Information Center

Baidu Yangquan Intelligent Computing Service Center is comprehensively upgraded and actively supports Jingxing Yiyan's market demand for large-scale intelligent network resources


With the explosion of AI applications such as ChatGPT and Jingxing Speakers, the market demand for network resources required for supporting AI has also soared. Baidu said that it would release the generative AI product "Jingxing Speakers" on March 16. In order to actively support the market demand of Jingxing Speakers for large-scale sorting, and further simultaneously achieve the approval of Jingxing Speakers' industrial development, Baidu Intelligent Cloud has also frequently announced the preparation of Jingxing Speakers' service facilities and facilities recently.

On March 14, various aspects of Baidu's intelligent cloud said that since the end of last year, Baidu's intelligent cloud has comprehensively upgraded its cloud service skills through three major actions: in December last year, it announced the first domestic "AI Big Base" of the whole stack self researched AI infrastructure, in February this year, it upgraded its AI research and development operation (MLOps) skills, and in March, Baidu Yangquan Intelligent Computing Service Center completed the upgrade. It is revealed that the network resource infrastructure behind "Jingxing Speakers" is actively supported by Baidu Intelligent Computing Service Center, and the subsequent multiple intelligent computing service centers of Baidu will also provide more underlying support for the approval of the industrialization of "Jingxing Speakers" for national industrial development.

The development of generative AI industry has been approved and has encountered three challenges, and the rules of cloud market are changing

Behind the boom of generative AI, the value of cloud sorting for private enterprises has undergone a fundamental qualitative change: from providing more infrastructure to providing more intelligent services, which has also consolidated more possibilities for the combination of generative AI goods and industrial development. Li Yanhong, the founder and CEO of Baidu, said that "Jingxing Speakers" is a big mathematical model created by the basic concept of Baidu's intelligent cloud core technology, which will fundamentally change the legal provisions of the cloud market.

This kind of subversive change also means that generative AI products such as Jingxing Speakers face multiple challenges in the approval of industrial development and industrialization. Specifically, the challenges are mainly reflected in the following three aspects: first, intelligent network resources. The data volume of generated AI products is huge, and the operation requires unprecedented AI network resources, which challenges the existing sorter architecture. The second is AI infrastructure. The sorting scenarios of artificial intelligent commodities such as Jingxing Speakers are complex and the sorting architecture is multidimensional. Private enterprises need to integrate chip, architecture, mathematical models and application fields into the whole stack AI infrastructure, and the whole process of end-to-end intelligent closed-loop AI research and development. Finally, it is the level of AI industrialization. When AI generative products are approved for industrial development practice, the basic concept and specific business development scenarios are needed to carry out secondary network connection research and development. However, the whole process of research, development and operation of AI mathematical model is highly complex, and any non-standard operation will affect the effect of AI mathematical model. The level of AI industrialization determines whether AI can break through the "final mile" approved for industrial development.

In the face of the new legal provisions of the cloud market, only cloud suppliers who solve the above three challenges for private enterprises are eligible for admission tickets. Shen Shao, executive vice president of Baidu Group and president of Baidu Intelligent Cloud Business Group, said that cloud services have been transformed from the digital era to the intelligent era. Previously, private enterprises chose cloud providers more for network resources, storage and other basic cloud services. In the future, the market demand of private enterprises for cloud will focus more on the level of intelligent services.

Jingxing spokesperson is about to announce that Baidu's intelligent cloud upgrades cloud services of three major service facilities

In the era of intelligent computing, private enterprises have encountered many challenges, such as the shortage of intelligent network resources, the complexity of the whole process of AI research and development, and the difficulty of AI industry development being approved for industrialization, Baidu Intelligent Cloud gave the answer by virtue of its unique advantage of "birth of cloud intelligence". Baidu said that it would provide more core skills in three aspects for industrial development, and help the intelligent transformation of industrial development onto the fast lane.

First, provide private enterprises with more massive high-performance intelligent network resources to actively support the industrial development of AI products such as Jingxing Speakers. Baidu has built intelligent computing service centers in Yangquan, Shanxi, Yancheng, Jiangsu and other places. Among them, Baidu Yangquan Intelligent Computing Service Center is the largest single intelligent computing service center in Asia, with a construction scale of 4 EFLOPS (40 billion floating point operations per second) AI network resources, which can meet the market demand for large-scale AI sorting in various industries. Song Chunxiao, the head of Kunlun Chip's science and technology strategy, said: "AI chip is the core of network resources. Kunlun Chip 2 has been widely approved in the application field of Baidu Jingxing's mathematical model, and provides more AI network resources for intelligent upgrading in all walks of life."

It is understood that Baidu Intelligent Computing Service Center continues to innovate the ranking structure, actively supports the ranking tasks of different business development scenarios of private enterprises during the intelligent computing period, and improves the work efficiency and innovation skills of business development of private enterprises. At present, Baidu Intelligent Computing Service Center has actively supported the advanced AI application fields such as Jingxing pre training mathematical model, biological sequencing, and automatic driving. At the same time, the basic concept, self-developed and innovative core technology can reduce PUE to 1.08, and achieve efficient and energy-saving operation, thus reducing customers' electricity charges and operation and maintenance costs.

Second, provide more new AI infrastructure for private enterprises, and realize cost reduction and efficiency increase of AI research and development. Baidu Intelligent Cloud launched the "Baidu AI Big Base" by encapsulating network resources, architecture, mathematical models and AI application fields. From high-end chip Kunlun core, to deep learning architecture of flying oars, to Jingxing pre training big mathematical model, and then to AI application field, we also realize end-to-end intelligent closed-loop, so as to turn the high threshold AI core technology into the same as hydropower for private enterprises to use on demand, greatly reducing the research and development costs of private enterprises and improving work efficiency. Song Fei, general manager of Baidu Intelligent Cloud Sorting Product Solution and Operation Department, said: "Baidu AI Base can continuously optimize the basic concept and actual business development data, so that the resource utilization rate is increased to 70%, and the R&D efficiency of private enterprises is improved by 100%."

In the end, it will provide more AI R&D and operation bodiless (MLOps) skills for private enterprises, and accelerate the rapid industrial development of big mathematical model commodities such as generative AI. Xin Zhou, director of Baidu AI middle office, said that Baidu AI big base will provide a series of AI R&D, operation and maintenance tools for private enterprises nationwide. After accessing Jingxing Speakers, private enterprises can easily complete network connection and secondary research and development with business development scenarios at low cost, and help industrial development break through the "final mile" approved by AI through AI industrialization skills.

Supported by low-cost and multi framework intelligent network resources, new AI infrastructure of "AI Big Base" and MLOps AI industrialization skills, Jingxing speakers will accelerate the arrival of intelligent industrial development and bring real AI benefits.