Insurance catches the big-model wave

Author: Su Jie

**Source:**China Insurer

Image source: Generated by Unbounded AI‌

Nowadays, ChatGPT and the large model behind it are triggering a wave of technology in the industry, and many technology companies have joined it one after another. With the rapid application of ChatGPT in the world, my country has also set off a wave of AI innovation. In the insurance industry, what kind of sparks will the combination of large-scale model technology and it create?

The wave of AI large models is coming

"The wave of technology comes once in 30 years. In the face of the wave of artificial intelligence represented by ChatGPT, if you don't get on this boat, you will miss the opportunity to go ashore." Gu Wei, vice president of Sunshine Insurance Group, pointed out that the era of technology is turbulent, represented by ChatGPT The large-scale model technology has an impact on various industries, and the insurance field is one of the most affected areas.

Yang Tao, deputy director of the National Finance and Development Laboratory and an academic member of the Fintech 50 Forum, wrote that in terms of financial applications, generative AI has gradually demonstrated its value on several levels: First, in terms of improving customer-oriented service capabilities , which can provide support for professional marketing and channel maintenance of employees of financial institutions; second, in terms of improving institutional work processes and efficiency, it can further enhance the intelligence of business chains and automation of office models; third, in terms of text processing, for financial institutions General texts, professional compliance documents, and business-needed information can provide technical support at a lower cost and more efficiently; fourth, it provides IT support for financial institutions, and its code generation function improves the efficiency of basic IT work.

Yang Tao said that the financial industry, as one of the "cornerstones" for the healthy operation of the economy and society, is also a special information processing industry. It should seize opportunities and create conditions to use AI to promote its own high-quality development, making services more efficient, convenient and effective. temperature.

In practice, existing insurance institutions have begun to try to lay out large models. For example, at the recently held World Artificial Intelligence Conference, ZhongAn Insurance released the first AIGC application map in the insurance industry, and brought ZhongAn AIGC Zhongtai Lingxi and the first batch of AIGC application tools for insurance vertical scenarios - Yichuang Content Operation Platform And Jizhi business analysis platform.

Gu Qingshan, deputy general manager of Sunshine Insurance Group Science and Technology Center, explained in detail the Sunshine Insurance Planning and Sunshine Zhengyan large-scale model open platform, and built the "three major robots" (sales robot, management robot) , service robots) to support the four major areas of insurance business, namely intelligent marketing, intelligent operation, intelligent risk control and intelligent management.

Gu Qingshan said that Sunshine Insurance's large-scale model technology has been applied and explored in the "Mengke Online Sales Robot" project. The accuracy of information extraction tasks has increased by 15%, and the accuracy of intention recognition and intelligent question and answer has improved significantly. In the future, Sunshine Insurance will build a proprietary large-scale model in the vertical field of insurance, support sales, underwriting, claims settlement, customer service and other industry applications, and build core technological competitiveness.

"In the next 5-10 years, large-scale model technology and capabilities will have a profound impact on the insurance industry. In the long-term, we hope to continue to explore and develop intelligent applications of large-scale models in the field of full-staff office, 'three robots' and more business products, and truly realize A machine serves the ultimate vision of a group of customers, thus realizing the replacement of humans by robots." Gu Qingshan said.

How to give full play to the advantages of artificial intelligence technologies such as ChatGPT in actual insurance business scenarios to achieve more accurate and efficient marketing communication? Luo Cheng, CEO of Beijing Mecca Intelligent Technology Co., Ltd., expressed his expectation for the emergence of real smart insurance. Luo Cheng said that the general-purpose large model is currently unable to subdivide downstream tasks and training paradigms in all fields. In addition, high-quality data in the industry is still relatively scarce.

Li Jin, an expert in artificial intelligence technology and the founder and CEO of Hangzhou Huali Intelligent Technology Co., Ltd., said that the application of large models in the field of virtual agents can complete the collection of customers' personal and family information, understanding of insurance needs and insurance recommendations. He pointed out that virtual insurance agents can play the role of linking insurance agents and customers, empowering insurance agents on the one hand, and serving and accompanying customers on the other.

