WAIC2022 ｜ Another domestic big model was born, but the computing power is a long way to go

Jiangyue Shanghai, a reporter from Southern Finance and Economics, reported that databases, processors, operating systems, AI frameworks ... Building an AI computing power base requires a multi -directional and long -term construction efforts. At the World Artificial Intelligence Conference on September 2nd, Huawei, which is building a "Shengteng" AI computing power base, announced new achievements in the construction of a scientific field and a national computing power.

On the same day, the release of the first industrial -grade fluid simulation model "Oriental. Yufeng" will help China Shangfei to build a large domestic aircraft faster. The re -expansion of China Computing Power Network will promote "allowing users to use computing services like electricity."

Huo Houkun, chairman of Huawei, said that "Shengteng" AI will continue to work in service scientific computing, so that scientists in biology, physics, meteorological and other fields use AI technology to solve scientific issues regardless of traditional or cutting -edge. Zhang Diyi, the president of Huawei Shengteng Computing Business, introduced to a reporter from Southern Finance and Economics that the computing power is currently scarce and expensive. Establishing a national large -scale computing power center will help the computing power more fairly flowing to all areas including scientific research.

Daping Model Published by fluid simulation

In recent years, "Foundation Model" has shown unprecedented power in early pre -training and has become a consensus of AI innovation. Earlier, the "Shengteng" AI, which had hatched the three -model model of "Zidong · Taichu", also united China COMAC on September 2nd, released the industry's first industrial -grade fluid simulation big model "Oriental. Essence Zhang Diyi said that the "intelligent fluid mechanics industry consortium" established on the day will open up the results of the results of scientific research innovation and industrial landing.

Hu Houkun pointed out that in promoting a more efficient AI development model, big models have become consensus. However, the development process of large models is very complicated and the threshold is also very high. Therefore, "Shengteng" provides a large model development platform for the industry, making AI development more convenient and rapid.

In the future, it will continue to work in the field of scientific computing, solving scientists in the fields of biology, physics, and meteorology to solve scientific issues of traditional or cutting -edge, and will provide AI technical support and accelerate the transformation of scientific research results.

It is understood that "Oriental. Yufeng" is a model for high -efficiency, high -precision AI simulation prediction models for large passenger -oriented wing -type streaming fields. A precise prediction can be performed on the strenuous regions in the flow field. The average error of the flow field is reduced to one thousandth of the order of 1,000 to meet the industrial -grade standards.

Wu Guanghui, an academician of the Chinese Academy of Engineering and chief scientist of COMAC, said that artificial intelligence is changing the future of aircraft design and aviation operations, changing the ecology and format of the aviation industry. "Artificial intelligence technology will bring new kinetic energy to large aircraft manufacturing, and the large Chinese aircraft that carry several generations of people's dreams are becoming reality."

China Computing Power Network expands again

Let users use computing power services like electricity, this is the goal of the current efforts of the power world. In June of this year, China Computing Power Network — the first phase of the Intelligence Network was officially launched, which means that under the layout of the national “East Digital and West” project and the national integrated computing power network, the China Computing Power Network Plan has fully launched Essence

On September 2, China Computing Power Network added 7 new nodes to further expand. The 7 new nodes include Shenyang, Guangzhou, Chongqing, Kunming, Fuzhou, Changsha, Hebei (Langfang), etc., which has made the total number of Chinese computing power networks more than 20 so far.

China Computing Power Network will access more national intellectual calculation centers, super computing centers, and "East Digital West Calcus" hub nodes to jointly build independent innovative computing power network technology systems to achieve the collaborative scheduling and efficient national computing power of the country Calculate to create the strongest power base for the digital economy, and realize "allowing users to use computing power services like electricity".

Zhang Diyi introduced that in the process of promoting the "East Digital and West Calculation" and integrated big data centers, the government in order to provide inclusive computing power, and the construction of an artificial intelligence computing power center in intensive land, so that the computing power can also be like railways and highways. The electricity is the same.

However, in his opinion, the current computing power is still scarce and expensive. Many computing power is in the enterprise, mainly based on the self -construction and self -use of the enterprise. Some artificial intelligence junior enterprises and scientific research institutions are obviously insufficient, which will slow down the release of scientific research innovation and scientific research results. Zhang Diyi pointed out that the construction of the computing power center can improve the "Matthew effect" of the current computing power distribution and make computing power more flow to scientific research and other fields.

- END -