Except for Baidu, there is no cloud manufacturer in China that can have leading self-developed technologies, products and ecology at every level. Baidu’s Kunlun chip in the AI ​​IaaS layer and the deep learning framework Fei Pao and Wenxin large model in the AI ​​PaaS layer have enabled Baidu to form a self-contained closed-loop intelligent path of “chip-framework-large model-industry application”. Truly end-to-end optimization.

New strategy: cloud and intelligence integration, in-depth industry

Based on the new strategy of “integration of cloud and intelligence, in-depth industry”, Baidu Smart Cloud released a new upgrade of the “integration of cloud and intelligence 3.0” architecture. The 3.0 architecture starts from the core scenarios of the industry. By creating industry benchmark applications, it drives and precipitates the capabilities of the AI ​​PaaS layer and the AI ​​IaaS layer, creating cost-effective heterogeneous computing power and efficient AI development and operation capabilities, and then optimizes existing ones. Apply and incubate new applications, transform the digital base downward, make the basic cloud more suitable for AI applications, and form the effect of spiral upward and continuous evolution.

Shen Dou said: “In the future, in order to intelligently solve the core needs of the industry, we must go deeper into the industry. The rich application scenarios in the industry provide the broadest space for the development of artificial intelligence and cloud.”

It is understood that the AI ​​cloud composed of AI IaaS and AI PaaS brings together Baidu’s leading self-developed technologies at all levels of artificial intelligence, allowing Baidu’s core self-developed technologies to be fully applied in the industry.

Self-developed technology runs through the cloud-intelligence integration 3.0 architecture

Baidu Smart Cloud has always insisted on independent research and development in terms of technology. Shen Dou said that cloud-intelligence integration 3.0 has formed a set of intelligent closed-loop paths of “chip-framework-large model-industry application” to achieve end-to-end optimization.

In the AI ​​IaaS layer of Cloud Intelligence 3.0, Kunlun Chip 2 has been deployed in Baidu search, autonomous driving, iQiyi and other businesses, as well as customers in industries such as finance and industry. As a 7-nanometer general-purpose GPU, the performance of the second generation of Kunlun core is up to 3 times higher than that of the first generation, and the cost performance is better than that of foreign products of the same level. In industrial quality inspection scenarios, it has been able to replace non-domestic chips, reducing costs by 65%. At present, the third generation of Kunlun chips is under research and development, and it is expected to be mass-produced in 2024, which will become a substitute for domestic high-end chip demand.

With the support of Kunlun Core, the Baige AI heterogeneous computing platform has been upgraded to Baige 2.0, which improves AI computing power in line with the needs of industrial intelligent development. Through the application of Baige 2.0, the training efficiency of the drug protein structure prediction model has been doubled, and the automatic driving iteration cycle of mass-produced vehicles has been shortened from the monthly level to the weekly level.

Under the strong support of the Flying Paddle deep learning framework and the Wenxin industry model, the overall AI PaaS capability of Baidu Smart Cloud has been greatly improved. The comprehensively improved AI PaaS has evolved into an “AI big engine”, which can make full use of AI computing power and reduce Use thresholds to speed up model iteration.

At the meeting, Baidu Smart Cloud launched the Smart Computing Center 1.0 solution. The intelligent computing center 1.0 has large-scale training scenarios, which can achieve low energy consumption and high-performance operation, and meet the development needs of advanced science and technology industries such as local urban brains, life sciences, and autonomous driving. Industry applications of Baidu Smart Cloud usher in a comprehensive upgrade

Baidu Smart Cloud released the first car cloud. From the car enterprise group cloud, the network connection cloud, and the supply chain collaboration cloud, the three levels are deeply involved in the digital upgrade of the automobile manufacturing industry, and solve the three major digital application problems of automobile industry production, automatic driving testing, and supply chain management. Baidu Smart Cloud is becoming an invisible digital force behind the rise of China’s auto industry. China’s top 15 automakers by sales, including FAW, Geely, Dongfeng Motor, and other 10 are Baidu Smart Cloud customers; among the top ten new energy vehicle companies, Baidu Smart Cloud serves five.

Kaiwu Industrial Internet platform ushered in a new upgrade. Kaiwu 2.0 has accumulated more than 200 industrial solutions and deposited 38,000 industrial models, which have been implemented in Chongqing, Guangzhou, Suzhou and other cities across the country. Turn AI into a production tool within the reach of SMEs.

The “Kyushu” district and county brain is the country’s first lightweight urban brain solution. As the country’s first lightweight urban brain solution, “Kyushu” can shorten the deployment time by more than 50%, shortening the deployment cycle efficiency from months to weeks. Through elastic expansion, “Jiuzhou” can quickly respond to the digital development needs of districts and counties, superimpose the application of digital agriculture, digital cultural tourism and other scene business features at any time, and quickly sink the governance experience of first-tier cities to districts and counties.

Launched industrial digital financial solutions to support the entire industry to cross the “digital divide”. Based on Baidu Netdisk, Baidu Smart Cloud has launched a personal cloud SME digital solution, providing a total of 100,000 SMEs with digital upgrade services such as storage, management, and sharing, and continuously lowering the threshold for SMEs to use AI technology.

It is a heavy historical mission to go deep into the industry and accelerate the deep integration of digital technology and the real economy. Facing the future, let us work together to enter a new era of China’s industrial intelligence.

Reviewing Editor: Peng Jing

Leave a Reply

Your email address will not be published.