From Ideals To Reality, Why Does China Quickly Implement Artificial Intelligence Computing Power Network?

At the end of May 2021, the National Development and Reform Commission, the Central Cyberspace Administration of China, the Ministry of Industry and Information Technology, and the National Energy Administration jointly issued the "National Integrated Big Data Center Collaborative Innovation System Computing Power Hub Implementation Plan". In this plan, about The coordinated and coordinated development of computing resources has set the tone at the macro level, and a new vision for the future network has emerged;

Just four months later, at the just-held Huawei Full Connect 2021, the China Institute of Scientific and Technological Information, AITISA (New Generation Artificial Intelligence Industry Technology Innovation Strategic Alliance), and Pengcheng Laboratory jointly released the "Artificial Intelligence Computing Center Development" White Paper 2.0 - From Artificial Intelligence Computing Center to Artificial Intelligence Computing Power Network", this new vision of the future network has completed the macro concept and roadmap design, and has begun to take steps to put it into practice.

QQ screenshot 20210928105642

The artificial intelligence computing power network is becoming one of the biggest waves in the "new infrastructure" wave, driving the development of China's AI industry into an epic acceleration process, and bringing the ultimate goal of artificial intelligence development, the intelligent society, closer. In this process, China's unique advantages in developing the AI industry have also been fully demonstrated.

The answer to computing power is "increasing the dimension" step by step. The emergence of artificial intelligence computing power network is inevitable.

To understand what the artificial intelligence computing power network is and why it appears, we must start from the process of the industry's continuous satisfaction of computing power needs.

With the accelerated implementation of AI scenario-based applications, among the AI troika, innovative algorithms are constantly emerging and data is becoming more and more abundant, but the challenge of computing power is becoming increasingly severe - the supply of computing power in the upstream of the AI "industrial chain" has begun to appear Insufficiency forces the evolution of various forms of computing power and "upgrading the dimension" step by step.

Stage 1: Basic software and hardware innovation

At the beginning, there was a huge development in basic software and hardware specifically for AI computing. This included the launch of specialized chip products (such as NPU), or further, heterogeneous computing based on the innovation of underlying computing architecture (such as Ascend AI). Local or cloud computing provides AI with more abundant computing power than in the past.

Phase 2: Construction of artificial intelligence computing center

Then, with the further development of the AI industry and the stronger demand for computing power, even with innovations in basic software and hardware, a single enterprise will face the problem of shortage of computing resources or high prices when deploying or purchasing computing power. .

As a result, in places with AI industry clusters, artificial intelligence computing centers built under the leadership of the government began to appear, providing enterprises with relatively more reasonable computing power prices and more flexible supply methods through intensive methods. The most typical ones have been implemented Shenzhen Pengcheng Cloud Brain II and Wuhan Artificial Intelligence Computing Center, with the blessing of Ascend AI, their computing resources are operating at full capacity, which shows the strong demand.

At the same time, there is also the industrial layout of artificial intelligence computing centers.

Unlike supercomputing centers that are also intensively deployed and only serve scientific research, artificial intelligence computing centers have an industrial mission. To help the development of smart cities, smart finance, smart manufacturing, smart transportation and other industries, pure computing power also needs to be combined with Industrial development is coordinated. Therefore, the construction of artificial intelligence computing centers by local governments has formed a pattern of one center and four platforms: relying on the artificial intelligence computing center, a public computing power service platform (inclusive computing power) and an industry application innovation incubation platform ( Create an AI application demonstration benchmark), an industry aggregation development platform (promote the intensive development of the AI industry), and a scientific research innovation and talent training platform.

Stage 3: Artificial Intelligence Computing Network

After the continuous development of computing infrastructure, including artificial intelligence computing centers, new problems have emerged:

AI computing power in different regions also has its own peaks and troughs, and large AI models developed on the basis of more abundant computing power use computing power in stages and with high density. Naturally, people in the industry began to think about allocating computing power across regions. To solve the problem of human resources, interconnect the computing power network between artificial intelligence computing centers in various places to realize resource sharing and collaborative scheduling.

As a result, the artificial intelligence computing power network emerged as the times require. It uses new network technology to connect artificial intelligence computing center nodes distributed in various places, and senses, allocates, and schedules artificial intelligence computing power within the region. According to the computing power resources of each center and the Computing power is dynamically allocated based on the demand in each region.

