GOVLA具身大模型
Search documents
细扒!入选全球前2%顶尖科学家榜单的6位中国具身智能大佬及背后技术布局!
机器人大讲堂· 2025-09-25 10:07
当前,全球具身智能领域的竞争正日益聚焦于核心人才的储备与创新能力。近日,斯坦福大学联合爱思唯尔发 布的全球前 2%顶尖科学家榜单,为中国在该领域的高水平科研力量提供了重要参照。 据不完全统计,本次名单共有来自跨维智能、智平方、 智元、 星海图、银河通用5家中国企业的6位相关科学 家入选, 他们不仅在学术研究上取得了国际瞩目的成果,更在推动具身智能技术产业化落地中扮演着关键角 色 ,反映出中国在具身智能前沿研究与产教融合方面的持续进展。 跨维智能创始人贾奎,排名为57162, 现任香港中文大学(深圳)终身教授,同时担任广东省"珠江人才计 划"创新创业团队带头人,并入选中组部第十二批海外高层次青年人才,是在人工智能、具身智能、计算机视 觉及机器学习领域的权威专家。学术方面,贾奎教授成果丰硕,已发表百余篇顶级论文,曾获CVPR 2019最 佳论文候选、ICDP最佳论文奖及广东省自然科学一等奖,并担任TIP、TMLR等顶刊副主编,以及ICCV、 ICML、NeurIPS等国际顶会主席,在学界享有广泛影响力。 在 技 术 研 发 与 实 践 领 域 , 贾 奎 教 授 成 绩 卓 著 。 他 首 创 Fantasia3 ...
半导体产线迎“硅基打工人”未来三年千台机器人上岗
Nan Fang Du Shi Bao· 2025-09-11 23:06
Core Insights - The collaboration between Zhifang Technology and Huike aims to deploy over 1,000 humanoid robots in Huike's global production bases over the next three years, marking a significant entry of embodied intelligence into the semiconductor display industry [2][3] - This deal is expected to be the largest single order in the global embodied intelligence sector, potentially reaching nearly 500 million RMB based on industry pricing estimates [2][3] Industry Developments - The initial application of the robots will focus on complex PCB operations, with plans to expand to logistics, material handling, assembly, and quality testing [3] - The flexibility of Zhifang's GOVLA model allows robots to adapt to various production environments without extensive modifications to existing lines, facilitating a shift from "humans adapting to machines" to "machines adapting to humans" [3] Data Utilization and Model Improvement - The deployment of robots in factories serves as a critical training ground for improving the performance of large models, with the accumulation of real-world production data enhancing the robots' capabilities [4] - Zhifang's approach emphasizes a "data closed-loop" system, where the data collected during operations will feed back into model iterations, creating a "smarter" robotic system over time [4] Future Commercialization Strategy - Zhifang plans to follow a path similar to that of smart automotive development, starting with semi-structured scenarios and gradually moving towards unstructured environments [5] - The company identifies high-end manufacturing and diverse public service scenarios as the most commercially valuable areas for future growth [5]
千台机器人“上岗”面板厂,智平方获惠科大单,金额或达5亿
Nan Fang Du Shi Bao· 2025-09-11 09:24
Core Insights - The collaboration between Zhifang Technology and Huike's subsidiary marks a significant entry of embodied intelligence robots into the semiconductor display industry, with over 1,000 robots expected to be deployed in the next three years [2][3] - This deal could potentially reach nearly 500 million RMB, making it the largest single order in the global embodied intelligence sector [2] Group 1: Strategic Collaboration - Zhifang Technology and Huike have announced a comprehensive strategic partnership, focusing on deploying robots in complex PCB operations as the initial demonstration scenario [3] - The robots will gradually cover the entire production process, including warehousing logistics, material handling, component assembly, and quality inspection [3] Group 2: Industry Context - The semiconductor display industry's flexible production processes have been challenging for traditional automation, which relies on preset programs and struggles with mixed-model production [5] - The GOVLA embodied model developed by Zhifang allows for full-body coordinated control of robots, enabling them to move flexibly across production lines without major modifications [5] Group 3: Data Utilization and Future Path - The industry consensus is that data is crucial for humanoid robots to become effective workers, with the most valuable data coming from real industrial environments [6] - Zhifang aims to create a "data closed-loop" through continuous operation on Huike's production line, allowing the robots to accumulate real-world data that will enhance the model's intelligence [6] - The CEO of Zhifang envisions a commercial path similar to that of smart vehicles, starting with semi-structured scenarios and progressing to unstructured environments, focusing on high-end manufacturing and diverse public service applications [6]