Data is the core competitiveness

Cao Feng, Director of the Artificial Intelligence Department of the Cloud Computing and Big Data Research Institute of the China Academy of Information and Communications Technology, said that at present, the large-scale model industry has entered an explosive period, and there are mainly traditional artificial intelligence companies, subdivided industries, and entrepreneurial and innovative companies doing large-scale model research and development. At present, everyone is more concerned about the effect of large models in actual application scenarios, rather than just focusing on the parameters of large models.

Wei Chenyang, deputy dean of Tsinghua University’s Financial Technology Research Institute and director of the China Insurance and Pension Research Center, believes that large models are now in the staking stage. Among the “three elements” of AI, compared with computing power and algorithms, data is the most important. core competitiveness. In his view, the insurance industry is rich in data and has a large number of application scenarios, which is an advantageous area for large-scale model applications. At the same time, he pointed out that the positioning of the large model is not to replace people, but to play the function of "co-pilot".

According to Liu Yi, a senior consultant expert on Microsoft's financial industry strategic technology, Microsoft classifies the application scenarios of large models according to data types, and provides practical experience in the application of large model technology for different scenarios. Now the big model has become a personal assistant, and the role of human beings has changed from executor to commander. For enterprises, how to introduce large-scale models more reasonably requires not only distinguishing scenarios, such as data security and the delay requirements of enterprise-level services, but also classifying application scenarios according to data types, such as Internet data, mixed data from internal and external networks, Intranet data, etc.

"The new technology of AI has brought new landing scenarios, and enterprises urgently need to coordinate efforts from the AI 'troika'—model, data and computing power levels, to drive business landing." Founder and Chairman of Samoyed Cloud Technology Group And CEO Lin Jianming pointed out that enterprises should maintain business acumen, look for opportunities in various industry tracks, thoroughly understand the business needs and challenges of different industries, develop customized digital intelligence solutions, and combine AI technology with data-driven business applications Combined, we will continue to absorb and train excellent AI talents. In addition, user experience and value creation should be given top priority, and continuous improvements should be made based on user feedback and data.

A researcher at the Institute of Computing Technology of the Chinese Academy of Sciences, a doctoral supervisor, and the leader of the natural language processing team pointed out that the training of large models requires a large amount of data resources, computing resources, and storage resources. Can be fully used for large model training.

Building a safe and trusted system is the key

At present, in the excellent performance of the large model, there are also doubts from all walks of life about issues such as credibility, data and privacy security, abuse risk, and ethics. But there is no doubt that the vigorous development of the large-scale model industry will help the AI industrialization process, change the interaction mode of massive applications, and create new growth space for the digital industry.

Yang Tao said that the financial industry needs to face the challenges faced by generative AI large models such as data protection, computing power protection, industrial policy protection, and industry regulatory protection.

"The safety and reliability of large-scale models has attracted widespread attention, and it is necessary to cultivate a new environment for large-scale models to be good and positive." Cao Feng said frankly.

Yang Xuan, founding partner of Warp Capital and founder of Baoguan Technology, believes that insurance is a model-driven industry, and a good agent and a good underwriting and claims expert are a good model. He pointed out that the current application of big models in the insurance industry is mainly in chatbots, marketing, salesman training, legal affairs, claims settlement and risk control. Regarding whether AI will replace insurance agents in the future, he believes that some jobs may be replaced first, but insurance agents must adapt to the era of generative artificial intelligence and co-create with AI. In addition, the application of large model technology in the insurance industry also faces challenges in training costs, employee skills, data security, etc., and the industry must actively respond.

At present, in addition to focusing on the exploration of specific business scenarios to empower the insurance industry with large models, Wei Chenyang believes that it is also necessary to actively discuss what is "more responsible AI", that is, how artificial intelligence can empower insurance in a more responsible and warm manner industry, and ultimately achieve safety, fairness, inclusiveness and sustainability.

"The next five years are full of imagination. How to grasp the opportunity of the large-scale model wave, we look forward to joining hands with various parties in the ecology to form a joint force and strive to produce high-quality large-scale model products in the vertical field and contribute to the industry." Gu Wei said.