The recent entry of the human-computer interaction model ChatGPT into the application field means that the development of artificial intelligence has reached a new highland. So, what is the application and future development trend of artificial intelligence in the financial field? What challenges will this pose to financial institutions? A few days ago, the Financial Times reporter interviewed Zhang Chenghui, president of the Sanya Economic Research Institute, on the above issues.
Zhang Chenghui, former director and researcher of the Institute of Financial Research of the Development Research Center of the State Council, enjoys the special government allowance of the State Council. Standing Director of China Modern Finance Society, Doctoral Supervisor of the Graduate School of the Chinese Academy of Social Sciences, Special Invited Advisor of the 7th China Council for International Cooperation on Environment and Development, Member of the Strategic Advisory Committee of PBC School of Finance, Tsinghua University. Key research areas: financial reform, SME financing, financial technology, over-the-counter capital market, ESG. He has participated in more than 30 major key projects of the Development Research Center of the State Council, presided over dozens of research projects, published more than 300 papers, and won the China Development Research Award more than 10 times.
Application scenarios and future development trends Ask questions
Reporter: Recently, ChatGPT has attracted attention from all parties, it has the strongest machine brain and knowledge base in history, which can not only help people better understand the world, but also break language and cultural barriers, promote cross-border communication and cooperation among human beings, and even change the way human thinking and cognition to a certain extent. How do you see the application and future development trend of artificial intelligence in the financial field?
Zhang Chenghui: In recent years, financial technology has been widely used and developed rapidly, which has profoundly changed the format of finance, the internal logic of finance and the behavior mode of the financial workplace. Big data, blockchain, artificial intelligence, Internet of Things, etc. are originally the main components of financial technology. With the improvement of artificial intelligence technology, financial technology will have a more extensive and in-depth impact on the financial industry.
First, customer service and digital marketing of financial products. Customer service is the application scenario where artificial intelligence can play a role and effect the most quickly. So far, various trading platforms have widely adopted robot customer service. But overall, the effect is not ideal. The reason for this is that the robot customer service lacks the ability to sense the language expression of customers, lacks empathy for customer needs, and does not have a wide range of knowledge. From the performance of ChatGPT, high-level artificial intelligence may be more experienced, knowledgeable and responsive than human customer service. It can be expected that in the near future, manual customer service will be completely replaced by intelligent robots, the number of customer service positions will be sharply reduced, and the huge customer service workplace will no longer exist, thereby greatly reducing the labor costs and management costs of financial institutions.
With the development of financial technology, digital marketing has played a great role in tapping a large number of "long tail" and "sleeping" customers of financial institutions. The artificial intelligence deepens the understanding and dialogue functions in the marketing process, improves the recognition accuracy, and can conduct high-quality one-on-one communication with the respondents, effectively solving the problems of high cost of manual follow-up, difficult management of human seats, and difficult real-time monitoring of data. In recent years, the highly demanded personalized service has put forward extremely high requirements for the professional ability of marketers, the accuracy of demand identification and the flexibility of response, and the wide application of artificial intelligence will help to quickly improve the product marketing capabilities of financial institutions.
Second, financial risk management. Risk prevention is the duty and core responsibility of financial institutions. As intermediaries of funds, financial institutions face a variety of risks, such as credit risk, market risk, management risk, liquidity risk, legal and compliance risk, etc. In the face of risks, financial institutions must first establish a sound internal control and risk management system, establish a sound risk management framework, and classify, assess and manage various risks. In this process, artificial intelligence can fully play a role, including monitoring the implementation of the system, rapid response to volatile markets, and scientific assessment of the type and degree of risk. Secondly, employees' understanding of risks and ability to implement systems are the key to risk management of financial institutions, and risk management personnel should especially have rich risk management experience and theoretical knowledge. The introduction of artificial intelligence in the process of providing systematic training and education for employees by financial institutions can effectively improve training efficiency and accurately detect the risk management capabilities and levels of personnel in key positions. Third, the disclosure and disclosure of risk information is the responsibility and obligation of financial institutions to the public. Information disclosure involves a large amount of data and information, and it is difficult to process this information scientifically, accurately and quickly by relying on human resources alone, and artificial intelligence can also greatly improve efficiency in this field.
Third, product pricing. The essence of financial product pricing is risk assessment, which requires risk assessment of customers according to factors such as their credit status, repayment ability, financial status, etc., and formulates different risk premiums or discounts to avoid losses that customers may suffer from default. Due to the diversity and complexity of financial products, a lot of knowledge and skills in mathematics, statistics, and economics are required. Taking actuarial as an example, a reasonable actuarial can not only protect the interests of insurance companies themselves, but also help protect the rights and interests of customers. Actuarial factors include, at a minimum, the risk of the insurance product (type of underwriting risk, risk level, insurance liability, insurance period, insurance amount, deductible, etc.), historical data and statistical analysis (average life expectancy, accident probability, weather changes, etc.), insurance product risk (predicted loss, probability distribution, time value, etc.), policies, regulations and regulatory requirements, economic environment (inflation, interest rates, etc.) and market competition, insurance company's underwriting capacity (balance sheet status and matching, investment portfolio, profit budget). , reserves, solvency), etc.
