AI agents are involved in the financial industry, and a large number of applications have been implemented, but commercialization will take time
Author：Eve Cole
Update Time：2024-11-22 14:18:01
AI Agent is a high-frequency word in the field of artificial intelligence in 2024, and the financial industry is also involved. At the 2024 Inclusion Bund Conference, AI agents were frequently mentioned, and the industry believes that AI agents will become one of the mainstream directions for the development of artificial intelligence. China Business News reporters also noticed that many financial intelligence applications are being implemented, bringing a more humane experience to the intelligent communication between humans and machines. However, in applications in the financial field, intelligent agents are prone to "hallucinations". Commercialization will take time to iterate, and the relevant regulatory framework also needs to be improved. AI agents involved in the financial industry refer to artificial intelligence agents, which are intelligent entities that can perceive the environment, make decisions and execute actions. They have the ability to gradually complete given goals through independent thinking and calling tools. From the user's perspective, intelligent agents have the ability to think deeply about specific problems and specific fields, and can communicate with people more like "humans". According to Han Xinyi, President of Ant Group, AI agents, as a product form of generative artificial intelligence, are the core direction of the current application of large models, allowing large models to grow "hands and feet." Sun Maosong, a foreign academician of the European Academy of Sciences and executive vice president of the Artificial Intelligence Research Institute of Tsinghua University, pointed out in an interview with China Business News that in English, "intelligent" is translated as "Agent", and the emergence of AI has given this term a new meaning. meaning. The concept is relatively broad and may refer to an intelligent robot or a digital human in the virtual world. It may not even have the concept of "human", but refers to a software or a tool that uses AI technology to help people solve problems. AI agents that can make large models have more human characteristics are the core direction of current large model applications. This trend has already spread to banking, insurance, financial management and other financial fields. In 2024, many large model manufacturers will explore the application development of AI agents. In terms of wealth management, risk assessment, customer service, etc. in the financial field, agents are showing their commercial value. For example, the "Postal Savings Brain" program launched by the Postal Savings Bank has transformed from perception and insight to generative creation, and is promoting scenarios such as intelligent business assistants, digital account managers, and virtual business halls; China Construction Bank's "Ark Plan" is gradually promoting financial The construction of large-scale models in the field promotes the implementation of intelligent agents in scenarios such as marketing, investment research reports, and risk control. AI agents also provide digital intelligence services to professionals and lower-threshold wealth consulting services to the public. For example, at this year's exhibition, Zhixiaozhu 2.0 is an intelligent assistant developed by Ant Group based on its self-developed large model and customizable agentUnierse agent framework with professional knowledge. At the scene, an audience asked it, "Please provide an investment and financial management plan suitable for the elderly." It gave corresponding investment suggestions, analyzed the target situation based on the current market conditions, and made conclusions from two aspects: risk and return. . "The application of intelligent agent technology is entering public life without showing off anything." Sun Maosong pointed out that we already have the ability to make better agents, mainly due to the ability to generate large language models, The "four major advances" in the ability to generate code, image and video processing capabilities, and 3D modeling capabilities have given new development opportunities to intelligent agents. However, Sun Maosong also believes that compared with large models that are limited to dialogue environments under general conditions, intelligence is a more complex upstream concept that is discussed in three-dimensional and four-dimensional spaces. However, from the perspective of industry and industrial implementation, Agents are downstream concepts of large models applied in various practical fields. Difficulties and Challenges The application of AI agents in the financial field is the only way to go, but this process is not smooth sailing, and there are still many problems and challenges to be solved. First, in the financial field, there are challenges in the commercialization of AI agents. "The commercialization of intelligent agent technology still needs time to iterate." Sun Maosong believes that this is a process in which quantitative changes lead to qualitative changes. It may take several years of iteration and accumulation before its performance will be significantly improved, and the commercial value of intelligent agents will be worth it. Looking forward to it, but having more patience with it. The road to commercialization of AI agents is also restricted by R&D capabilities. According to Huang Xuanjing, a professor at Fudan University, intelligent agents are considered a promising path toward general artificial intelligence. It can give full play to the professionalism of large language models and bring about iterative upgrades of various services. But at the same time, the research and development of intelligent agents currently faces four major challenges: insufficient base model capabilities, lack of a unified interactive framework, lack of self-learning and self-evolution, and safety and ethical issues with intelligent agents. In addition, the problem of large model illusion is a major challenge for the application of AI agents in the financial industry. Industry insiders believe that the financial industry's fault tolerance rate is very low. Whether it is knowledge question and answer or content extraction, it puts forward very high requirements for intelligent agents. The accuracy of the model has become an important obstacle to the adoption of generative AI in the financial industry. "Large models are very popular and their abilities are also very powerful, but there are conditions for this ability to be strong. For example, it performs very well in a dialogue environment, but once you get out of that environment, your ability is questionable." Sun Maosong pointed out that through intelligent agents and other , let the capabilities of large models be further improved, growing from "co-pilot" to "main driver". "Although large models have shown great potential in the financial business field, their comprehensive application in the financial industry still faces many challenges." According to Wu Lianfeng, vice president and chief analyst of IDC China, the current combination of general large models and the industry as a whole It is still in its infancy. In addition, the financial industry’s strict supervision and strong compliance requirements also make the implementation of large models need to be more cautious and the cycle will be longer. Overall, the industry ecosystem of large models is not yet fully mature, and widespread application requires overcoming challenges such as technology, industry evolution, regulation, and technology ethics.
