The list of Beijing’s General Artificial Intelligence Industry Innovation Partner Program was announced, and JD Technology was selected as a “Computing Power Partner”

On July 2, at the Artificial Intelligence Summit Forum of the 2023 Global Digital Economy Conference, the second batch of the Beijing General Artificial Intelligence Industry Innovation Partnership Program (hereinafter referred to as the "Partnership Program") was announced. Products and operational capabilities such as cost-effective computing resources, storage resources, general computing resources, and AI development platforms have become the second batch of "computing partners" in the partnership program.

The partner program is jointly initiated by the Beijing Municipal Bureau of Economy and Information Technology, the Beijing Municipal Science and Technology Commission, the Zhongguancun Management Committee, and the Beijing Municipal Development and Reform Commission. It recruits computing power partners, data partners, model partners, application partners, and investment partners through various channels. partners, aiming to build an open cooperation platform for artificial intelligence large models, establish a collaborative cooperation mechanism, significantly improve high-quality computing power and high-quality data supply support capabilities through continuous optimization of the industrial chain layout, and cultivate a group of applied large-model technologies to achieve breakthroughs A benchmark enterprise for growth.

It is understood that the industry has actively responded to and widely participated in the solicitation for the second batch of "Partnership Program". As of June 30, a total of 416 large-scale model R&D and application companies from inside and outside Beijing have applied to settle in. According to extensive demonstrations by experts from academia, industry and investment circles, and combined with market mechanism evaluation, a total of 63 companies have been successfully selected. The partners are composed as follows: 10 computing partners, 10 data partners, 10 model partners, 24 application partners, and 9 investment partners. In addition, 30 model observers were evaluated.

Currently, artificial intelligence technology represented by large models is developing rapidly, and the training of large models puts forward higher requirements for computing power. JD Cloud believes that artificial intelligence computing power services should have a division of labor, and different efficient tools should be used to do different things. More efficient heterogeneous infrastructure resources should be used as much as possible to support different computing power needs.

JD Technology has begun to establish the world's first supercomputing center in Chongqing early in 2021, mainly conducting scientific research and exploration in fields such as artificial intelligence and quantum computing. Faced with the growing demand for computing power, JD Technology's computing and storage networks fully support high-performance RDMA, which can provide high-performance, high-bandwidth and low-latency computing clusters for large model training, autonomous driving, scientific computing, etc.

In addition, JD Technology’s GPU-backed cloud hosts and bare metal servers are equipped with 25G dual network cards and self-developed Jingang smart network cards, which reduce virtualization losses to zero, greatly improve computing, storage, and network performance, and help enterprises in In a hybrid multi-cloud environment, we can quickly build stable and secure high-performance computing capabilities and comprehensively improve the efficiency of large model training.

Flexible scheduling of computing power clusters is the underlying foundation for large model training. Jingdong Technology's independently developed hybrid multi-cloud operating system Yunjian, on the basis of its original support for hybrid multi-cloud CPU computing power pooling capabilities, has further increased the scheduling required for AI applications with the pan-computing power pooling capabilities required for large model training. Management capabilities, including card management, node management, heterogeneous resource scheduling management, etc., provide one-stop computing power pooling solutions for a variety of AI applications including large model training, thereby comprehensively improving resource utilization.

As the second batch of "computing power partners" of the partnership program, JD Technology will use the world's leading computing power cluster to assist large model R&D iterations and demonstration applications, and help Beijing accelerate the creation of a source of artificial intelligence technology innovation and a highland for industrial development. A general artificial intelligence industry development pattern will be formed as soon as possible with complete elements, leading technology, and complete ecology, which can effectively support the high-quality development of the digital economy.

The above is the detailed content of The list of Beijing’s General Artificial Intelligence Industry Innovation Partner Program was announced, and JD Technology was selected as a “Computing Power Partner”. For more information, please follow other related articles on the PHP Chinese website!