Data is blowing out, the value of data is blowing out, and the computing power of the driving engine behind the data is facing huge demand.
From 5G to artificial intelligence to the metaverse, the innovation of technological concepts is changing rapidly, and to truly land, it is inseparable from more big data support and the continuous improvement of computing power.
Wang Jinqiao, a researcher at the Institute of Automation of the Chinese Academy of Sciences, a professor at the University of the Chinese Academy of Sciences, a doctoral supervisor, and the secretary general of the Multimodal Artificial Intelligence Industry Alliance, believes that artificial intelligence has now reached a new stage. Typically, as more and more data from artificial intelligence is generated, the amount of parameters in the model becomes larger and larger.
Wang Jinqiao mentioned the GPT-3 model disclosed by OpenAI in June 2020. He believes that this model has the ability of a single model to multitask, which opens a new beginning for artificial intelligence from dedicated intelligence to general intelligence. Of course, this new model will bring greater challenges, because to train such a model, it depends on a lot of hashrate support.
The new situation of the development of artificial intelligence "big model + big data + big computing power" is putting forward new demands for China's artificial intelligence industry. In China, after nearly two years of rapid development, the artificial intelligence computing center has been included in the key layout and planning of major cities across the country.
In the next step, the artificial intelligence computing center is no longer just an independent system, but should move towards interconnection, better convergence and sharing of computing power, data, and algorithm resources, and better meet the new situation and new needs of China's high-quality economic and social development.
Large computing power accelerates scientific research and innovation
Two major alliances were formed
The full play of data value is inseparable from the demand for computing power, and now massive data is becoming one of the new production factors, and digital transformation is gaining momentum, which has also spawned a huge demand for computing power.
Especially in the artificial intelligence industry, under the continuous development of technology, the exploration of large models and deep learning has gradually deepened, with Pengcheng. The ultra-large-scale pre-training model represented by Pangu has begun to empower all walks of life, making the industry's demand for computing power continue to rise; at the same time, under the national "double carbon" goal this year, the artificial intelligence computing center also needs to coordinate the construction and improve the utilization rate to further reduce emissions and consumption.
In this context, the network development of artificial intelligence computing centers has become a new development trend, and the flow and sharing of artificial intelligence elements such as computing power, large models, data sets, and industry applications have become an important way to cope with new demands. That is to say, in the future, after the gradual development and improvement of various artificial intelligence computing centers, they will also gradually connect with each other from independent systems to form a powerful computing network.
On May 31, the wuhan artificial intelligence computing center with the nature of public service in the country was put into operation, creating the "Wuhan model" of the construction of the artificial intelligence computing center of "one center and four platforms"; it also allowed the industry to witness the "Wuhan speed" of 5 months of rapid launch and saturated operation.
At the 2021 East Lake International Artificial Intelligence Summit Forum held on December 18, relying on the achievements of Wuhan Artificial Intelligence Computing Center in the past six months, Wuhan Ascend Artificial Intelligence Ecological Innovation Center, Intelligent Remote Sensing Open Source Ecological Alliance, Multimodal Artificial Intelligence Industry Alliance, Wuhan Artificial Intelligence Research Institute and other organizations were established.
The Intelligent Remote Sensing Open Source Ecosystem Alliance was established
At the same time, the Hubei Artificial Intelligence Talent Development Acceleration Plan was released, relying on the Artificial Intelligence Computing Center will further promote scientific research and innovation, accelerate the incubation of industry application innovation through scientific research and innovation, so as to bring together more enterprises, institutions and organizations to form industrial aggregation and development, while attracting and cultivating high-quality AI talents and jointly promoting the high-quality development of the artificial intelligence industry.
Wuhan Answer from the Artificial Intelligence Computing Center
Since its establishment for half a year, the Wuhan Artificial Intelligence Computing Center has handed over an excellent report card.
Wang Tao, Managing Director of Huawei and Director of the ICT Infrastructure Business Management Committee, introduced at the 2021 East Lake International Artificial Intelligence Summit Forum that under the guidance of the Ministry of Science and Technology, Wuhan took the lead in realizing the construction mode of artificial intelligence computing power of "one center and four platforms", and achieved fruitful results, "one center and four platforms" from Wuhan to the whole country, becoming a Chinese model for the development of the artificial intelligence industry, and accelerating the intelligent upgrading of related industries.