Taking into account the actual development of artificial intelligence computing centers and China's regional economic conditions, the future artificial intelligence computing power network may refer to the "three plus one" policy proposed by relevant experts in terms of specific interconnection methods: relying on the Yangtze River Delta and the Guangdong-Hong Kong-Macao Greater Bay Area. The interconnection construction in the District, Beijing-Tianjin-Hebei, Chengdu-Chongqing Economic Circle and other regions is proceeding step by step.

It is worth mentioning that in addition to computing power, data and algorithm resources are also gathered and integrated nationwide through the artificial intelligence computing power network. The value concept of "one network, three convergences" was also proposed in the white paper. Simply put, it is:

One network is the artificial intelligence computing power network;

Computing power aggregation, that is, high-speed interconnection of nodes, unified management and operation and maintenance;

Data aggregation means realizing the safe opening of public data resources at different nodes and building high-quality artificial intelligence public data sets;

Ecological convergence, that is, the opening of large model capabilities and the sharing of application innovation results at different nodes, is intended to strengthen cross-regional scientific research and industrial collaboration.

It can be seen that, step by step, from basic software and hardware innovation to artificial intelligence computing power network, the supply of computing power resources is becoming more and more abundant, and the collaboration of computing power, algorithms, and data is becoming more and more closely integrated with the industry. Of course, these three stages are not in strict chronological order and are currently being carried out simultaneously.

Why can only China build a good artificial intelligence computing power network?

The emergence of artificial intelligence computing power networks is an inevitable necessity for the development of industrial demand. However, this necessity can only happen in China. It is determined by China's unique AI technology and industrial development reality, and it is also China's unique advantage.

The reasons include at least four aspects:

1. The general background of the simultaneous development of new infrastructure

On the one hand, the artificial intelligence computing power network itself exists as a new type of infrastructure; on the other hand, it also needs to be supported by lower-level new infrastructure construction.

For example, the coordination of the computing power of different nodes requires a communication network that can support the transmission and access of massive data. According to the plan, in the next ten years during its implementation, broadband will achieve a leap from Gigabit to 100G, and IP Such resources must support a hundredfold increase in capacity. In addition, powerful edge computing software and hardware infrastructure is also needed to realize the center + edge distributed computing power model to achieve a hundredfold capacity increase.

From the artificial intelligence computing power network to industrial applications, this process also requires a lot of new infrastructure. From the production of computing power to the final use and good use of computing power, it cannot be implemented without a good new infrastructure foundation.

China's new infrastructure is already in full swing from macro policies to specific implementation. Artificial intelligence infrastructure has been clearly identified as the core task of the new infrastructure. At the policy level, it is also the infrastructure that supports technological self-reliance and the development of the digital economy. 5G, wifi6, Technologies such as IPv6 are becoming widely available, which is unique in the world.

2. The rapid development of the "node" itself

From the formation of the artificial intelligence computing power network, we can find that this is a macro-level overall planning. Its implementation process is not a construction process from nothing to a comprehensive one. The "node" of the artificial intelligence computing center becomes the core resource. If there is not a large number of "Nodes" may not be able to promote the rapid construction and implementation of "nodes", and the artificial intelligence computing power network can only be a castle in the air.

This is why there are a lot of discussions about artificial intelligence computing power networks, but few can be put into action.

In China, the Artificial Intelligence Computing Center has long been the implementation entity of the State Council's "New Generation Artificial Intelligence Development Plan" to "build high-efficiency computing infrastructure". So far, with the construction led by local governments and the participation of technology companies such as Huawei, artificial intelligence computing centers have been built and put into operation in cities such as Shenzhen, Wuhan, and Xi'an. Artificial intelligence computing centers such as Chengdu and Zhongyuan are under construction. In addition, there are The construction of artificial intelligence computing centers in Beijing, Nanjing and other places is also being planned.

WeChat screenshot_20210928105651

These dense nodes, which only appear in China, promote the rapid implementation of artificial intelligence computing power networks.

3. The convergence of demands for industrial cluster development

Without the support of huge demand, or the existing demands in the market cannot be connected to effectively "supply" computing power, the artificial intelligence computing power network cannot be implemented. In other words, the implementation process itself is a waste of resources and energy.

At this time, the value of China's unique industrial cluster development path in the process of social and economic development is revealed.

