He Han, member of Chinese People’s Political Consultative Conference： Accelerate the expansion of the application scenario of artificial intelligence big model and attach importance to the generation
(Member Chinese People’s Political Consultative Conference, Deputy General Manager of Tianyu Mathematics Department, Chairman of Shanxi Data Flow Valley He Han)
Our reporter (chinatimes.net.cn) Yu Yujin reports from Beijing.
As an important driving force of the new round of scientific and technological revolution and industrial transformation, artificial intelligence big model technology is continuously injecting new kinetic energy into economic and social development. From GPT to Sora, from unimodal to multimodal, from single intelligence to general intelligence, large model technology has led a new round of global artificial intelligence innovation upsurge. On the occasion of the two sessions this year, He Han, member of Chinese People’s Political Consultative Conference, deputy general manager of Tianyu Mathematics Department and chairman of Shanxi Data Flow Valley, also brought a number of proposals, including accelerating the expansion of application scenarios of artificial intelligence big model technology, promoting the interconnection of intelligent computing resources, alleviating the contradiction between supply and demand structure, promoting the interconnection of data trading institutions and promoting the coordinated development of multi-level data factor markets.
In an interview with the reporter of China Times, He Han said that it is of great significance to accelerate the expansion of the application scenarios of artificial intelligence big model technology to promote the application of artificial intelligence at a higher level, cultivate new quality productivity and better support high-quality development. In addition, he also believes that 3D data is the basic element of the next generation of big models.
Promote the ecological construction of large-scale model application
The application of large-scale model empowerment industry is booming. Based on the generalization ability and emergent characteristics of large-scale model, it shows amazing performance and great development potential in more and more vertical industry scenarios, which has attracted great attention from all industries. Scene-driven has become an important model for technological innovation and industrial growth of large-scale model.
To speed up the expansion of the application scenarios of artificial intelligence big model technology, He Han put forward four suggestions. First, he encouraged the construction of a multi-level application scenario supply and demand docking platform to provide the market with "visible and tangible" application scenarios and promote the transformation of application scenario expansion from "issuing documents" to "giving policies" to "building mechanisms" and "creating opportunities"; The second is to implement the large-scale model empowerment demonstration application promotion plan nationwide; Third, vigorously promote the ecological construction of large-scale model application and build an open, shared and collaborative industrial ecological system; Fourth, pay attention to the training of talents facing the era of general artificial intelligence, and train a group of comprehensive talents who know both industry technology and big model technology.
"Industry, finance, commerce, transportation, medicine, government affairs, education, cultural tourism, media and other industries should be encouraged to dig deep and actively create various scenarios in which artificial intelligence big model technology can be applied, and promote the application of big model empowerment industry Internet platform." He Han further stated that the existing production, service and management methods will be upgraded through large-scale model technology to realize business process innovation and reengineering, and at the same time, new functions will be developed, new products will be formed, and new applications will be expanded to realize the deeper integration of artificial intelligence technology and industry.
In 2023, the "Hundred Models War" was launched in the AI field, but the development paths at home and abroad were not the same, and the domestic large model focused more on application landing.
In this regard, He Han analyzed to our reporter, "There are three necessities to focus on the expansion of application scenarios this year. First, large models are currently used in media, finance and other industries, and there are few cases in other industries, let alone popularization. Promoting the combination of large model technology with wider industry applications such as manufacturing, energy and transportation will help the whole economy to reduce costs, improve quality and increase efficiency; Second, there will be new formats in the process of (large model application), such as AI teaching assistant, AI diagnosis, etc., which is a new growth point; Third, feedback from industrial applications will in turn be helpful for large-scale model technology innovation. "
He Han also stressed that the "artificial intelligence+"action mentioned in this year’s government work report is also aimed at promoting the deep integration of artificial intelligence technology with various fields of economy and society, which will also help achieve the growth target of around 5% this year.
Improve 3D data generation and acquisition
As a new factor of production, data plays an increasingly prominent role in improving production efficiency and promoting the economy. High-quality data is the magic weapon for the value jump of large models.
