓机器人助推具身智能从“模仿”走向“推理”