China-Singapore Jingwei, Feb. 25 (Xinhua) -- What changes will SORA bring to the financial industry?
Written by Li Lihui, former Governor of Bank of China.
On November 30, 2022, Microsoft's OpenAI launched ChatGPT, and 14 months later, on February 16, 2024, OpenAI's Sora was launched. Sora is capable of producing 1-minute hi-fi based on the user's text prompts, and can achieve seamless connection and extension.
ChatGPT is probably the most advanced generative AI model at present, while SORA is currently the most advanced generative model. This is a milestone in the iteration of artificial intelligence technology.
The first is to initiate the transformation of content production methods and human-computer interaction methods. Generative AI models can learn from unstructured data formats to generate new unstructured content, including text, audio, images, and more, that can be adapted to a variety of tasks, making a significant impact on the entire digital technology industry chain. **The core underlying layer of the generative model SORA integrates the diffusion program and the transformer program, and the core of the underlying technology lies in the ability of multimodal perception, learning and interaction. SORA breaks through the limitations of text interaction, is able to perceive, understand and simulate the dynamic physical world, and is able to interact and learn from the real world. It can be expected that multimodal AI will change the operation mode and business model of the advertising, media, animation, film and television industries, and has the potential to accelerate the innovation of intelligent driving technology based on vision (rather than radar), and the innovation of intelligent biomedical technology based on gene mapping and cell image analysis. The AI digital ** program is called Co-pilot (co-pilot), which can perform tasks such as knowledge learning, environmental perception, and action planning on behalf of the owner.
The second is to expand from the general large model to the vertical model (vertical model). The basic and applied research of artificial intelligence is integrated into various fields of digital technology, including big data, cloud computing, blockchain and the Internet of Things, including natural language processing, virtual augmented reality, human-computer interaction and knowledge graph, computer vision, biometrics, as well as robotics, space technology, biomedicine, optoelectronic technology, autonomous driving, etc., forming a complex system with AI technology as the core.
The third is to directly create business value. AI models can directly reduce the cost of knowledge application, thereby creating business value. It can automate routine tasks and improve the input-output ratio; It can improve the degree of automation of industrial, logistics and service processes and save marginal costs; It can diagnose the operation defects in all aspects of production and operation, and improve production efficiency and management efficiency.
Smart finance is still in the early stages of assistant + assistant. One is to improve product innovation and customer service. For example, ICBC's digital intelligence trading system covers more than 100 business scenarios of exchange rate, interest rate and commodity trading, China Merchants Bank's AI Xiaozhao intelligent assistant realizes the best financial intelligent advisory service for tens of millions of users, and the digital employees of Pacific Insurance can provide daily office, software skills, knowledge quizzes, data processing, professional scene task execution and other services as personal assistants, Guotai Junan's Junhong investment and financial intelligent customer service APP has **, options, foreign exchange, wealth management, margin financing and securities lending, Cross-scenario business delivery capabilities such as investment advisory. The second is to improve operation management and risk control. For example, MYbank's Bailing system applies human-computer interaction technology to achieve personalized risk control for millions of users, Ping An Property & Casualty's natural disaster risk management platform applies spatial data and satellite remote sensing images to provide remote survey, accurate damage assessment and rapid claim settlement, and Taikang Insurance's underwriting and claims recognition platform can provide core functions such as medical imaging and medical record quality inspection, customer health assessment, and false claim screening.
In the field of digital finance, generative and multimodal AI technologies have the potential to bring about new changes.
The first is to realize the humanization of human-computer interaction with high fidelity. For example, an intelligent financial robot that applies the underlying technology of multimodal SORA may be able to dynamically capture, perceive in real time, correctly understand the customer's language and expression, accurately judge the customer's risk preference and business demands, and provide customers with the best service plan by using human-warm expressions, so as to solve the problem of cold machine service. Intelligent financial robots can not only replace blunt machine customer service, but also have the potential to become a landscape of counter service.
The second is to realize the intelligence of the whole process of image management. For example, the intelligent image management system using SORA's underlying technology is applied to the health and medical insurance business, which can conduct medical professional quality inspection and classification of customers' medical records and medical images, approve health assessments, and screen false claims; Applied to banking business, it can audit the authenticity of various bills and contracts in the front and middle and back offices, screen cloned bills or false contracts, correctly extract data and incorporate it into the accounting system in real time, and improve service quality and operational efficiency.
The computing power level of AI depends on computing hardware, algorithm software, and data resources. Computing infrastructure includes computing hardware and algorithm software. Technology has borders. The competition for computing power will be national-level competition between major economies, as well as enterprise-level competition between capital giants and technology giants.
In terms of data resources, China and the United States have their own strengths and weaknesses.
