3D intelligent interactive digital employees take up their posts,AI pre-loan investigation reports are quickly generated,and bond intelligent assistants enhance operational efficiency...

This year,the financial industry has ushered in a surge in the application of large models,with many banks increasing their investment in AI technology,continuously exploring innovations,and striving to improve the level of intelligent service systems.

Industry insiders and experts have stated that currently,personalized customer experience has become the key to bank competition.

The application of AI technology enables banks to more conveniently provide customized financial products and services,which helps to drive performance growth.

However,it cannot be ignored that AI is a "double-edged sword"; the broader its application in the financial field,the greater the pressure on banks to ensure data security and information protection,which poses higher demands on the technical level and management capabilities of banks.

State-owned large banks have entered a period of explosive development,promoting comprehensive industry upgrades.

In recent years,with the rapid development of artificial intelligence technology,state-owned large banks have focused their attention on the development of AI large models.

According to the Global Financial Large Model Patent Innovation Ranking released by MIT Technology Review in 2024,12 Chinese institutions have entered the top 20 globally.

Among them,Ant Group,Ping An Group,and ICBC rank in the top three,with CCB,BOC,Consumer Finance,WeBank,and ABC also among the top 10.

Reporters have combed through annual reports and public information to find that nearly 20 banks have disclosed their exploration in the AI field this year,with state-owned large banks being the main force.

Among them,ICBC and Postal Savings Bank of China (PSBC) place relatively higher importance on large models,with the number of mentions and the length of the discussions far exceeding those of other large banks.

ICBC mentioned in its semi-annual report that in the first half of this year,the bank deepened the construction and empowerment of large model technology,becoming the first in the financial industry to complete the full-stack independent controllable training and inference deployment of enterprise-level financial large models; it promoted the deep integration of large model technology and business,achieving innovative applications in various fields,including empowering the entire process of investment,financing,and trading in the financial field,and building a marketing intelligent assistant based on large models.

The term "large model" appeared 15 times in PSBC's semi-annual report this year,the most among all listed banks.

PSBC disclosed that it has completed the pilot construction of large model computing power cloud resource pools,achieved cross-cloud application of distributed object storage,added more than 500,000 vCPU resources in the first half of this year,effectively supporting the smooth migration of system resources.

Currently,the bank's large model heterogeneous computing power cluster has initially possessed the ability to support the training of large models at a scale of hundreds of billions.

CCB has continued to work hard in creating financial image and text recognition products,with a system that can support the recognition of more than 140 types of bills,covering 75% of bill recognition volume,and increasing the efficiency of bill review information entry by 120 times,winning the championship in the seal text detection track of the global artificial intelligence document image analysis and recognition competition (ICDAR 2023).

BOC disclosed that in the first half of this year,the bank promoted the application of privacy computing,the Internet of Things,and other technologies,focusing on the three key elements of computing power,algorithms,and data,piloting large model applications such as code assistance,and conducting pre-research on anti-quantum cryptography and quantum computing technology.

Zeng Gang,Director and Chief Expert of the Shanghai Financial Lab,stated that banks of different types should choose a digitalization path suitable for themselves,with state-owned large banks paying more attention to security and technical requirements.

It is expected that a considerable number of financial positions may be replaced by artificial intelligence in the future,and banks should quickly build new core competencies to achieve high-quality development.

Small and medium-sized banks are "running fast in small steps" to seek new opportunities for AI empowerment.

Reporters have found that the research and innovation of artificial intelligence in finance have characteristics such as high investment costs,relatively slow results,increasing marginal returns,and scale effects.

Compared with large and medium-sized banks,city commercial banks,rural commercial banks,and other small and medium-sized banks,although they cannot be compared in terms of cost investment,are not waiting to die.

They are doing their best to catch up with the digital wave under limited resources.

According to the combing of this year's mid-year reports,Bank of Beijing is undoubtedly one of the most active small and medium-sized banks.

