On September 26, the 2023 China International Forum on Financial Inclusion and the Academic Summit of Digital Economy Open Research Platform was held in Beijing, and Zhang Xiaoyan, Deputy Dean of PBC School of Finance and Chair Professor of Finance of Tsinghua University, was invited to participate and delivered a keynote speech on "Application and Challenges of Artificial Intelligence in the Financial Field: Taking Big Language Model as an Example". In the speech, Zhang Xiaoyan believed that big language models have shown surprising technological breakthroughs in the field of natural language processing, and the outstanding capabilities of big language models in data analysis and decision support will promote the improvement of global production efficiency. She also summarized the application and challenges of big language models in the financial field based on case studies.
The picture shows Zhang Xiaoyan giving a keynote speech
A large language model is a large-scale artificial intelligence model used to process natural language information. Large language models are trained on large-scale text data to learn grammatical, semantic, and contextual information in the language, so that they can understand and generate human language and perform a variety of tasks, and the main application scenarios include knowledge question answering, sentiment analysis, article generation, knowledge extraction, abstract summary, etc. The international Open AI's GPT3.5 model has 175 billion parameters, while GPT4 is expected to reach 1 trillion; the closed source version of InternLM of the domestic Shanghai AI laboratory has 104 billion parameters, which is trained on a corpus containing 1.8 trillion tokens. The large parameter characteristics of large language models have brought breakthroughs in various abilities, and Zhong et al. (2023) shows that the ability of large language models in many exams has approached and exceeded the human level. These include the Academic Aptitude Test (SAT) for U.S. High School Graduates, the GRE, China College Entrance Exam, the Bar Exam, and the National Civil Service Exam.
The picture shows the forum site
Zhang Xiaoyan said that big language model technology has brought about the improvement of global economic production efficiency. First, a questionnaire released by McKinsey in 2023 shows that the big language model has been used in many industries to improve industry efficiency. Among all industries, respondents in financial services and technology used generative AI the most; Among different departments of the enterprise, respondents in marketing, customer service management, R&D, and software development use generative AI the most. Second, the large language model will promote economic growth, and its wide application in various industries will bring about the improvement of production efficiency in the whole society. Third, the big language model is driving job market change, which can replace some simple text processing jobs while also providing entirely new jobs. The state maintains a positive and encouraging attitude towards artificial intelligence technologies such as large language models, and the industry is in a new situation of orderly market development and continuous improvement of supervision.
Zhang Xiaoyan believes that the potential impact of big language models on the financial industry may be even greater. The big language model can be used in investment scenarios for investment decision-making, risk assessment, market analysis, document processing, automated customer service, and public opinion analysis. For example, in macroeconomic analysis, the attitude of the US central bank towards monetary policy can be analyzed based on a large-language model. In the banking wealth management business, the application scenarios of generative AI run through all links of the bank's entire industry chain. Compared with other industries, the financial industry needs to process a variety of information, including news, analyst research reports, government policies, etc., and large-language models can extract and analyze massive amounts of text more conveniently. Second, the timeliness of financial information requires analysts to make quick decisions, and large language models can analyze the market and make recommendations in seconds. In addition, a large number of financial services require language communication, and in the future, large language models may be able to answer customer inquiries, provide investment advice, and even partially replace human customer service and investment advisors.
With the rapid development, the application of large language models in the financial field also faces some technical challenges. First, the lack of knowledge in the field of financial expertise (that is, the corpus of large language models in the financial field is insufficient); The second is the cost of training large language models, and the cost of computing power required to train large language models is high; Third, financial information has timeliness requirements, and the training corpus of large language models may have lag problems; Fourth, the accuracy of financial decision-making requirements; Fifth, the high dynamics of the financial field, such as the dynamic increase of professional vocabulary in the financial field, the same term may change in meaning at different times.
There are solutions to the current challenges. The first is to retrain the large language model in the financial field based on the financial corpus; Second, based on an open source big language model, through different fine-tuning techniques, the ability of the big language model is aligned with human needs, and then a financial big language model that meets specific scenarios is created; The third is to combine the word vector database and Langchain technology to update the word vector database in time to allow the large language model to have the latest and accurate information.
Finally, Zhang Xiaoyan summarized and looked forward, she believes that although the big language model has appeared for a short time, it is having a positive impact on the global and Chinese economy, and many business scenarios in the financial industry are very suitable for the application and landing of the big language model. In order to standardize the healthy and orderly development of the big language model in the financial field, relevant regulatory authorities need to formulate rules and regulations in a timely manner to guide the healthy development of the industry.