Relying on the advantages of data resources accumulated by long-term development, the United States and other developed countries in the West have built a Western-led data resource supply pattern in the fields of knowledge and academia. For example, the MEDLINE of the National Library of Medicine is the world's most authoritative biomedical literature database, including more than 5,200 biomedical journals published in more than 70 countries and regions since 1950, with an annual increase of 300,000-350,000 records, covering basic medicine, clinical medicine, environmental medicine, nutrition and hygiene, occupational diseases, health management, health care, microbiology, pharmacy, social medicine and other subdivisions. There is an exponential gap between the "Chinese Biomedical Literature Database CBM", which was put into operation by the Chinese Academy of Medical Sciences in 1994, and Medline.
Relying on the advantages of population size, market size and mobile services, China has formed a world-leading data resource supply pattern in the field of market transactions and citizen behavior. Massive data on market transactions and citizen behavior is a valuable resource for China's digital finance development.
It is important to note that the limitations of the data sharing model may affect the in-depth development of data value. For example, the financial data and behavioral data involving residents and enterprises are scattered in different local systems such as financial institutions, financial supervision, business administration, taxation, and customs, and the level of sharing is not high, forming an administrative data gap. For another example, the scale of mobile payment users in China is as high as 900 million, digital payment has become the main data entrance, and the Internet platform has super-large-scale personal and enterprise data, but the data association and data sharing between the Internet platform and financial institutions have not yet reached a mature model, the efficiency of data sharing is not high enough, and the value of data resources has not been fully explored.
It is important to pay attention to the constraints imposed by the geopolitical environment on the supply of data resources. The technical barriers set by the United States and Western countries against China have been escalating, and now they are high-end chips and core software, and the next step may extend to the field of data resources.
Attention must be paid to the restrictions imposed by intellectual property protection on the application of data resources. On December 27, 2023, the New York Times filed a lawsuit against OpenAI for scraping copyrighted text, unveiling the prelude to data property rights disputes in the context of generative AI technology.
AI models in different scenarios have different requirements for data resources. Whether it is the application of multimodal artificial intelligence technology to achieve intelligent financial iteration, or the application of big data to create short-term and long-tail inclusive finance, it is necessary to build a high-quality and efficient data element sharing system.
The Opinions of the Communist Party of China on Building a Basic Data System to Better Play the Role of Data Elements (Article 20 of Data) clarifies the norms of the data property rights system, the data element circulation and trading system, the data element income distribution system, and the data element governance system. Improving data quality, expanding data scale, promoting data circulation, realizing data sharing, exploring data value, and protecting data security are the driving forces for the development of the digital economy. The focus is to improve the market system and mechanism of data elements, fill the data gap, enhance the sharing and inclusiveness of data elements, improve the quantity and quality of data element supply, effectively prevent and resolve various data risks, and deepen openness and cooperation to achieve mutual benefit and win-win results.
In terms of digital security, the cautious view is that the latest artificial intelligence technology represented by generative AI models and sora models is in its infancy, and the algorithms and models are not clear and transparent enough, and the immature artificial intelligence technology will be put into the high-risk financial field, which may amplify existing risks and generate new risks. Therefore, intelligent financial innovation must be based on the premise of AI trust and AI security, achieve financial equality in line with ethical standards, ensure financial efficiency in line with security standards, and create an innovation model that conforms to economic laws.
For smart financial innovation, relying too much on self-regulation may lead to monopoly and industry risks, while too strict regulation may inhibit innovation and industrial development. The principles that can be considered are: technological innovation strives for "high school", "high" is to lead innovation to occupy the high ground, and "medium" is the down-to-earth Chinese plan; Risk management and control strives to be "small at the beginning", "at the beginning" is to have the ability to eliminate risks in the bud, and "small" is to minimize risk probability and risk costs.
This requires accelerating innovation in smart financial regulation. For example, laws and regulations should be formulated to clarify the boundaries of responsibilities of all participants in smart finance, including the basic principles of smart finance supervision, the responsibilities and powers of regulators, and the smart finance business norms of financial institutions; Establish a penetrating, integrated, and cross-local intelligent financial collaborative supervision system to achieve regulatory information sharing; Establish an intelligent financial technology audit and certification system, and improve the testing platform, tools, standards and methods of AI large models; Establish an intelligent financial risk analysis and monitoring system to identify, assess and warn abnormal transactions and market manipulation in advance in a timely manner, and proactively prevent systemic risks; Breakthrough innovations in smart financial services are allowed to be piloted under the regulatory sandbox mechanism, and regulatory experience and data support will be accumulated; Actively participate in the construction of international rules and general standards for digital technology for the digital economy, strengthen international regulatory cooperation and exchanges on smart finance, and strive for China's right to speak in the construction of international rules for the digital economy, and for China's "right to position" in the construction of general standards for digital technology. (Zhongxin Jingwei app).
This article is selected and edited by the Sino-Singapore Jingwei Research Institute, and the works produced by the selection are all rights reserved, and no unit or individual may use it in any other way without written authorization. The views involved in the selected content only represent those of the original author and do not necessarily represent the views of Sino-Singapore Jingwei.
Editor in charge: Zhang Zhihan.
For more exciting content, please pay attention to the official WeChat of JWVIEW***