融资14亿美元，Skild AI打造跨任务、跨硬件形态的通用具身大脑

Core Insights - The development of AI progresses from specialization to generalization, eventually entering valuable niche scenarios based on general capabilities [1] - Skild AI aims to create a universal robotic brain, Skild Brain, that can adapt to any hardware form and perform various tasks, moving beyond the limitations of previous generations of robots [6][4] Funding and Valuation - Skild AI secured $300 million in Series A funding in 2024 and recently completed a $1.4 billion Series B funding round, led by SoftBank, with a valuation exceeding $14 billion [3] Founders and Team - The founders, Deepak Pathak and Abhinav Gupta, have over 110,000 paper citations and have led significant breakthroughs in robotics over the past decade [4] - The team comprises experts from leading tech companies and universities, enhancing its capabilities in embodied intelligence [4] Technological Advancements - Skild Brain is designed to control various types of robots without prior knowledge of their physical forms, enabling it to handle tasks ranging from simple household chores to complex physical challenges [6] - The architecture of Skild Brain includes a layered structure for operation and navigation strategies, allowing seamless transitions between different tasks [8] Learning and Adaptation - Skild Brain utilizes end-to-end motion control driven by online visual and proprioceptive feedback, enabling real-time adaptation to new environments [9][11] - The model incorporates context learning, allowing robots to learn from failures and adapt to unfamiliar situations, enhancing their versatility [12][15] Data Infrastructure - Skild has built a robust data infrastructure that allows Skild Brain to learn from various sources, including large-scale simulations and real-world deployments [16][17] - The company has achieved significant revenue growth, reaching $30 million in a few months through deployments in security, construction, and warehousing [19] Future Outlook - The company aims to bridge the generalization gap in physical AI, potentially leading to exponential value creation as it matures and expands its applications [20]