Huatai Partners with AI World Expo: Anticipating AI-Driven Industry Restructuring

On the morning of July 5th, Huatai Securities hosted the "Discourse on Intelligent New Trends" Technology and Finance Innovation Forum at the 2024 World Artificial Intelligence Conference. This marks the fourth year of in-depth cooperation between Huatai Securities and the World Artificial Intelligence Conference. The forum gathered a group of the company's tech ecosystem partners, including representatives from various niche enterprises and industry experts, to discuss the development trends of cutting-edge fields such as AI+Healthcare, Embodied Artificial Intelligence, and AI Computing Networks.

Han Zhencong, a member of the Executive Committee and Chief Information Officer of Huatai Securities, stated in his speech that the "three new systems" of artificial intelligence as a new production tool, data as a new factor of production, and computing power as new infrastructure are accelerating their integration with industries, which will have a profound impact. This includes enhancing the service capabilities of industries, transforming the paradigms of industry development, and fostering the emergence of new fields. Relying on technology, Huatai Securities continuously leads the company's transformation and transcendence. At the beginning of this round of generative artificial intelligence wave, the company quickly promoted the application of large models in business scenarios, using them to promote construction. It looks forward to exploring the frontier fields of AI and maximizing the commercial value of AI with more ecosystem partners through continuous docking and deep integration of technical capabilities, business scenarios, and customer resources.

Advertisement

Focusing on AI+Healthcare and Embodied Intelligence

AI+Healthcare and Embodied Artificial Intelligence are two cutting-edge fields that are highly regarded in both scientific research and investment circles. Two senior industry experts shared their insights on these topics at the forum.

Pu Zhengrong, Vice President of Siemens Healthineers Greater China and Head of Clinical Therapy Systems, introduced the global development trends of AI+Healthcare in his speech. He pointed out that with the empowerment of AI, healthcare will become more personalized and precise, providing better services for more patients. AI can promote the intelligent development of medical imaging, as well as image-guided treatment and testing equipment, supporting the clinical process from disease prevention and early detection, diagnosis, treatment and monitoring, to follow-up, replacing some routine work of doctors. AI can also support and assist doctors in making key clinical decisions and help patients with health management throughout their life cycle. He specifically mentioned that digital twin technology will have a huge application scenario in healthcare. By utilizing patient data accumulated from various examinations, a physiological model of the patient can be constructed. Based on this, different treatment plans can be simulated and results observed, thus helping patients choose the best personalized treatment plan.

Professor Wang Donglin, Director of the Artificial Intelligence Field at the School of Engineering, Westlake University, and founder of Westlake Robotics, combined the three principles of the embodied intelligence hypothesis proposed by Linda Smith (Professor at Indiana University, USA) in 2005, and shared the direction of the field in the next step: First, when facing robot scenarios and tasks, it is not possible to rely on predefined customized complex logic. For this point, multimodal large models and robot embodied large model technologies have already had universality, but there is still a need to further solve problems such as hallucination, generalization, and efficient reasoning. Second, there is a need for an evolutionary learning mechanism that adapts to dynamic environmental interactions. The current application of reinforcement learning on robots is mainly limited to the design of manual reward functions related to cerebellar functions, lacking universality in adapting to environmental scenarios and tasks, and also lacking deep coupling technology with large models to meet the integrated fusion of "brain" and "cerebellum". Third, the environment is very important for shaping physical behavior and cognitive institutions. For this, it is necessary to deal with the differences between simulation and real environments through the deep integration of robot software and hardware, and to improve the performance of embodied intelligence in real environments.

Addressing AI Computing Power Challenges

This round of artificial intelligence wave has put forward an exponential growth requirement for intelligent computing power, and computing power has become a hot spot in current infrastructure construction. How to overcome the AI computing power bottleneck more efficiently and turn challenges into opportunities? In the roundtable on "Trends and Applications of AI Computing Networks", Liu Cheng, head of Hua Tai Innovation Investment Equity Investment Department II, discussed with four outstanding start-ups in the niche track.