The risk model of a financial institution is a complex system that requires a combination of knowledge and skills in risk assessment, data collection, mathematical modeling, model validation and risk management. In fact, in the process of establishing and applying and testing risk models, financial institutions have applied a lot of financial technology, and the addition of high-level artificial intelligence will further improve the scientificity of these models, and it is possible for artificial intelligence to replace actuaries to a certain extent.
Fourth, insurance investigation and claim settlement. The biggest risk that insurance companies face after the insurance policy is issued is fraud. Billions of dollars in fraudulent claims occur every year. In order to reduce such risks, insurance companies must conduct necessary investigations and reviews on claims applications, carefully identify the authenticity and extent of losses, and provide a basis for claims decisions. In addition, due to the large number of insurance policies, involving multiple fields, and the high degree of complexity, the investigation and settlement of claims is often time-consuming and laborious. AI can greatly simplify this process, eliminate human error, and improve the science and speed of claims processing.
Fifth, investment advisors. Fintech has begun to be widely used in the field of securities investment, including quantitative investment, providing customers with personalized investment advice and recommendations, and optimizing customers' portfolios under the premise of ensuring risk control and maximizing returns. However, in the field of PE and VC investment, artificial intelligence mainly appears as an investment object rather than as an investment decision-making tool. In the future investment advisory scenario, artificial intelligence should be able to use its powerful database, knowledge base and analysis capabilities to help the private equity investment industry make more scientific investment decisions and improve the return and risk control capabilities of the portfolio.
Sixth, family asset management. Compared with the investment advisory industry, which mainly serves institutional and high-net-worth clients, family asset management is still largely blank in the mainland. There are reasons why Chinese families have traditionally found it difficult to accept fee-based services, as well as barriers between different industries, and the lack of financial institutions in the ability to provide families with life-cycle customized services. For example, in addition to the traditional deposit and loan foreign exchange business, bank relationship managers can only recommend a small number of products such as funds and wealth management, and customers must have direct contact with financial institutions that provide corresponding products if they want to buy securities and insurance. The data processing capabilities of artificial intelligence will help financial institutions and third-party service providers to develop the huge market of home asset management, thereby further improving the efficiency of financial services.
Based on the huge power of artificial intelligence in improving response speed and work efficiency, financial institutions will use artificial intelligence more in the future, and thus give birth to more financial service scenarios and new profit models. It can be expected that the development of financial technology will continue to develop in the direction of digitalization, intelligence, personalization and cross-border, so as to further deepen the differentiation of financial services and make different types of financial services more integrated and innovative.
Data security is a key factor in transformation
Reporter: Data security is a risk factor that must be paid attention to in the era of artificial intelligence, and it is also a key factor affecting the digital transformation of financial institutions. How should data be secured?
Zhang Chenghui: At present, data security in a broad sense involves customer information and personal privacy security, data security of the whole business cycle and various business processes of financial institutions, security of information systems and infrastructure of financial institutions, and data security in the operation of online and offline business scenarios of financial institutions. The most sensitive thing for society and the public is the risk of "batch leakage" after personal data is collected and used centrally.
The paradox is that maximizing the value of data must rely on the aggregation, flow, processing and analysis of a large number of diverse data, and in this process, it is inevitable to encounter problems such as hacker attacks, user mismanagement, and malicious use of data. Verizon's 2021 Data Breach Investigation Report states that 85% of data breaches involve human factors, and human negligence is the biggest threat to data security. In order to properly handle data flow and security issues, the EU has successively issued a number of data protection-related regulations since 1995, clarifying certain principles in the process of data acquisition, and greatly improving the level of data protection by imposing high fines, establishing government supervision agencies, and requiring enterprises to add new data protection officers. In May 2022, the Data Governance Bill was approved into law by the Council of the European Union after being voted by the European Parliament, further enriching and refining the connotation of data governance.
The Measures for the Prevention and Control of Computer Viruses promulgated by the mainland in April 2000 for the first time regulated the management of threats caused by personal data from the aspect of computer viruses, and a number of relevant laws and regulations have been promulgated since then. In 2016, the Cybersecurity Law of the People's Republic of China was promulgated. The Law establishes the principles of personal data processing, establishes a domestic storage and export security assessment system for personal information and important data. The Data Security Law of the People's Republic of China, passed by the National People's Congress in June 2021, pays more attention to the security protection of data itself, proposes to formulate a catalogue of important data, promote hierarchical classification of data, and regularly carry out data risk assessment. The Personal Information Protection Law of the People's Republic of China, which came into effect in November 2021, establishes rules for the processing and cross-border provision of personal information and establishes a complete framework for personal information protection.
