Promote the implementation of large-scale industry models and accelerate the upgrade of industries

At the beginning of 2023, IDC (International Data Corporation), an authoritative international consulting organization, released the "2022 China Large Model Development White Paper," serving as a reference for the development of large models in China. In recent years, major domestic and foreign enterprises have actively and rapidly promoted large model training, with countless models with hundreds of billions of parameters emerging one after another.

The application of large models can significantly help industries improve quality, reduce costs, and increase efficiency, driving the development of the digital economy. However, its implementation requires multi-party participation and high requirements for platforms and algorithms. Despite the continuous expansion of training volume for large models and their applications in various industries and fields, the development level and speed of their applications vary greatly. In some fields, it is still difficult to truly implement the models and achieve effectiveness. Therefore, the opportunities and challenges of promoting the implementation of large models in industries coexist.

Application areas of large models: Exploration and limitations

Large models represent a significant milestone in the evolution of artificial intelligence from discriminative to generative approaches. Their undeniable prowess has ushered in the era of AI generalization, empowering various industries to tackle efficiency optimization challenges, drive industrial transformation, and ultimately enhance quality, reduce costs, and boost efficiency, fostering industrial transformation, upgrading, and development. Presently, the application of large models in China is relatively mature in four key sectors: finance, media, cultural tourism, and government affairs.

The financial sector, being an information-intensive industry, represents one of the best application scenarios for large model technology. Currently, numerous domestic fintech platforms have invested in the research and development of large model technology. However, the high cost of developing new applications, as well as challenges posed by data compliance and security, still loom large over the financial industry, demanding further optimization and handling.

The media sector is deeply applying large model technology to promote intelligent transformation and upgrading. For example, the integration of Baidu's Wenxin large model with the media industry has brought new models to different scenarios such as "strategy, gathering, editing, and publishing". Traditional media are stepping out of their comfort zones, applying cutting-edge technology, and keeping up with the pace of industry iteration.

In recent years, various regions across the country have been accelerating the construction of large-scale cultural and tourism models, utilizing these models to precisely promote cultural and tourism content to target audiences, achieving rapid transformation. For instance, Tencent Culture and Tourism, in collaboration with Nanjing Cultural Investment Group, has developed "Nanjing Wendu Smart Brain," the first city-level cultural industry large-scale model in China, providing Nanjing's cultural and tourism sector with richer and more intelligent cultural experiences. Meanwhile, the integration of domestic large-scale models with the cultural and tourism industry is still in its infancy. There is a need to further enhance the quality of cultural and tourism supply, improve the overall efficiency of cultural and tourism production, deeply grasp industry needs, and achieve growth through digital upgrading.

In the government affairs field, CETC Tai Chi, a subsidiary of China Electronics Technology Group Corporation, has released the generative AI large model product "Xiao Ke" to assist government, political, and enterprise users in intelligently carrying out their business. Currently, a series of intelligent applications such as text generation assistant, coding assistant, and intelligent plotting have been launched, addressing pain points in related fields such as repetitive work, insufficiently scientific decision-making based on personal experience, and low work efficiency. These applications meet the needs of cost reduction, efficiency improvement, process optimization, and work innovation. However, there are still many application scenarios in the government affairs field that lack the promotion of corresponding large model applications, which remains to be further explored.

Countermeasures and suggestions to accelerate and facilitate industrial development

To comprehensively promote the implementation of large models and empower the development of the digital economy, we can start from four aspects: top-level design, regulatory improvement, industry enthusiasm mobilization, and talent cultivation.

Firstly, we should strengthen the top-level design of large models. First, we should encourage continuous innovation in large model technology within the industry. Second, we should encourage enterprises nationwide to develop large models in an open-source manner and collaborate to address common issues such as stability and transparency. Third, we should focus on promoting scenario innovation, using it to drive technological iteration and upgrading, accelerate the implementation of large model technology in various fields, and better empower high-quality economic and social development. Fourth, we should enhance the safety and reliability of large model technology, simultaneously develop privacy protection-related technologies, and provide support and guarantees for the security of large models.

Improve domestic regulatory measures for large models. Firstly, establish a unified regulatory framework. Currently, foreign countries have relevant legislation and regulatory legal policies in the field of large model applications, but domestic regulatory measures still need to be improved. Secondly, explore targeted regulatory measures for key areas including finance, artificial intelligence, and public services, and optimize the industry development environment.

Mobilize industry enthusiasm and promote all sectors to explore the application of large models in specific scenarios. Firstly, strengthen financial support. Increase fiscal funding to primarily support innovative applications of large models. Leverage the guiding role of government investment and coordinate and integrate resources effectively. Secondly, create a more conducive industrial environment for innovation, encourage enterprises to overcome difficulties, and increase R&D investment in large models, especially those specialized for specific industries.

Focus on cultivating comprehensive talents. Cultivate a group of comprehensive talents who understand both industry technology and AI training. Firstly, universities should improve the construction of disciplines related to generative artificial intelligence, perfect the teaching system, and carry out targeted interdisciplinary integration to cultivate multi-functional and composite talents. Secondly, accelerate the integration of industry, academia, and research, vigorously build joint research and development centers between schools and enterprises, closely integrate teaching content with the front line of industry, cultivate talents with practical experience, and prevent the emergence of "patterned" talents. Thirdly, encourage innovation and allow students to freely exert their abilities and talents, laying an innovative foundation for the development of large model applications in various industries.

In today's world, promoting the application and implementation of large-scale industry models has become an inevitable trend. Only by uniting various social stakeholders, working together, and fostering collaborative development, can we continuously advance the innovation and maturity of large-scale model technology, facilitate its application and implementation in various fields and industries, thereby supporting industrial optimization and upgrading, and driving the development of the digital economy.