The intellect is here!Looking for a new soil of the digital economy

The intellectual calculation industry requires more efficient and greener soil

World Internet giants have bet on artificial intelligence computing power and have invested huge investment. At the same time, China's intelligent computing is achieving breakout. On August 30, Alibaba Cloud fully opened the Flying Smart Calculation Platform and enabled two super -intellectual centers to provide one -stop services from intellectual calculations to AI development tools to support autonomous driving, new drug research and development, meteorological prediction, industrial energy sources Waiting for various intellectuals scenarios to greatly improve the efficiency of AI training.

The two super intellectual centers represent the peak practice of China's intelligent computing. In a smart calculation center in Zhangbei, the construction scale will reach 12 EFLOPS, which can be calculated in 120 billion times a second. The other is located in Ulancha, with a construction scale of 3 EFLOPS, which can be calculated by 30 billion times a second.

Cai Yinghua, CEO of Alibaba Cloud Smart Global Sales, said that the intelligence is not only large in scale, but also requires efficient, economical, green, open and industrial practice. Intelligent computing is a huge and complex system. There is no systematic core technical capabilities. Heaping hardware does not have computing power, and it cannot bring actual industrial value.

It is understood that Alibaba Cloud's Feitian Intelligence Platform uses advanced technical architecture, which can support a variety of chips, and increase the utilization of computing power resources by more than 3 times, increase the efficiency of AI training by 11 times, and increase reasoning efficiency by 6 times. At the same time, the application of green technologies such as cold and high -density servers such as cold and high -density servers can make PUE a minimum of 1.09.

Some experts said that intelligent computing is different from other types of computing power, and a large amount of data needs to be trained with massive data. A large amount of computing power is lost in data migration and synchronization. Many traditional intellectual centers resource utilization rate is even less than 10%. This has led to a high cost of intelligent computing power and restricting industrial development.

Sun Weijie, CEO of the momentum technology of investing in life science research, said that the new algorithm puts forward many new requirements for computing power, and the total amount of computing power is richer and more elastic. The service platform not only improves training efficiency and reduces costs, but also brings a unified R & D platform to downstream customers with potential technology.

Another expert believes that if we think of AI computing power as the soil of the intelligent industry, first of all, the nutrients of this soil must be particularly sufficient, and the cost of this nutrient cannot be high. Secondly Bee pollination and ants flippage soil can grow well on the new species above.

In this regard, Cai Yinghua said that the Feitian Intelligence Platform exported Alibaba's accumulated technology and practice for several years, and each level came from the polishing of large -scale practice, focusing on solving the core pain points. It not only provides computing power, but also provides the ability to integrate AI big data. Disposal to the end of the end to solve the pain points of the industry.

Not long ago, Xiaopeng Automobile constructed the "Fuyao" in Ulanchabu Based on Feitian Intelligent Coupons. The scale of computing power reached 600PFLOPS. It is the largest autonomous driving intellectual center in China. The daytime time was shortened to 1 hour, and the training efficiency was speeded up by nearly 170 times. Based on Feitian Intelligence, the efficiency of 128 calories and the efficiency of 128 cards exceeded 96%, reducing the training cost of autonomous driving model by 62%, and the training speed increased by 110%, which greatly shortened the model iteration cycle.

In the field of industrial manufacturing, SAIC Group uses high -performance computing to increase the efficiency of industrial simulation by 25%, accelerating the research and development of new models; Panang Steel Group uses industrial intelligent brain to calculate through real -time simulation of desulfurization technology, and use optimal parameters with optimal parameters Adjusting the material ratio, every ton of steel is produced to save one kilogram of iron, saving more than 17 million per year.

In the field of life science, after the use of the flying intelligence platform, the efficiency of the entire training and reasoning ranges by 2-6 times, and the training cost is reduced to 30 %. The School of Chemistry and Molecular Engineering of Peking University uses the Flying Smart Calcum Platform for targeted drugs to increase the efficiency of data integration by 100 times.

In the field of urban governance, the accuracy of water conservancy scheduling prediction is as high as 95%through remote sensing data and simulation deduction, which significantly improves the water supply efficiency of mountain cities. Through urban simulation, Kunming City provides real -time signal lights to provide real -time signal lights at the intersection, and achieves "one policy" traffic guidance, achieving an average daily speed increase of 7%.

Li Donghong, deputy dean of the Global Industry Research Institute, pointed out that every time the computing power index is increased by 1 point, it will bring a 3.5 ‰ contribution to the digital economy and promote GDP to increase by 1.8 ‰.

In the fierce competition of the global digital economy, China is standing out in multiple fields of computing power such as cloud computing, big data, and artificial intelligence, and provides sufficient and surging computing power for the digital transformation of the industry.

- END -