He Han believes that the main challenge facing the technical development of AI big model and its deep integration with industrial applications at this stage is insufficient matching of data elements. It is necessary to strengthen the collection, utilization and development of data from various industries and reserve high-quality industrial data sets. He further explained that although we have many overall data resources, high-quality industrial data sets are still scarce resources. Recently, the United States began to restrict data transmission to China, which is also aggravating the problem. It is imperative to strengthen the collection, development and utilization of data in various industries, and establish high-quality industrial data sets in all fields.
He Han particularly emphasized that in addition to traditional data, attention should also be paid to the generation and collection of 3D data, which is the basic element of the next generation of 3D large models.
Compared with the traditional big model, He Han believes that the 3D big model can provide more accurate and comprehensive analysis and decision-making, and has obvious advantages in dealing with tasks related to spatial perception and interaction.
According to the reporter’s understanding, the 3D big model can be applied to many scenes including autonomous driving, industrial design and manufacturing, architecture and engineering, medical care, film and television special effects, etc. Taking medical care as an example, 3D large model can make personalized 3D medical devices or organ models according to patients’ CT or MRI scanning data, which can help doctors to plan and train operations.
He Han said that 3D large models need 3D data to feed, and 3D data can capture depth information, which enables machines to have a more accurate understanding of the shape, structure and position of objects, thus creating more realistic models and scenes, providing more intuitive visualization effects, and conducting more accurate analysis and measurement; At the same time, 3D data can support more interactivity, and users can operate, rotate, zoom and view different angles in the 3D environment, providing a better user experience.
Talk about new quality productivity in detail
During the two sessions this year, the hottest word was "new quality productivity". As early as the beginning of 2024, the Political Bureau of the Central Committee focused on the development of new quality productivity for the first time, and proposed that "developing new quality productivity is an inherent requirement and an important focus for promoting high-quality development." The government work report released on March 5 even pointed out that we should accelerate the development of new quality productive forces, give full play to the leading role of innovation, and promote industrial innovation with scientific and technological innovation.
He Han believes that new quality productivity is a combination of new technologies and new elements. The field of digital economy is the combination of big model technology and data elements. He said that at present, the mode of driving growth by the low-cost comparative advantage of traditional production factors has peaked, and data elements have the characteristics of sharing and copying, breaking the limit of limited supply of traditional factors on growth. To achieve the growth target of about 5% this year, we must give new impetus to the growth of new quality productivity.
He Han combined with his own industry explanation, taking Shanxi Data Traffic Valley as an example, in fact, it is a common phenomenon to develop digital economy in the central and western regions, lacking talents, funds, technology and platforms, but for data elements, the eastern, central and western regions are at the same starting line.
"We have created a low-cost comparative advantage in terms of data elements by means of market-oriented industrialization, so as to attract the continuous convergence of data elements in various industries to form resource advantages, so as to drive talents, capital and technology to land along the industrial chain, supply chain and service chain, and combine them with traditional advantageous industrial scenes such as Shanxi coal coke, steel electricity and new energy, transform old industries, and at the same time give birth to new formats, and embark on a new path of developing digital economy in the central and western regions." He Wei said.
The reporter of China Times learned from Shanxi Data Flow Valley that in the past three years, the transaction scale of data elements in Shanxi Data Flow Valley was 4.3 billion yuan, and the accumulated revenue of enterprises entering the valley was 35.2 billion yuan. The average input of data elements of 1 yuan produced about 8 yuan. Economic output.
He Han believes that the above data is the intuitive embodiment of the multiplier effect of data elements, and the embodiment of new growth brought by new quality productivity. The reason why there is such a large multiplier effect, He Han analyzed, on the one hand, we continue to promote the combination of large model technology with rich application scenarios and industry data elements of Shanxi’s characteristic advantageous industries, and train to build large models in vertical fields to help traditional industries realize digital transformation, improve production efficiency and quality, reduce production costs, and reduce energy consumption and environmental pollution; On the other hand, with low-cost data traffic supply, it has also cultivated new formats such as short video, live e-commerce, e-sports games, virtual digital people, meta-universe, and new individual economy, and has become an important carrier for attracting employment and incubating entrepreneurship.
Editor in Charge: Huang Xingli Editor in Chief: Han Feng