Observation | What problems are highlighted by the recent frequent software “crashes”? What is the urgent need to break through in the “Battle of 100 Models”?

As new generation information technologies such as large models accelerate innovation, human society is entering a development stage marked by digital productivity. Software undoubtedly plays an important role, becoming a new infrastructure that integrates elements of people, machines, and things to support social and economic development.

On December 2, the "China Software Conference" hosted by the China Computer Federation was held at the Shanghai International Convention Center. Registered representatives came from 10 countries and regions, with more than 2,300 registered people and 326 participating units, creating China Software The number of attendees and units at the conference reached new highs.

What are the opportunities and challenges in the software industry today? What problems have been highlighted by the recent frequent "crashes" of software systems? How do colleges and universities cultivate software talents? Jiefang Daily·Shangguan News reporter interviewed several experts attending the meeting.

"We are entering a digital era, where software defines everything and the world." said Mei Hong, academician of the Chinese Academy of Sciences, chairman of the China Computer Federation, and professor of Peking University.

In Academician Mei’s view, any industry is inseparable from software, especially in the current context of digital transformation, where many system functions need to be implemented by software. The manufacturing industry is undergoing a process of digitization and softwareization. For example, "for new energy vehicles, the car can be regarded as a mobile computing system. With the help of an on-board operating system, many car functions can be software-based."

Benjani Straustroup, the inventor of the C language, once said that human civilization will run on software. Academician Mei also said that intelligent software integrated with artificial intelligence technology will promote the development of the digital economy and society, and is also an important opportunity facing the software industry.

At the same time, the rapid development and application of artificial intelligence deep learning has also brought many challenges to the software industry. "Faced with the current 'black box' risks of artificial intelligence, how to use 'white box' software to define the world and make it more credible and reliable is an important challenge we face." Academician Mei believes.

[New application scenarios may bring new opportunities]

A few days ago, Didi Chuxing App experienced a glitch, and many users heatedly discussed that large platforms require continuous investment in software research and development.

Talking about the reasons for the "collapse" of software systems, Jin Zhi, executive director of the China Computer Society and professor at Peking University, said that the development of software systems is a complex system engineering, but compared with the structure of other engineering products, software is "easy to "Modifiability" makes software products easier to evolve and update, that is, it is convenient to add new functions, but it also brings great challenges to the assurance of software quality. "In the process of software development and software update and evolution, testing or verification is needed to ensure that the software product does not malfunction."

"In fact, some domestic application software already has a good reputation, such as WeChat and Alipay, but our basic software technology started late, and there are still many core technologies that need to be broken through." Jinzhi said, however, new application scenarios may Bring new opportunities. The previous operating system platform mainly managed computing resources. In the future, under the Internet of Everything, more physical devices will be connected. "If we are the first to enter new application scenarios, we may occupy a better position in this track."

In addition, software is mainly updated iteratively through use. Jinzhi therefore called for the need for a wide range of application scenarios of considerable scale across society to promote the development of domestic software.

[Software talent training is closely related to application scenarios]

Should the training of software talents be more important in universities or in enterprises?

Jinzhi believes that this requires universities to systematically teach knowledge points, and it is also closely related to application scenarios. Enterprises can provide application scenarios to train talents. Therefore, universities and enterprises are complementary to each other in cultivating software talents. She specifically mentioned that there were instantaneous information load peaks during the "Double 11" in previous years, and this application scenario can only be encountered in actual applications.

Regarding the relatively lagging issue of software textbooks in colleges and universities, Jin Zhi introduced that starting from 2020, the China Software Conference has arranged a research forum on software education to help standardize and update the core teaching material system. “On December 1, we just released an updated version. A core curriculum system for software engineering graduate students to promote the cultivation of software engineering graduate talents."

At present, the large-scale model industry is very hot. Industry insiders jokingly call it no longer a "war of hundreds of models" but a "war of hundreds of models." "What needs to be paid attention to is that many models are only specialized models in a certain field formed on the basis of various open source large models. What is urgently needed now is the base large model that supports the development of large models." School of Computer Science and Technology, Fudan University Dean Yang Min said that some leading domestic manufacturers and well-known universities are exploring this and have begun to lay out the research and development of hardware platform technology based on domestic computing power, as well as technological breakthroughs in platform support software.

In February this year, Fudan University released MOSS, a large conversational model similar to ChatGPT. This is the only high-impact open source large model in China that is independently released and continuously maintained by a university. In June this year, the largest model among Chinese universities was launched. The scientific research intelligent computer on the cloud has now entered a stable operation stage. In fact, from the birth of "my country's first electronic analog computer" in 1956 to the support of important conferences, high-speed rail fault detection, and driverless safety evaluation in recent years, Fudan computer scientists have created many firsts in China's computer teaching and research, cultivating A large number of talents with international vision and competitiveness have been recruited.

"In the past, the training of software and hardware talents in domestic universities was completely separate. Fudan has now launched cross-talent training and implemented part-time teachers in departments related to computer science to cultivate high-end talents with a cross-section of software and hardware." Yang Min introduced that Fudan recently The university has also been approved as a national artificial intelligence industry-education integration innovation platform, and will carry out scientific research and talent training around artificial intelligence technology theory, platform and application.