Behind the Artificial Intelligence Boom—A Perspective on New Trends in Computing Power

The development of artificial intelligence technologies such as ChatGPT has led to a further increase in global demand for computing power. To welcome a new round of artificial intelligence boom, where will China’s computing power grow? What kind of computing power is more accurate and efficient? The reporter recently interviewed industry experts to look forward to the development trend of computing power.

Big models call for “big computing power”

In the new wave of global artificial intelligence technology development, large models are a key word. ChatGPT and other similar models make significant progress precisely by learning from massive amounts of data.

Chen Runsheng, an academician of the Chinese Academy of Sciences, said that large artificial intelligence models are undergoing tremendous changes at the application level and will trigger changes in many fields in the future, while also consuming a lot of computing power.

“Large models must have strong computing capabilities,” Zheng Weimin, an academician of the Chinese Academy of Engineering, said that ChatGPT is the result of relying on large-scale computing infrastructure.

Recently, the New Generation Artificial Intelligence Development Research Center of the Ministry of Science and Technology released a research report titled "Research Report on China's Artificial Intelligence Large Model Map" at the Zhongguancun Forum. The report shows that the number of large models developed in my country ranks second in the world.

Lu Zhonghua, a researcher at the Computer Network Information Center of the Chinese Academy of Sciences, said that society as a whole has a growing demand for computing power, and the demand for computing power in artificial intelligence applications is even more prominent, and we must embrace this trend.

General computing power Special computing power

In recent years, the proportion of my country’s artificial intelligence computing power has been increasing. According to estimates, artificial intelligence computing power has grown by an average of about 70% per year in recent years, while other computing power has increased by about 30%. " said Yu Xiaohui, President of China Academy of Information and Communications Technology.

Chen Wenguang, a professor in the Department of Computer Science at Tsinghua University, also observed that with the deepening penetration of artificial intelligence, the utilization rate of intelligent computing centers has become higher than that of traditional supercomputing centers.

Cao Zhennan, deputy director of the National High-Performance Computer Engineering Technology Research Center, said that although computing power is often divided into supercomputing and artificial intelligence computing, the two have many similarities, and the biggest difference lies in the different requirements for computing accuracy.

"There is currently a large gap in artificial intelligence computing power, and some artificial intelligence applications have lowered their accuracy requirements in pursuit of computing performance. However, if computing power becomes easy and cheap, some artificial intelligence applications may put forward some requirements for accuracy. High requirements." Cao Zhennan said.

According to expert opinion, “The combination of general-purpose computing power and dedicated computing power will be the key to building artificial intelligence computing power infrastructure.”. On the one hand, it must meet a wide range of application scenarios, be universal, and be universal; on the other hand, it must also support some personalized application scenarios that require high calculation accuracy and efficiency, be efficient, and be dedicated.

On February 14, 2023, scientific researchers from the National Supercomputing Chengdu Center inspected the equipment in the computer room. Photo by Xinhua News Agency reporter Liu Kun

Building a computing power network “highway”

In the opinion of experts, connecting existing computing power centers with different architectures through the network can reasonably configure, share, schedule, release more computing power, and lower the application threshold.

Under the leadership of the High-tech Department of the Ministry of Science and Technology, the national supercomputing Internet work has been launched. By the end of 2025, the national supercomputer interconnection network will become a "fast track" to support the construction of Digital China, in line with the plan.

Cao Zhennan introduced that one of the construction goals of the supercomputing Internet platform is to solve the problem of uneven distribution of computing power facilities and achieve interconnection and resource sharing of computing power resources.

Qian Depei, an academician of the Chinese Academy of Sciences, said that computing power infrastructure should provide not only computing power, but also the software or application services required by users, so that computing power resources can be maximized.

I hope that in the future, big data computing power can be spread to thousands of households like electricity, realizing the vision of being able to use it at any time and anywhere. "Qian Depei said.

Source: Xinhua News Agency

The above is the detailed content of Behind the Artificial Intelligence Boom—A Perspective on New Trends in Computing Power. For more information, please follow other related articles on the PHP Chinese website!