China businessman, strategy, Chen.
In the era of artificial intelligence, AI computing power is electricity, and AI computing center is power plant.
In the era of electricity, we have built a "power grid". Now, with the artificial intelligence computing centers all over China landing one after another, we are weaving an AI computing network.
At present, the development of artificial intelligence has been promoted to the national strategic level. Accelerating the development of artificial intelligence industry and ensuring and providing abundant AI computing power have become urgent problems for countries, cities and enterprises that have won in the AI era.
No computing power is like no electricity.
What bothers the heads of many local governments and enterprises in China recently?
The answer may be two words: lack of electricity.
However, this troublesome thing has not been solved. In different countries around the world, even in different areas within a country, they are beginning to face the key factors that determine the national economy and people's livelihood like electricity.
This key element is AI computing power.
AI computing power, as its name implies, is the computing power that supports AI. The calculation here is not addition, subtraction, multiplication and division, but the calculation of everything, which is highly complex and ubiquitous under the internet of everything and artificial intelligence.
Different from traditional computing power, AI computing power needs extremely high parallel processing power in order to support the development, training and reasoning of AI models, so it needs special AI chips and frameworks.
For example, AI chips with powerful floating-point computing capabilities can provide high-quality AI models that meet the intelligent transformation of industrial enterprises through training and continuous iterative optimization. In complex model training, hundreds of billions of floating-point parameters need to be fine-tuned by hundreds of thousands of steps, which requires fine floating-point expression ability. Without a powerful training chip, it is difficult to ensure the efficiency of the algorithm model output. The release of the 100 billion Chinese NLP (Natural Language Processing) model Pengcheng Pangu and the biomedical platform Pengcheng Shennong are inseparable from the support of AI chips.
For another example, 90% of the AI framework that is regarded as the "AI domain operating system" is developed based on the AI framework layer. In this field, domestic technology companies have made great achievements: Ascension MindSpore, the industry-leading AI computing framework, is a deep learning and training reasoning framework, which supports all scenarios of end, edge and cloud. In addition to the advantages of automatic control, a set of frameworks can support diverse applications such as AI+ scientific computing. At present, the cumulative download volume of MindSpore community in Shengsi exceeds 600,000, and more than 100 colleges and universities choose MindSpore in Shengsi for teaching.
It is with the AI computing power released by these AI chips and AI frameworks that we can accelerate into the era of Internet of Everything and artificial intelligence.
Today, from everyone's mobile phone, to the cloud platform of the enterprise, to the brain of the city ... our production and life are increasingly dependent on AI, and we are getting more and more strength from AI.
For China, AI is the key to the transformation and upgrading from a manufacturing power to a manufacturing power. In recent years, many cities are striving for various resources to enhance their development and competitiveness, and AI computing power is the most important "resource" for future development.
In the world of artificial intelligence, there is no AI computing power, just like there is no electricity.
AI computing power has penetrated into every aspect of our life and production. Take the familiar hospital pharmacy as an example:
Many people have personal experience of the long wait and anxiety when they get the prescription list and wait in line in front of the pharmacy, and the medical staff take the prescription and distribute the medicine according to the list. Now, some enterprises have developed robots that use artificial intelligence technology to automatically reissue drugs. Using 3D visual positioning, robot intelligent grasping and intelligent visual recheck technology, the accuracy of 100% drug replenishment is guaranteed, and the efficiency is higher. The delivery speed can reach 2500 boxes per hour, and an order can be processed in 8 seconds. In the process of drug distribution, the system can dispatch and transport drugs independently without manual intervention.
The most intuitive change it brings is that the time for drug sorting can be shortened from 50 seconds to 3 seconds, and patients can get drugs in just one minute.
In this process, how to identify prescriptions and how to accurately sort and distribute drugs? To realize these functions, it depends on how fast, how good and how accurate the AI system for this robot can be. This is the computing power of AI.
If there is electricity, there must be power plants and power grids.
AI computing power is so important, but many companies lack sufficient funds to build their own AI computing power. So how to meet the demand for AI computing power, and how can countries and cities provide enough AI computing power support to promote the development of AI industry and win the competitiveness in the AI ? ? era? The answer is to make AI computing power a public resource and build a new infrastructure.
