Baidu AI Cloud Releases New Generation Cloud Strategy
At the 2022 Intelligent Economy Summit held on September 6, Baidu AI Cloud, an intelligent cloud computing brand under Chinese technology giant Baidu, released a new strategy consisting of the “Integration of Cloud and Intelligence, Deepening the Industry” and “Integration of Cloud and Intelligence 3.0”.
Dou Shen, executive vice president of Baidu and chief of Baidu AI Cloud, said at the forum: “Aside from Baidu, there is no other cloud service provider in China that has such leading technologies developed in-house, nor similar products nor an ecology in every field of cloud services. Baidu has Kunlun chip in the AI IaaS (infrastructure as a service) field and PaddlePaddle, a deep learning framework, and Wenxin, a machine learning model in AI PaaS (platform as a service) field. These three products make Baidu, form a closed-loop intelligent path of “chip – framework – model – industry application”, which truly achieves end-to-end optimization. “
In the AI IaaS field of Integration of Cloud and Intelligence 3.0, Baidu‘s self-developed AI chip Kunlun (2nd generation) has been deployed in Baidu‘s search engine, automated driving, video streaming platform iQIYI and other businesses, as well as customers in financial and industrial fields. As a 7-nm general GPU, the performance of Kunlun (2nd generation) is up to 3 times higher than that of the 1st generation, and its cost performance is better than that of foreign products of the same level. In terms of industrial quality inspection, the chip has been able to replace foreign-made chips and reduce costs by as much as 65%. At present, Kunlun (3rd generation) is already under R&D and it is expected to be mass-produced in 2024. The chip will become a substitute product for domestic high-end demand.
With the support of Kunlun, Baidu AI Cloud’s AI heterogeneous computing platform Baige has been upgraded to version 2.0, which closely follows the needs of industrial intelligent development and overall improved AI computing power. Through the application of Baige 2.0, the training efficiency of drug protein structure prediction models has been improved twofold, and the iteration period of mass-produced automated driving vehicles is shortened from months to weeks.
SEE ALSO: Baidu’s Robin Li: Next Commercial Stage of Autonomous Driving Is L4, Not L3
At the conference, Baidu AI Cloud launched the 1.0 version of its Intelligent Computing Center. It supports large-scale training, consumes low amounts of energy but offers high performance operations, and should meet the development needs of advanced science and technology industries such as “city brain,” life sciences and automated driving in local cities.
At the forum, Baidu AI Cloud released the automobile cloud for the first time, which covers three levels of the automobile manufacturing industry: automobile enterprise, network connections and supply chain collaboration. It aims to solve digital application problems including automobile production, automated driving tests and supply chain management.