At the 2024 Yunqi Conference hosted by Alibaba, He Xiaopeng, Chairman and CEO of Xiaopeng Motors, appeared driving the "AI car" P7+.
It is reported that this car is equipped with an industry-leading end-to-end large model. In the past two years, the AI computing power scale jointly built by Xiaopeng Motors and Alibaba Cloud has increased by more than four times. He Xiaopeng stated that he will continue to deepen the AI computing power cooperation with Alibaba Cloud, accelerate the expansion of end-to-end large-scale models for autonomous driving, and raise the lower limit.
It is reported that in order to improve the efficiency of intelligent driving model training, Xiaopeng Motors joined hands with Alibaba Cloud to build China's largest autonomous driving intelligent computing center in Ulanqab as early as 2022, which has increased the efficiency of autonomous driving model training by more than 600 times. In the past two years, due to the rapid development of big model technology, Alibaba Cloud has expanded the computing power reserve of this intelligent computing center by more than 4 times to 2.51 Eflops, providing a stable and efficient computing power base for Xiaopeng Motors, supporting the rapid iteration of big models, and achieving "nationwide usability" of intelligent driving.
In May of this year, Xiaopeng Motors achieved end-to-end autonomous driving mass production in China and quickly landed nationwide. Xiaopeng Motors announced that in order to further maintain its first mover advantage, it will invest 3.5 billion yuan annually in research and development, of which 700 million yuan will be used for computing power training. It will also continue to deepen cooperation with Alibaba Cloud to accelerate the implementation of end-to-end large-scale models.
It is understood that Xiaopeng Motors has also deeply applied large models to cockpit scenes. Xiaopeng Motors has fully upgraded its in car assistant based on its self-developed "global language model" X-GPT and Tongyi Qianwen. Xiaopeng Motors has integrated with Tongyi Wanxiang in its official app, and in the research and development scenario, the code review adoption rate of Tongyi Lingma is as high as 50%.