This new computing infrastructure is the artificial intelligence computing center. To use the analogy of the electrical age, it is to build power plants and power grids.
The first is to speed up the construction of artificial intelligence computing centers.
Artificial intelligence computing center is a complete system based on artificial intelligence computer cluster built on artificial intelligence chips, covering infrastructure, hardware infrastructure and software infrastructure. Its core function is from the computing power of the bottom chip to the full stack capability of the top application providing artificial intelligence, that is, the output AI computing power.
The artificial intelligence computing center is not only a platform for providing public computing services, but also an incubation platform for application innovation, a platform for industrial agglomeration development and a platform for cultivating scientific research and innovative talents. Only by playing these roles well at the same time can we get through "political use in Industry-University-Research", concentrate our greatest strength, form industrial integration and enhance the competitiveness of AI.
At present, the whole world is accelerating the construction of artificial intelligence computing centers. The United States, in particular, on the one hand tries its best to suppress the development of other countries, on the other hand, it invests huge sums of money to strengthen the development of its own artificial intelligence. The Biden administration once announced a $300 billion investment plan to defend the leading position of the United States in the field of artificial intelligence. One of the most important investments is to strengthen the new infrastructure of data centers and intelligent computing centers.
Of course, China will not easily miss the opportunities brought by the development of artificial intelligence industry. As early as 20 17, the State Council released a new generation of artificial intelligence development plan, and stressed the need to "build an efficient computing infrastructure". During the epidemic last year, the central government further put forward a new infrastructure strategy, and strengthening the construction of data centers and artificial intelligence computing centers was the top priority of the whole new infrastructure.
Because, without strong computing power, it is difficult to achieve its construction goals in seven major areas of new infrastructure focusing on digitalization.
Under the guidance of national strategy, the call of market prospect, and even the pressure of economic transformation and upgrading, many local governments have taken active actions to lead the construction of artificial intelligence computing centers, and on this basis, improve the local computing capacity and build the core competitiveness in the digital age.
On May 3, this year, in the 15 National Artificial Intelligence Innovation and Development Experimental Zone approved by the Ministry of Science and Technology, Wuhan Artificial Intelligence Computing Center was first built and put into operation. In the future, the artificial intelligence computing center in Xi 'an has also started, and the construction of artificial intelligence computing centers in other provinces and cities is also under planning.
After the operation of Wuhan Artificial Intelligence Computing Center, it provided computing support for economic development, scientific research innovation and enterprise transformation in Wuhan and even Hubei.
For example, relying on Wuhan Artificial Intelligence Computing Center, Wuhan University has built the world's first special framework for remote sensing, Wuhan Luojia Network, which has made a major breakthrough in the full image analysis of "large-format, multi-channel" remote sensing images and the minimum reading and processing of data sets.
For another example, with the support of the center's computing power, the Institute of Automation of Chinese Academy of Sciences developed the world's first multi-modal large model of video generation-Zidongtaichu. As the industry's first 100-billion-level three-mode model, Zidongtaichu's video understanding and description performance has achieved the first in the world. It not only has the ability of multi-task joint learning, but also can realize graphic search and audio, short video and MV production based on AI through learning, which greatly shortens the creation time of audio and video.
In September, the first artificial intelligence computing center in northwest China was located in Xi 'an, a major town in northwest China. The first phase of Xi 'an Future Artificial Intelligence Computing Center with 300P AI computing capacity is planned, which will provide accurate and reliable model training and reasoning based on the construction of Shengteng AI basic software and hardware platform.
Xi 'an Future Artificial Intelligence Computing Center has signed several projects, including xidian university Remote Sensing Project, Northwestern Polytechnical University Voice Model Project, MindSpore Laboratory of Shaanxi Normal University, etc. While supporting the development of Xi 'an's "6+5+6+ 1" modern industrial system, it will also expand Xi 'an and even the whole northwest region's artificial intelligence industrial cluster, providing computational support for the development of the northwest region's artificial intelligence industry.
The computing capacity utilization rate of Xi 'an Future Artificial Intelligence Computing Center, which has just been launched, is close to full capacity. The thirst for computing power of local enterprises, scientific research institutions and universities can be seen.