This year,Bank of Beijing released the AIB artificial intelligence innovation platform,integrating up to 80 large model services and 7 GPT creation tools,building a digital system covering voice interaction,image recognition,natural language processing,intelligent decision-making,knowledge graph,robot automation,virtual digital people,large models,and other fields.

It is worth mentioning that Bank of Beijing has upgraded and iterated the "smoke index" risk early warning model,accelerating the construction of a new generation of credit risk management systems,and has built the Jingxin Miaobi intelligent report platform with the generative writing ability of "big data + GPT large model".

Nanjing Bank is also very active.

According to the disclosure,in the first half of this year,the bank launched a human-computer collaborative intelligent call,which effectively improved the efficiency of reaching target customers,successfully landing more than 9 major categories and more than 40 service scenarios such as marketing assistants and office assistants,and created the "Xiaohe Chat" business Q&A assistant,providing various large model services more than 300,000 times.

Some small and medium-sized banks also focus on continuously deepening the application scenarios of new technologies such as large models.

For example,Shanghai Bank has created a "digital intelligence + humanization" online service system for corporate customers,exploring the use of large models to improve precise Q&A capabilities,and establishing a corporate expert service team with front,middle,and back office collaboration; Chongqing Bank has conducted in-depth exploration of multimodal AIGC capabilities,piloting the integration of large model technology into traditional fields such as OCR to improve intelligent recognition levels.

In addition,compared with the technological research and development advantages and talent reserve strength of large and medium-sized banks,small and medium-sized banks are at a disadvantage in technological innovation,data governance,and ecosystem construction.

Therefore,some small and medium-sized banks also choose to rely on external forces to achieve complementary advantages and ensure that they do not fall behind.

Ningbo Bank recently announced a strategic cooperation with Ant Group,and the two parties will jointly explore joint innovation in large models,big data,high-performance computing,blockchain,and other fields,and jointly promote the research and landing of new technology applications.

Jiangnan Rural Commercial Bank announced a cooperation with JD Cloud,introducing AI digital people,and with the help of the "Yanxi" platform,it has achieved 3D digital people customer service,simplifying the user interaction process.

Opportunities and challenges coexist,and AI needs to better empower finance by adhering to "long-termism".

Industry insiders have stated that the banking industry is not smooth on the road to "AI+".

On the one hand,the rapid iteration of technology requires banks to continuously update systems and train talents.

On the other hand,data security and privacy protection issues have also become challenges that cannot be ignored in the process of intelligence.

Ping An Securities research report pointed out that in 2024,against the background of lack of elasticity in corporate credit pricing,the growth of retail loans will become the key to stabilizing the bank's asset side income level.

However,it cannot be denied that risk control is complex,supervision is getting stricter,and there is a shortage of talents...For banks,there are still a series of specific challenges to face in the process of digital transformation.

According to the "2024 Financial Industry Generative AI Application Report",although generative AI is reshaping the global financial industry in an unprecedented way,the high cost of developing and using large models is also the main challenge to its application in the financial industry.

The high cost comes from the difficulty of obtaining high-performance GPUs,high electricity consumption driven by large-scale parameters,a lack of large model talents,and the deployment of localization under data security,which may cause some small and medium-sized banks with tight budgets to be out of reach of generative AI.

Industry insiders believe that as one of the industries with the highest dependence on information technology,banks need to rely on the underlying IT system for daily operations,risk management,and customer service.

However,banks themselves have heavy IT assets,and it is difficult to upgrade large-scale systems.

Therefore,in order to adapt to the development of artificial intelligence,the banking industry needs to introduce advanced technology to reduce costs and increase efficiency,and better support the digital transformation of the upper-level business.

Cao Haifeng,a non-bank financial analyst at UBS Securities,stated that the potential reshaping effect of generative AI on the financial industry is more direct,mainly due to its large amount of data accumulation,high labor intensity,and a high proportion of language-related work content.

However,when applying it to risk prevention and control,on the one hand,it is necessary to ensure the security of internal data,and on the other hand,it should be cautious when using GPT technology in serving customers,such as directly recommending stocks and funds,to avoid touching regulatory red lines.