The 2023 fiscal year annual reports of the six major state-owned banks and most of the major joint-stock banks have been disclosed one after another, and the focus of the industry on fintech has begun to shift from "artificial intelligence" to the exploration of "large models". According to preliminary statistics, at present, there are 6 large state-owned banks, as well as a number of joint-stock banks such as China Merchants, CITIC, and Industrial Bank, and a total of nearly 20 financial institutions have disclosed the R&D and business application in the field of large models in their 2023 annual reports. Rural commercial banks are in the exploratory stage and have less information to disclose.
The six major banks have steadily increased their investment in "large models", and they have a wide range of trial business scenarios
Titanium media APP sorted out the main descriptions of the large model in the annual reports of the six major banks, and it can be seen that the Industrial and Commercial Bank of China, Postal Savings and CCB have more descriptions of the large model, compared with the Bank of China, the Bank of Communications and the Agricultural Bank of China, and the six banks except the Agricultural Bank of China have only disclosed information about the "large model" in the 2023 annual report. This move by large banks shows that the banking industry attaches great importance to the exploration and application of large model technology.
From the information sorted out, it can be seen that the six major banks are currently focusing on the exploration of large models in intelligent services, including digital humans, intelligent customer service and platforms, office assistants, etc., and the applications mainly include financial information, content and picture text generation, intelligent customer service online interaction, etc., according to the data, the landing application of the banking and financial model has increased the number of cases handled and services received.
Among the six major banks, it is worth paying attention to the Bank of China, the Industrial and Commercial Bank of China and the China Construction Bank, don't look at the Bank of China's description of the large model only one sentence, but it reveals that the scope and scenarios of digital construction are wider than other banks. In just three lines of text description, it is mentioned that "privacy computing" is a patented technology called "a bank customer information protection method, system, equipment and storage medium" of Bank of China, which is mainly to protect the privacy information of customer information of overseas branches, and establish a personal privacy data scope for overseas bank customers. Intelligent search technology intelligently searches international and industry regulatory requirements, national standards and norms, national laws and best practices, etc., to identify the information supervision needs of overseas banks, establish a customer personal privacy data information database of overseas branches, and further sort out data assets on this basis, and establish a formal data classification and hierarchical management mechanism.
The second is "computer vision technology", which aims to complete the processing, understanding and decision-making of images, videos and other data by simulating human visual perception functions, and finally realize the autonomous response to complex and changeable environments, such as when used in the operation of outlets, it conducts visual analysis and behavior analysis of video and image data collected by outlets, and "notifies" the person in charge of violations of laws or potential risks for timely improvement and processing.
The third is "assisted coding", which uses artificial intelligence technology, machine learning algorithms, natural language processing and other technologies to help developers write and edit code, which can provide developers with intelligent suggestions, automated help, optimized coding and other functions, and improve the overall productivity of R&D personnel.
ICBC integrates financial technology capabilities to create one-stop, integrated and digital treasury service solutions for enterprises, improves treasury information system to help enterprises improve their capital management level and prevent capital management risks, and enriches treasury management cloud services to create a differentiated and matrix treasury management service system. In addition, it adheres to the principle of "technology-driven and value creation", and accelerates the construction of science and technology enforcement and digital ICBC. Consolidate the foundation of safe production and operation, enhance the ability to drive innovative technologies, deepen the D-ICBC digital ecosystem, deepen the reform of the scientific and technological system and mechanism, and create a new engine that meets the requirements of high-quality development.
It is worth noting that many banks are paying special attention to the construction of the treasury, according to the titanium media APP, the treasury information system launched by the Industrial and Commercial Bank of China can establish a treasury management ecology with treasury management as the core, and tailor-made accounts, funds, bills, financing, investment, risk control, and decision-making integrated solutions for enterprises.
CCB took the lead in laying out computing infrastructure, and the bank's "CCB Cloud" provided an overall financial cloud solution of "financial technology base + financial business operating system + high-quality application", with a computing power scale of 463.34PFlops at the end of 2023, of which new computing power such as graphics processing units (GPUs) accounted for more than 20%, and the overall computing power scale and service capabilities remained leading in the industry, providing an underlying platform for building a large model platform. In addition, the bank continued to improve the layout of "multi-region, multi-location, multi-stack and multi-core", integrating multiple computing power such as general and intelligent, and has the ability to carry ultra-large-scale financial core, key business and intelligent services.
In addition, the "Ark Project" implemented by CCB is to promote the construction of the financial model, and in terms of basic capacity building, the bank's financial model has seven first-level capabilities, including information summary, information inference, information expansion, text conversion, security and values, complex reasoning, and financial knowledge.
From the perspective of investment in science and technology and talents, the investment of the six major banks in financial technology has increased year-on-year, of which the investment of four banks is more than 20 billion yuan, the investment of ICBC is the highest, the investment of Bank of Communications accounts for the largest proportion of operating income, and the investment of China Construction Bank has increased the most year-on-year; 。
Large joint-stock banks are suspected to have marginal decreasing benefits in their investment in "large models", and the whole is in the stage of deploying open source large models
Titanium Media APP has sorted out the main descriptions of large models in the 2023 annual reports of large joint-stock banks, and it is intuitive to see that large joint-stock banks have significantly more descriptions of large models than the six major banks. Compared with the description of the large model of the six major banks, the large joint-stock banks describe the application scenarios, the magnitude of open source and the system in more detail, and the more noteworthy joint-stock banks are Ping An Bank, China Merchants Bank and Industrial Bank.