In addition, after the Wuhan Artificial Intelligence Computing Center was planned, a series of achievements were made in scientific research and innovation, application incubation, industrial aggregation and talent training, etc. Relying on the abundant computing power of the Wuhan Artificial Intelligence Computing Center, the Institute of Automation of the Chinese Academy of Sciences trained the world's first three-mode large model Zidong. Too early.
At present, the first phase of the Wuhan Artificial Intelligence Computing Center is built with 100 PFLOPS (in terms of FP16 accuracy, 1 PFLOPS is 100 billion times per second) AI computing power, and is about to complete the expansion of 100 PFLOPS to meet the local AI computing power demand.
Driven by the artificial intelligence computing center, Hubei's artificial intelligence industry is booming, and the industry intelligence has entered the fast lane in an all-round way. At present, the Institute of Automation of the Chinese Academy of Sciences, Wuhan University and other colleges and universities and 100+ enterprises and Wuhan Artificial Intelligence Computing Center have been launched project cooperation, which is the first artificial intelligence computing infrastructure with public service nature in China. In the past six months, the Wuhan model, Wuhan speed and Wuhan architecture have also undergone new developments, and are creating a Chinese model for the development of the artificial intelligence industry.
Recently, the Ministry of Science and Technology organized the "National New Generation Artificial Intelligence Innovation and Development Experimental Work Promotion Symposium" in Wuhan, and the relevant person in charge of the Ministry of Science and Technology said that the Wuhan Artificial Intelligence Computing Center will be used as a pilot to improve the construction of a public computing open innovation platform.
At present, major cities are focusing on the layout and planning of computing infrastructure such as artificial intelligence computing centers and integrated big data centers. Wang Tao believes that in the future, the computing power infrastructure in various places will be connected into a network to form a nationally integrated diversified computing power network to achieve the sharing of computing power, data and ecology, which will also become the key direction of digital infrastructure.
Build a network of diverse computing power
Innovate a new model for the development of artificial intelligence industry
Wuhan, which is connected by nine provinces, is also expected to become a core node in the construction of an interconnected computing power network in the future.
In China, after nearly two years of rapid development, artificial intelligence computing centers have been included in the key layouts and plans of major cities across the country, and cities such as Shenzhen, Wuhan, Xi'an, and Chengdu have built artificial intelligence computing centers and put them into operation.
At present, the construction of domestic artificial intelligence computing centers is still led by the government, and local governments are based on the innovative development of independent technical routes in the process of overall planning and investment to promote the high-quality construction of artificial intelligence computing centers, so as to accelerate the deep integration of artificial intelligence and the real economy.
At the same time, during the same period of the start of the construction of artificial intelligence computing centers in various places, the government began to fully consider how to operate effectively, set up an artificial intelligence ecological innovation center through a designated operating company, carry out the computing power and ecological operation of the artificial intelligence computing center, provide public computing power services, application innovation incubation services, industrial aggregation and development services, scientific research innovation and talent training services, so as to fully release the value of computing power and realize the small cycle of business and the big cycle of industry.
However, on the one hand, with the wide application of new generation information technologies such as 5G, artificial intelligence, cloud computing, and big data in all walks of life, the diversity of industry applications has also brought about the diversity of data and computing power. On the other hand, the pain points of the development of China's artificial intelligence industry are putting forward new demands for artificial intelligence computing centers.
At the 2021 East Lake International Artificial Intelligence Summit Forum, Xu Yingtong, president of Ascend Computing Business, mentioned that weak foundation, expensive computing power, data closure and high professionalism are the problems faced by the artificial intelligence industry, "The largest scale of our single machine damage center is 100 million, the entire scale is still limited, give play to our institutional advantages, (we must) realize the network networking of multiple cities." At the same time, he said that the core is to open up the entire path from the computing center of artificial intelligence to the computing center of the industry and the application of the industry.
In the future, the artificial intelligence computing center is not satisfied with only being an independent system, but to gradually move towards an interconnected computing power network, connecting the artificial intelligence computing center nodes distributed in various places to form a network of perception, distribution and scheduling of artificial intelligence computing power, better gathering and sharing computing power, data, and algorithm resources, and better meeting the new situation and new needs of China's high-quality economic and social development.
Xu Yingtong expects, "In time, we can imagine that the large models and applications developed by the Wuhan Artificial Intelligence Computing Center can be quickly applied to the urban operation and management of the Greater Bay Area through the computing power network, so as to achieve regional synergy and industrial complementarity." ”