Secondly, the computing power resources of the artificial intelligence computing center are used efficiently.
When more and more artificial intelligence computing centers are built and put into operation, how to make their computing power more efficient and serve more industries and enterprises? How to avoid uneven distribution of computing power and inconsistent use efficiency? How to enable places where there is no artificial intelligence computing center to enjoy the enabling of AI computing power? The interconnection, cooperation and sharing between artificial intelligence computing centers have become issues that need to be considered in all walks of life.
This requires an artificial intelligence computing network, just as the power grid is to power plants and power users.
With the help of computing power network, we can connect the nodes of artificial intelligence computing centers distributed all over the country, dynamically and in real time realize the overall distribution and scheduling of computing tasks according to the state and demand of computing power resources, form a nationwide computing power network for sensing, distributing and scheduling artificial intelligence centers, and then gather and * * * enjoy computing power, data and algorithm resources on this basis.
Most importantly, with this network, more industries and enterprises can use AI computing power as they do now.
So, what role will the network of computing power play in the whole artificial intelligence industry?
The first is the integration of computing power, that is, the high-speed interconnection of computing power resources in different regions and cities, realizing the reasonable scheduling of computing power across nodes and the elastic allocation of resources, which is conducive to improving the utilization rate of artificial intelligence computing centers and saving the overall energy consumption. In the future, it can support cross-node distributed learning and provide super computing power for the research of large models.
Second, data aggregation. The government takes the lead in cooperating with enterprises in various industries. After the public data in the field of artificial intelligence is opened, it can rely on the artificial intelligence computing center to aggregate high-quality open source and open artificial intelligence data sets, which can promote the algorithm development and industry landing in the field of artificial intelligence.
Finally, integrate ecology, unify the interconnection standards and application interface standards between artificial intelligence computing centers, realize the opening of large-scale model capabilities in the network and the enjoyment of application innovation results, strengthen cross-regional scientific research and industrial cooperation, provide more resource choices and more convenient cooperation methods for users across the country to carry out artificial intelligence application innovation, accelerate industrial integration, and activate industrial integration.
Simply summarizing the computing power network is to gather big data+big computing power, enable big models and major scientific research innovations, and incubate new applications. Then, the computing power is networked to reduce the cost of computing power and improve the computing energy efficiency.
In the three-year action plan, the Ministry of Science and Technology pointed out that it is necessary to "lay out a number of artificial intelligence computing centers to form a wide-area collaborative artificial intelligence platform". Under the guidance of this plan, artificial intelligence computing centers have landed in many cities. On the just-concluded HC202 1, the common ground of more than 20 cities with artificial intelligence computing centers lit up the "artificial intelligence computing network".
This computing network in the artificial intelligence industry has begun to be woven.
The construction of AI computing power is not long-term but imminent.
In 2020, Charles Leiserson, a computer scientist at MIT and a pioneer of parallel computing, wrote in Science magazine:
Deep learning is approaching the computing power limit of existing chips.
In fact, in the past ten years, the demand for computing power of the best AI algorithm has increased by almost 6.5438+0.00000 times, doubling every 3.4 months on average.
In contrast, the growth of global AI computing power is very limited.
The huge gap between supply and demand urges governments of all countries, especially China, the United States, Europe and Japan, which are leading in AI technology, to make great efforts to build AI computing power.
Without strong AI computing power, a country or region will be at a disadvantage in the future scientific and technological competition.
Judging from the current progress of computing infrastructure construction, Shenzhen, Wuhan, Xi 'an and other cities have all built artificial intelligence computing centers and put them into operation, Chengdu, Henan and other cities are under construction, and the artificial intelligence computing centers in Beijing, Nanjing and Shanghai are accelerating their construction, which is also ready to go.
In the future, once all the artificial intelligence computing centers are completed, the artificial intelligence computing network will be formed, which will not only provide cross-regional and sustainable supercomputing capabilities for the society. But also can realize cross-regional scientific research and industrial cooperation, enable large-scale models and major scientific research innovations, and incubate new applications in thousands of industries.
In the end, artificial intelligence will empower more industries and scenarios, allowing us to be invincible in the future industrial and technological competition between countries.
——END——
All rights reserved. Reprinting without permission is prohibited!