AI Large Models Expand Across Industries

[baoshi.rao]

Since the beginning of this year, domestic and international tech companies have been competing to expand their commercial footprint around large models. According to statistics, by the end of July, China had released a total of 130 large models, with 64 launched in the first seven months alone. Domestic large models are entering a period of rapid development, accelerating digital and intelligent transformation across various industries.

Empowering Scenario Applications

The ongoing Hangzhou Asian Games is the first-ever "Cloud Asian Games," with 100% of its core systems running on Alibaba Cloud. At the Hangzhou Asian Games Digital Technology Experience Center, visitors can use the art installation "Imagine Future Asian Games" to generate Asian Games-themed images with Alibaba Cloud's painting creation model, "Tongyi Wanxiang."

"Alibaba Cloud is one of the earliest tech companies to invest in large model R&D, starting its layout in 2019," said Zhou Jingren, CTO of Alibaba Cloud. This year, Alibaba Cloud has successively launched multiple self-developed Tongyi large models and applications, which now possess the capability to process or generate multimodal content such as text, speech, and images, and are widely applied across various industries.

On September 21, Huawei Cloud released the Pangu Medical Model at Huawei Connect 2023. This model has clinical assistance capabilities and will effectively support applications in medical testing, clinical diagnosis, and personal health management. Zhang Ping'an, Executive Director and CEO of Huawei Cloud, introduced that the RunDa Medical Model · Liangyi Xiaohui, developed based on the Pangu model, will be deployed in 4,000 hospitals to assist doctors in report interpretation and diagnosis.

On September 5, the 2023 Baidu AI Cloud Summit announced that the Baidu Intelligent Cloud Qianfan Large Model Platform now serves nearly 10,000 monthly active enterprises, covering over 400 scenarios in industries such as finance, manufacturing, energy, government, and transportation.

Huang Shi, Principal Researcher at the China Mobile Research Institute, stated that China Mobile has released two industry-specific large models: the "Jiutian · Haizheng Government Model" and the "Jiutian · Customer Service Model." The company is deepening collaborations with market demand in fields like government affairs, healthcare, judiciary, and education to advance the application of large model technology in real-world scenarios.

Technical Challenges Remain

Behind the transformation of industries by large models lie numerous complex technical challenges.

"As parameters and data volumes increase, the demand for computing power from large models also rises, leading to a shortage of GPU chips that provide this computing power," Huang Shi suggested. The next step is to build a new infrastructure centered on computing power networks, leveraging network strength to alleviate the shortage of computing power. "We need to connect ubiquitous computing resources through computing power networks, utilizing the advantages of network clusters to break through the performance limits of single-point computing power. By optimizing the allocation and use of network-wide resources, we can mitigate the issue of insufficient computing power."

"Finding suitable application scenarios is the biggest challenge for large models," said Xin Zhou, General Manager of Baidu Intelligent Cloud AI and Big Data Platform. Baidu has studied over 400 scenarios, each requiring different methods and industry expertise for large model application. In some benchmark industry scenarios, Baidu's algorithm experts summarize common features for different customers and needs, continuously accumulating experience, tools, and refining the toolchain.

Zhang Yuxin, CTO of Huawei Cloud, believes that as large model capabilities emerge, model parameters will soon reach trillions, placing higher demands on stable computing power, large-scale parallel training, and overall architecture design. Meanwhile, with the rapid growth of dataset sizes, challenges such as efficient data storage, cleaning, and labeling still need optimization.

Building an Open Source Ecosystem Together
Artificial intelligence is driving transformations in production, lifestyle, social governance, and scientific research, providing new momentum for high-quality economic and social development. Collaborative efforts to build a large-scale model ecosystem and lower usage barriers have become a shared goal among tech giants developing such models.

"Alibaba Cloud aims to collaborate with the entire industry to build an open-source ecosystem for large-scale models, further reducing the barriers to their adoption," said Zhou Jingren. On one hand, Alibaba Cloud has independently developed the Tongyi large-scale model series; on the other, it provides services for startups and developers working on large-scale models to foster a thriving ecosystem. The Alibaba Cloud ModelScope community, an open-source hub for large-scale models, has already attracted over 2.3 million AI developers and hosts more than 1,200 high-quality AI models, with total downloads exceeding 85 million.

Qiu Yong, Secretary of the Party Committee of Tsinghua University, emphasized the need to focus on global technological frontiers and national strategic demands to drive research innovation. He highlighted the importance of cultivating top-tier AI talent domestically to support breakthroughs in core technologies and enhancing the integration of innovation and industrial chains in AI to strengthen education's role in high-quality development.

To safeguard the development of large-scale models, robust top-level design and security measures are essential. Wang Zhigang, Minister of Science and Technology, stressed the importance of seizing opportunities in the new wave of AI development. He advocated a systematic approach to understanding AI's trends and patterns, emphasizing foundational strength, growth promotion, and security assurance. He called for structured R&D efforts, the integration of innovation and industrial chains to build an innovation ecosystem, and a governance framework guided by the principle that "development is the greatest security."