SenseTime expects to ship over 1,000 robotic dogs next year, expanding into diverse scenarios such as cultural tourism, retail, and smart cities

With the acceleration of artificial intelligence (AI) technology iteration, multimodal large models are becoming the focus of industry competition. Wang Xiaogang, co-founder of SenseTime (020) and chairman of DaXiao Robotics, revealed that the upcoming mechanical dog is equipped with a navigation box that does not rely on high-precision maps and can achieve outdoor autonomous navigation based on ordinary electronic maps, supporting remote command issuance and multi-machine collaborative management. He also mentioned that through an integrated product route of software and hardware, the company expects to ship over a thousand mechanical dogs next year and expand into diverse scenarios such as cultural tourism, retail, and smart cities, with customer pilot projects already initiated in the cultural tourism and smart city sectors. Consumer-grade mechanical dogs priced from tens of thousands to hundreds of thousands DaXiao Robotics is an independent robotics company incubated by SenseTime, which will launch China's first open-source and commercially applicable "Enlightenment" world model 3.0 this month. Regarding the commercialization model of DaXiao Robotics, Wang Xiaogang stated that the short-term focus is on easily monetizable performance and inspection robots, including four-wheel-drive mechanical dogs; the medium-term goal is to enter commercial service scenarios such as industrial handling and retail sorting; and in the long term, they hope to enter the home service market in about five years. In terms of pricing, Wang Xiaogang mentioned that the current hardware costs in the industry are relatively high, with consumer-grade mechanical dogs priced from tens of thousands to hundreds of thousands of yuan, and they plan to reduce the price of To B (business-to-business) scenario products to around 100,000 yuan through scenario-based hardware improvements and large-scale production. In the face of global chip supply chain fluctuations, SenseTime co-founder and chief scientist of AI infrastructure and large models, Lin Dahua, stated that SenseTime started domestic chip adaptation planning six years ago and has now formed an efficient model adaptation methodology, with new models capable of completing domestic chip deployment within 1 to 2 weeks. He revealed that the current proportion of computing power from domestic chips has reached double digits and is on a continuous upward trend. New AI model expected to launch in spring next year Regarding model implementation, Lin Dahua pointed out that the core lies in shifting from "generative AI" to "creative AI." Many current large models have reached a certain level of general capability, but they still need to meet actual industry demands and cross over to "99-point professional level" to truly be helpful. He stated that SenseTime has currently completed the construction of basic reinforcement learning capabilities, and the next step will focus on advancing "AI autonomous growth," with the goal of launching a new model in spring next year, which will achieve significant upgrades in performance compared to existing versions. Given the different development paths of AI models in the East and West, Lin Dahua emphasized that technology must "break industrial red lines" and reach practical levels for commercial closed-loop formation. He also admitted that the realization of generative artificial intelligence requires long-term technological accumulation, and in the short term, the industry will still focus on scenario-based applications