Although a legal framework has been established to protect data security, data security will remain a major challenge for society as a whole. At present, telecommunications network fraud has become a prominent problem affecting public order, and in some provinces and cities, the number of fraud cases has accounted for half of all criminal cases, and it is very difficult to solve cases and recover stolen funds. For financial institutions, relevant laws and systems have increased the legal responsibility for the development of data resources, and it is urgent to accelerate data security capacity building from both business and technical levels, establish a data security system, improve data risk verification capabilities, risk early warning capabilities and data protection capabilities, and improve data management organizational systems and organizational structures. And these measures will increase the cost to varying degrees.
Challenges for financial institutions in developing AI
Reporter: The process of deep integration of artificial intelligence technology and economy and society carries great opportunities, but also faces a series of potential challenges. What challenges will the application of AI technology bring to financial institutions?
Zhang Chenghui: Although artificial intelligence will occupy an increasingly important position in financial technology, financial institutions will definitely face many challenges in the process of developing artificial intelligence. Specifically, there are mainly the following aspects:
The first is how to effectively improve computing power. Machine deep learning underpinning artificial intelligence involves massive amounts of data and complex algorithms, requiring supercomputing power and large amounts of energy. For small and medium-sized financial institutions, there are inherent deficiencies such as weak financial strength and lack of talents, and it is difficult to bear the huge investment required to improve computing power. One solution is to put artificial intelligence algorithms on the cloud, but many small and medium-sized financial institutions still have great concerns about going to the cloud due to the privacy and confidentiality of data.
For example, in 2022, the six major state-owned banks will invest more than 10 billion yuan in fintech, and the proportion of fintech investment of many joint-stock banks has increased, and the growth rate of fintech investment of some banks has increased by more than 20% year-on-year. Nevertheless, in terms of improving computing power, large financial institutions still face bottleneck constraints such as basic chips, operating systems, and databases. The challenges in expanding application scenarios and data architecture management capabilities across time and space and multi-technology integration are also greater. In addition, due to the huge energy consumption of computing power infrastructure, how to use better energy-saving technologies and introduce more green and clean energy in the process of green and low-carbon writing is also a problem that large financial institutions need to face.
The second is how to deal with the bias of AI development agencies. In reality, the bias of artificial intelligence technology and the discrimination caused by it are receiving more and more attention. This bias is caused by program development. Due to the differences in nationality, religious belief, race, gender, and region of residence held by developers will lead to cognitive bias to a certain extent, coupled with differences in the data source and frequency of the training algorithm behind AI, the final model and system must be double-standard. For example, it is predicted that Asians have more plagiarism than Europeans, and blacks and low-income families have a much higher risk of crime than whites and wealthy families. Unfortunately, at present, the technical level and independent controllability of the overall artificial intelligence computing power basic software and hardware in the mainland still have a certain gap with foreign countries, and the deep learning framework has no competitive advantage. In terms of intelligent computing chips, mainland enterprises are still largely subject to advanced chip manufacturing processes including chip design, chip production, packaging production, cost testing, etc., and there is also a significant gap with foreign advanced products in the software ecological environment. In this case, it is often difficult to avoid the occurrence of AI bias.
Third, the development of artificial intelligence is constrained by the degree of industrial digitalization. As an infrastructure, computing power needs relevant data to support it. In recent years, China's financial technology has achieved "overtaking on a curve", its development level ranks among the forefront of the world, and the degree of government affairs digitalization has also been rapidly improved. In contrast, industrial digitalization has been relatively slow. A series of basic industries such as intelligent perception, automatic control, intelligent equipment, network connection, and industrial software necessary for the construction of industrial Internet platforms are highly dependent on foreign countries, and the power and ability of traditional industrial digital transformation are lacking. In addition, data barriers between government departments and industries still exist, which restricts the development and application of artificial intelligence in the financial field to a considerable extent.
Fourth, regulatory pressure. The subversive development of artificial intelligence represented by ChatGPT has challenged human legal systems, standards and rules, rights protection and judicial remedies. While technology is changing life, it is also changing the traditional laws of society, and the potential impact and risks cannot be ignored. In particular, infrastructure, new technology applications, data and operation service providers, and supply chains are mixed and intersected, and the widespread use of remote operation and control technology brings extremely complex regulatory difficulties.
In short, the era of artificial intelligence has arrived, and artificial intelligence will be used more widely and deeply in the financial field in the future. For financial institutions, it is better to actively embrace the intelligent era than to passively accept. However, as far as single market entities are concerned, it is still necessary to do what they can. At the same time, in the face of the problems and challenges in the development of artificial intelligence, it is necessary for the government and industries and enterprises to cooperate closely to strengthen communication, accelerate the completion of shortcomings, and jointly promote the healthy development of artificial intelligence under the premise of ensuring safety.
If you still have questions, please leave your questions and we will continue to answer them
I have a question
Xiaobian online wait oh~
Source: Financial Times Reporter: Liu Li Editor: Yang Jingyi Email: [email protected]