Regarding the current bottleneck issues in the development of domestic AI computing power, Chen Qiyan, CEO of DaoCloud, believes that it is not possible to rely solely on hardware investment. What is more important is the joint efforts of the entire industry ecosystem. Chinese technology companies have a certain influence in the basic software industry chain. The upstream and downstream of the industry chain play to their strengths, combine the entire IT architecture, manage the computing power infrastructure well, and optimize the scheduling and management of computing power resources to make computing power more free. Kang Liang, Chief Marketing Officer and co-founder of Zhuanxin Semiconductor, believes that the AI computing power network is completely different from the past network in terms of scale, latency, and power consumption. For domestic companies, a big bonus is that the AI era computing power network has not yet unified standards, and companies have the opportunity to grow. However, in terms of computing power chip hardware, Zhang Yujun, Senior Vice President of Zhongke Yushu, pointed out that domestic companies do indeed need some time to catch up, but in some important scenarios such as DPU with ultra-low latency, domestic companies can already fully compete with international giants, and there is no generational gap in technology.DaoCloud CEO Chen Qiyan introduced the company's newly released integrated computing power solution at the forum.

Global Industry Chain Restructuring in the Era of AI Large Models

Against the backdrop of the global industry chain restructuring, how will the technology and industry related to AI large models develop? Based on the in-depth research report "Global Industry Chain Restructuring in the Era of AI Large Models," Huang Leping, Chief Analyst of Technology and Electronics at Huatai Securities Research Institute, shared four important judgments:

- The global server industry scale has already exceeded the scale of the smartphone industry, and the semiconductor industry scale is expected to double from 2023 to 2030. Servers and other AI computing power-related equipment are expected to replace smartphones as the largest category of technology hardware.

- The parallel development trend of the China-US computing power hardware industry chain is clear, and there are optimistic views on the development opportunities of the domestic computing power chain. In the new round of industry chain restructuring, Vietnam, Malaysia, Japan, and Mexico are expected to benefit.

- The hardware form in the AI era is still mainly PCs and smartphones, so Apple product updates can be focused on in the short term. However, with the rise of AI interaction capabilities, the long-term development opportunities of XR (Extended Reality, extended reality technology) and embodied intelligence-related products are optimistic.

Huang Leping, Chief Analyst of Technology and Electronics at Huatai Securities Research Institute, shared the changes in hardware form and future development direction from wired Internet to mobile Internet, and then to the AI era.

- In the AI server revolution, the iteration of key server technologies such as liquid cooling, optical modules, and HBM (High Bandwidth Memory, a new type of CPU/GPU memory chip) is expected to give birth to a batch of companies with a market value of tens of billions, bringing investment opportunities for both quantity and price increase.

In addition, at the forum, Du Changqing, Head of the TMT Industry Department of Huatai United Securities Investment Banking Business Line, invited four important partner companies to discuss the progress of AI's landing application in the industry. Wu Weijie, Vice President of ZhiPu AI, combined the company's experience in landing large models in multiple industries, pointing out that to further deepen the capabilities of large models, end-cloud integration, multimodality, and open platforms will be three important directions. Wang Kaijing, Director of the Financial Business Department of SenseTime Technology, pointed out for the financial industry that customers need to "use well" and "afford to use" to truly exert the value of model empowerment for business. Zhang Xianzheng, President of 360 Zhiniao, and He Zhiyang, Dean of the iFlytek Medical Research Institute, shared their company's solutions for security and privacy risks in the industrialization of large models.

In the new wave of artificial intelligence, Huatai Securities will further leverage its financial professional advantages, deeply layout the upstream and downstream of the industry chain, serve the growth of new quality productive forces, and will also use equity investment as a link to deeply layout cutting-edge fields such as cloud-native and artificial intelligence, and jointly promote the construction of the scientific and technological innovation ecosystem with more outstanding scientific and technological innovation enterprises.