On the one hand, since the development of the AI industry in various regions is carried out in a clustered manner, the demand is huge and concentrated, and the effective demand in the region can be brought together to achieve a full connection between supply and demand;

On the other hand, the huge demand from industrial clusters means that each region generally has its own peaks and troughs of computing power. Under the artificial intelligence computing power network, effective computing power supply demand can always find the right users locally or across regions.

This allows the artificial intelligence computing power network to "make the best use of everything", maximize value and stay away from "the art of slaying the dragon", forming a healthy development.

Ultimately, the high-efficiency operation of the artificial intelligence computing power network will promote the rapid development of various industrial clusters across the country, and a positive feedback loop will be formed.

4. Technology companies with internal drive to break through are working hard

Looking further into the implementation details, the artificial intelligence computing power network requires a lot of technological innovation to support it. For example, only in terms of the integration standard of the computing power network, it is necessary to complete the computing power network architecture and interface, application and computing power perception research, and computing power perception research. Strength demand quantification and modeling research, computing network resource credibility and collaboration and other standards construction.

If these solid technologies do not break through, they will become the shortcomings of the barrel, seriously affecting the computing power and industrial promotion effect of the artificial intelligence computing power network.

However, since the artificial intelligence computing power network itself is a brand-new thing, many matching technologies are also brand-new, and it is difficult to find reference objects on a global scale. At this time, Chinese technology companies that have the internal drive for independent innovation under a particularly complex situation will be more willing to invest in these technologies. This is an important opportunity to catch up with the world and establish a technological voice.

It's not that foreign technology companies can't produce these technologies. It's just that, represented by Huawei, the solution provider of the artificial intelligence computing center, Chinese technology companies are more motivated on the one hand, and on the other hand, in the new era In the field, based on existing computing power and new infrastructure experience (such as Ascend full-stack AI), it will be better in terms of technical understanding.

From technological innovation and industrial advancement to technological highlands, the value of artificial intelligence computing power networks is highlighted

With a sword in hand, what role can the artificial intelligence computing power network play? There are too many answers to describe in detail, but at least there are three aspects that are of certain value.

First, it is strong support for breakthrough technological innovation.

In this regard, multimodal large models are typical. As an important technological innovation for the future of strong artificial intelligence and general artificial intelligence, multi-modal large models have been proposed in the field of artificial intelligence for many years, and there have been many technological breakthroughs in the industry. However, multi-modal large models will go further. Power demand is rising exponentially, and general computing infrastructure will soon be unable to cope.

The transition from separate training of images, text, and voice to dual-modal and tri-modal allows artificial intelligence to flexibly respond to the transformation of different modalities, as naturally as humans interacting with the world. In this regard, artificial intelligence calculates There will be obvious advantages in using the power network to drive it.

In the same way, there are many technological innovations that require explosive growth in computing power and will be rapidly advanced with the support of artificial intelligence computing power networks.

Then, it is to fully meet the needs of the entire chain of industrial development.

As we all know, a wide range of application markets is an important advantage for China in the development of all technologies. Putting technology on the market to generate value and feed back technology research and development, with a rolling increase, has become a standard model for technological progress in many industries.

However, in artificial intelligence, upstream supply and downstream demand have become equally important. The emergence of artificial intelligence computing power network means that the development of China's AI industry and technology still has a huge scene application space to realize commercial value. On the other hand, it has the world's leading computing power supply capabilities, and its relative advantages are more obvious, promoting the prosperity of the AI industry and accelerating it.

Finally, it is to create a true technological highland.

In the past, in the development of artificial intelligence, although the troika of computing power, algorithms and data have made rapid progress, in the final analysis, there have not been many breakthrough developments. Most of them have been continuously optimized and improved within the existing framework. , For example, if the hardware is not good, then improve the chip performance, and if it is not good enough, stack it up to win by quantity.

The emergence of artificial intelligence computing power networks may have similar breakthrough significance as the creation of basic models at the algorithm level (for example, the BERT model in the field of NLP), directly raising the dimension of AI development in a collaborative manner, and in artificial intelligence computing. The center then further broke through the old common form of computing power supply, allowing AI to truly get rid of the shackles of computing power and embrace new thinking and practice. This also shows that with the power of integration, China is occupying a new technological highland belonging to artificial intelligence.

All in all, in the past many years, China has continued to lead the world in the fields of Internet technology, artificial intelligence and other fields. Now, this lead is continuing and deepening. With the support of artificial intelligence computing power network, China is expected to be the first to enter the intelligent society and truly gain Look at the global technology and industrial chain voice.