Ping An Bank said that it will use the large model technology in the auto finance AI car inspection scenario, which is a unique application scenario compared to the information disclosed by other banks, and it is not surprising that Ping An Bank has this business and other banks do not, and the subsidiaries of Ping An Group have a wide range of businesses, especially those with two types of special businesses, insurance and lending, so customers have a wider demand for automobile-related services. According to the titanium media APP, the AI car inspection service mainly uses face recognition, OCR recognition and other technologies to collect document information or VIN code, and generate vehicle information in combination with data verification, which will automatically approve the process, so as to facilitate the follow-up of car loan or insurance and other services, save the manual cross-verification of data, images and other information, and improve business efficiency.
Image source@Ping An Bank APP
China Merchants Bank said in its 2023 annual report that it should seize the deterministic opportunity of large language models and actively carry out the construction and application of large language models. On the one hand, it actively strengthens the construction of large models, introduces 100 billion parameter large models, optimizes, trains, and adapts to industry scenarios with its own corpus, actively follows up the development of open source large model technology, and develops 10 billion parameter large models in professional scenarios. On the other hand, we actively explore the application of large models in retail, wholesale, and middle and back offices. At the same time, a large-scale model experience platform is built and a number of domestic mainstream large-scale models are accessed.
According to the titanium media APP, China Merchants Bank is the first bank in China with assets of more than 10 trillion yuan to complete the full cloud migration of the system, and its "system comprehensive cloud project" is not only to physically shut down the traditional centralized architecture, cut all traffic to the cloud, and implement single-track operation, but also to realize the migration of all debit card accounts, credit card customers, corporate accounts and all application systems of the head office and branches to the cloud from the business scope. In addition, the "Cloud Migration Project" includes two modes: one is the "Financial Transaction Cloud" (FTC) as a substitute host, which mainly carries the core accounts of CMB's debit and credit cards, as well as steady-state services such as retail transfers, quick payments, and digital RMB, and the other is the "Native Cloud" (ACS) system that carries systems on traditional open architectures other than the host system, which mainly carries sensitive services such as channels, retail credit and other business processing, and customer operation.
IB has built a "1+5+N" digital IB system, 1 is IB mobile banking, 5 is the five major platforms, IB Puhui, IB Butler, IB Life, Money Manager, and Bank-Bank Platform, and N is the N scenario ecology.
In 2023, in the construction and application of the large model, IB's "Satellite Remote Sensing Enabling Financial Services" project won the second place in the Asia-Pacific region. The service is mainly used in inclusive financial scenarios, and the satellite remote sensing application system explores the effective identification of crop categories, planting areas, and growth conditions by introducing "new data" from satellite remote sensing, combined with data from the government and industry, establishes crop planting portraits, and gradually improves the crop asset assessment and risk early warning mechanism, providing scientific and technological support for the accurate credit of the planting industry. In the green finance scenario, it enables remote monitoring of the construction progress of forestry assets, forestry carbon sequestration and new energy projects, and post-loan risk analysis, so as to realize the monitoring of the whole process of credit, reduce costs and increase efficiency.
From the perspective of technology investment and talents, the proportion of large joint-stock banks in financial technology investment remained stable between 3.5% and 6%, which was higher than that of the six major banks. Ping An Bank did not disclose the number of fintech employees in 2022, and IB had the highest proportion of fintech employees in the total number of employees, mainly due to the "Ten Thousand Talents in Science and Technology Talents Program" that IB has been vigorously implementing.
Several information can be deduced from the above data: first, on the whole, the proportion of joint-stock banks' investment in financial technology and the proportion of financial technology employees are higher than those of the six major banks, which shows that the transformation constraints of joint-stock banks are lower, but due to the higher absolute amount of investment by large banks, the overall digital transformation process of the six major banks is faster at this stage; In the period of "reducing costs and increasing efficiency", there has been a phenomenon of reducing costs in various departments; thirdly, the digitalization process has achieved initial results, and there are now generative codes instead of simple programming work, so the demand for programming employees is more sophisticated; fourth, due to the demand for fintech talents in the banking industry and the high difficulty in cultivating middle and senior talents, there is a phenomenon of job-hopping and poaching between industries, and one trades off the other, so the growth of fintech personnel in some banks has reduced the number of fintech personnel in some banks.
It is worth noting that Bank of Ningbo, as a leading fintech company, and other top-ranked city commercial banks, Bank of Zhengzhou and Bank of Qingdao, have not mentioned the construction of "large models" in their 2023 annual reports. According to industry insiders, the technical threshold of the large model is the main reason for its widespread use in the banking industry, and the cost of capital, personnel, and time has also led to the inability to popularize the technology in all banks. Because of this, at present, only large and medium-sized banks and other leading financial institutions in China are exploring the research and development and application of financial models, employees from different levels of banks said that the rise of the technology may aggravate the survival crisis of small and medium-sized banks, it will continue to widen the gap between banks, in addition, with the maturity of large model technology, the original small and medium-sized banks regionalized business may also be occupied by large models. (This article was first published in Titanium Media APP, author|Li Jingying, editor|Liu Yangxue)
For more macro research dry goods, please pay attention to the official account of Titanium Media International Think Tank: