Letter from an expert | Liu Jun from Inspur: AI computing will become one of the core supports of the "new infrastructure"

Liu Jun | General Manager of Inspur AI & HPC Division
Quantum Bit | Official Account: QbitAI

Editor’s Note:

In the "new infrastructure", the infrastructure construction of artificial intelligence is an important part, and the development of artificial intelligence is inseparable from the support of computing power. This article comes from an industry insider who has been engaged in computing infrastructure construction for a long time.

Liu Jun, general manager of Inspur Group's artificial intelligence and high-performance business unit, believes that AI computing will become the core support for related infrastructure.

As the amount of data increases and computing models become more complex, the demand for computing power continues to increase. Traditional computing platforms need to be transformed and upgraded.

From the perspective of AI applications, as AI technology penetrates into various industries, industrial AI will bring a larger market, but it also faces three major challenges: infrastructure investment, talent gap, and underlying research and development.

Ultimately, China's artificial intelligence needs to achieve a continuous self-cycle in its development.

about the author:

Liu Jun, General Manager of Inspur Group's Artificial Intelligence & High Performance Computing (AI&HPC) Department, has 19 years of professional experience in technology and applications in the HPC field. He participated in the initiation and organization of the Academic Supercomputing Competition (ASC) and won the second prize of the National Science and Technology Progress Award and the first prize of Beijing Science and Technology Award.

Key Points

• New infrastructure not only needs to "maintain growth", but also needs to promote "structural adjustment", and is an important support for industrial structure adjustment.

• The rapidly growing amount of data and more complex models bring greater challenges to computing.
  

• The Internet accounts for 62.4% of the AI ​​computing power investment market, followed by the government industry, the financial industry and the manufacturing industry.

• Industrialized AI will bring about a market worth trillions of dollars.

• Industrialized AI faces three major challenges: insufficient investment in infrastructure, talent gap limiting development potential, and weak basic research.

full text:

(Subheadings added later)

Recently, the central government has proposed to accelerate the construction of new infrastructure such as 5G network base stations, artificial intelligence, big data centers, and industrial Internet. The concept of "new infrastructure" has become a hot topic of widespread concern and discussion among all sectors of society.

There are many opinions about new infrastructure, but a widely recognized consensus is that "new infrastructure" is not only a means to promote China's economic recovery, but also an important support for China's industrial structure adjustment. Simply put, the mission of "new infrastructure" is not only to "maintain growth" but also to promote "structural adjustment" .

So, what are the elements of the new AI infrastructure? How to build a new AI computing infrastructure? And in what ways will AI promote the transformation and upgrading of my country's industrial structure?

AI computing infrastructure: two parts: soft and hard

There is a big difference between "new infrastructure" and traditional infrastructure. One of the core differences is to promote the construction of digital economic infrastructure to support the development of traditional industries towards networking, digitization and intelligence. This is actually a process of continuous expansion from the physical world to the information world, and from the information world to the intelligent world.

AI computing infrastructure consists of two parts: “hard” and “soft”.

On the one hand, computing, storage, and network hardware infrastructure together form a hardware infrastructure system that supports the widespread application of the new generation of artificial intelligence; on the other hand, diverse machine learning frameworks, algorithms, and related tool software, PaaS platforms, services, etc. together constitute a soft infrastructure system that supports artificial intelligence application development and continuous innovation.

The growing demand for AI computing power places greater demands on infrastructure

AI computing infrastructure will become the top priority of the entire new round of AI infrastructure construction. Currently, the rapidly growing massive data and more complex models are bringing greater challenges to computing.

If computing power cannot grow rapidly, we will have to face a terrible situation: when huge amounts of data are used for artificial intelligence training and learning, the amount of data will exceed the carrying capacity of memory and processors, the entire training process will become extremely long, and even the most basic artificial intelligence will not be achieved at all.

Data shows that the amount of global data is currently growing at an average annual rate of 50% .

IDC predicts that the total amount of global data will reach 44ZB in 2020, and China's total data will exceed 8ZB, accounting for 18% of the world's total data. The total amount of global data is expected to reach 175ZB by 2025. At the same time, deep neural networks are also developing rapidly, and deeper and larger algorithm models and more complex architectures are becoming a trend.

Inspur is ready for the increasingly severe computing challenges facing artificial intelligence.

We will achieve a comprehensive upgrade of artificial intelligence servers this year, providing customers with a richer range of products with stronger computing performance and higher interconnection bandwidth, supporting users in conducting AI training and deployment of larger data scales and more complex models.

At the same time, we will also carry out more extensive cooperation with partners to jointly create an agile, easy-to-use and rich software environment for the development of artificial intelligence algorithms, models and applications.

AI computing infrastructure supports the development of AI in the industry

The value of AI computing infrastructure lies in promoting transformative progress in production efficiency , which is one of the core values ​​of artificial intelligence technology.

Therefore, artificial intelligence should aim at industrial application, and through deep integration with industries such as manufacturing, finance, transportation, and medical care, it should bring about quality changes, efficiency changes, and power changes in various industries through intelligent upgrades. AI computing infrastructure will be the core supporting force to promote this transformation.

At present, the scope of artificial intelligence application is already very wide in the industry and maintains rapid growth every year.

According to the 2019-2020 China Artificial Intelligence Computing Power Development Assessment Report jointly released by IDC and Inspur, the Internet occupies 62.4% of China's artificial intelligence computing power investment market share, maintaining the first position. Typical application scenarios include: e-commerce precision marketing, image recognition and intelligent customer service, video content review, face recognition and intelligent writing, etc.

Government industry Following closely, typical application scenarios are mainly concentrated in urban operation and management platforms such as safe cities, smart cities, and smart transportation.

The third one is the financial industry. Typical application scenarios mainly include identity authentication in the financial industry, face recognition in the payment process, fraud analysis and investigation, etc.

Fourth, the quality management (QC automation) and smart factories of the manufacturing industry are developing rapidly.

Telecommunications industry Entered the top five for the first time, with typical application scenarios including relatively mature intelligent customer service and precision marketing.

Industrial AI brings a bigger market

In the specific practice of the industry in the past few years, we have a very profound understanding that the industrialization of AI will bring a market opportunity of hundreds of billions of dollars, and the larger industrial AI market will generate a large market of up to trillions of dollars.

However, in order to promote the rapid implementation of artificial intelligence, we will encounter the pressure of diversified and contradictory demands. The development of industrial AI faces both technical and commercial challenges.

So, how to drive the upgrade of industry AI amidst the dual challenges? To this end, Inspur proposed the "Metabrain Ecosystem", hoping to achieve the integration of the entire AI industry through a new model of ecological co-construction.

The uniqueness of the metabrain ecosystem lies in the fact that it does not belong to a single enterprise, but is composed of three core elements.

The first is the left-hand partner, which is a company with core capabilities in AI development;

The second is the right-hand partner, which is a system integrator and software developer that has been engaged in the informatization, digitization and intelligentization of each industry field for a long time and has the ability to implement the overall interactive artificial intelligence solutions of the end customer industry;

Third, in the Metabrain Ecosystem Plan, Inspur will share three core platforms, including an efficient and innovative AI computing platform, an agile and collaborative AI resource platform, and an instant delivery AI algorithm tool platform.

The three parties will work together to form an open and integrated ecosystem, efficiently aggregate industrial forces, and promote the integration and implementation of industrial AI.

At present, the "Metabrain Ecosystem" has gathered more than 100 left-hand partners and more than 1,000 right-hand partners, and jointly created dozens of artificial intelligence solutions based on different scenarios covering smart finance, smart education, smart industry, smart medical care, etc.

Industrial AI still faces three major challenges

The development of the artificial intelligence industry is at a critical stage. On the one hand, we see the huge potential energy contained in it and it is being sought after by all parties. On the other hand, it is also facing three major challenges: insufficient infrastructure investment, talent gap limiting development potential, and weak basic research.

The first is insufficient investment in AI computing infrastructure , and new infrastructure will promote the solution to this problem.

At present, China has fallen behind in the construction of AI computing infrastructure. The United States and Japan are already building government AI computing platforms, such as Summit in the United States and ABCI in Japan, while in my country, apart from BAT, which are building large-scale AI computing systems, most of them are still based on traditional scientific computing.

Traditional scientific computing platforms are not suitable for the needs of artificial intelligence applications. Computing platforms that rely solely on the construction of some enterprises have limited applicability. Only by viewing computing platforms as infrastructure investments for transformation and upgrading can we meet the major opportunities for the development of AI in our country's industries.

Inspur is actively collaborating with governments and industrial organizations at all levels to promote the construction of the government's artificial intelligence computing basic platform.

The second is that the talent gap limits development potential .

Relevant reports show that China's AI talent gap has exceeded 5 million. Although my country has achieved certain results in AI talent training in recent years, there is still a large gap with developed countries in the construction of AI-related disciplines and talent training in colleges and universities, which is mainly reflected in the lack of high-level leading talents, innovation teams and interdisciplinary innovation platforms, few breakthroughs in basic theories and original algorithms, and weak guidance in the training of compound talents.

To address this issue, the government has included artificial intelligence in the scope of support for the "Special Enrollment Plan for the Training of High-level Talents Urgently Needed in Key National Fields", encouraging and supporting universities to develop artificial intelligence disciplines. Currently, 35 universities will open artificial intelligence majors to cultivate talents that are crucial to technologies such as autonomous driving.

In addition, Inspur is also exploring the cultivation of compound talents in artificial intelligence and supercomputing through the ASC World University Supercomputing Competition with hundreds of universities around the world.

The third is that in addition to the application field, we should also pay attention to basic technology and underlying research and development , which will affect the future. On the one hand, this is closely related to talent training, and on the other hand, it also requires the government to focus on and support resource allocation, and more importantly, the extensive participation of social capital.

With the core support of AI computing infrastructure, through the innovative power brought by AI talents, clear policy guidance and support, and a large-scale application market, we hope and firmly believe that we can see China's AI development achieve a continuous self-cycle, which will drive our country to become an AI superpower.

The author is a contracted author of NetEase News and NetEase "Each has its own attitude"

-over- ​ ​

How to pay attention to, learn and make good use of artificial intelligence?

Every weekday, QuantumBit AI Insider selects the latest global technology and research developments, summarizes new technologies, products and new applications, sorts out the hottest industry trends and policies of the day, and searches for valuable papers, tutorials, research, etc.

At the same time, the AI ​​Insider Group provides a platform for communication and sharing, which can better meet everyone's needs for obtaining AI information and learning AI technology. Scan the QR code to subscribe: