ChatGPT leads the new wave! AI large-scale model drives financial technology innovation and development, enabling business to continue a new chapter

Original: Mu Chen

Source: Seven Finance

Image source: Generated by Unbounded AI tool

The advent of the Internet changed the way information flows and drove the form of new corporate paradigms. And the changes are still going on, especially the emergence of ChatGPT, which brings people a more in-depth intelligent interactive experience.

Based on GPT (generated pre-training change model), the all-round AI dialogue robot - ChatGPT may make mistakes when answering questions, but its logical ability in the process of chatting with people is astonishing.

For a time, there was an upsurge of artificial intelligence large models represented by ChatGPT, which attracted unprecedented attention from the market. Baidu, 360, Alibaba Cloud, SenseTime, HKUST Xunfei, etc. joined in this large-scale model melee.

As far as the financial industry is concerned, as a highly digitalized and specialized field, it naturally becomes the best scenario for large-scale model implementation.

How does the big model empower the financial industry to show its "supernatural power"?

According to a report jointly released by the Boston Consulting Group (BCG) and the China Development Foundation, it is estimated that by 2027 about 23% of jobs in China’s financial industry will be disrupted by artificial intelligence, and the remaining 77% will be disrupted by artificial intelligence. Jobs will be powered by artificial intelligence and work hours will be reduced by about 27%.

The prediction of the impact on the labor force in the financial industry proves that AI is no longer a robot boy who is programmed into emotions in Spielberg's "Artificial Intelligence", but has truly penetrated into the entire business chain. And now, with the wave of general-purpose large models rolling up, the financial industry also has new expectations for artificial intelligence.

The two sides agree on the fact that the financial industry produces and processes a large amount of data, and the large artificial intelligence model, especially the large model based on deep learning, is good at dealing with this kind of data-intensive environment. This ability is very important for risk assessment, fraud detection, and market expectations etc. are crucial.

Moreover, financial data usually contains complex patterns. Artificial intelligence models have unique advantages in dealing with complex patterns, and can better deal with high noise, high dimensionality, and nonlinear characteristics in the data, thereby helping financial institutions identify market trends and make more accurate decisions. decision making. In addition, large-scale artificial intelligence models can efficiently process and analyze large-scale financial data in a short period of time, enabling financial institutions to quickly respond to market changes and identify abnormal situations.

According to the relevant person in charge of Mama Consumer Artificial Intelligence Research Institute, in terms of intelligent interaction, financial knowledge and product-related information are added to the knowledge base one by one through the deployment of robot customer service. and accuracy, because its recognition ability is limited, and it plays more of a role of assisting human customer service. The large model itself has a lot of general knowledge. In addition to financial common sense, for other special content, it can be given to the large model through knowledge injection, and through continuous and sufficient training, the large model can be equipped with more accurate semantic understanding and Powerful natural language generation capabilities. Naturally, the big model becomes an "expert" who understands finance.

In addition, Rong 360 said that artificial intelligence large models can help financial institutions improve customer service quality. By analyzing vast amounts of customer data, these models can personalize services, predict customer needs, and provide tailored recommendations. Not only that, but the AI general-purpose large model can also improve the efficiency and accuracy of risk assessment. The capabilities of large models include technologies such as deep learning and natural language processing, which enables it to process and understand large-scale information, bringing more efficient and precise risk management to the financial industry, thereby enabling financial institutions to make more informed loans decision making.

And artificial intelligence large models can greatly enhance the ability of fraud detection. It can analyze and understand a large amount of structured and unstructured data, so it can identify fraudulent behavior and abnormal patterns hidden in huge data sets, and continuously improve the performance of fraud detection. Accuracy and efficiency, so that financial institutions, e-commerce platforms, etc. can detect fraud in a timely manner, reduce financial losses and protect the interests of users.

Promote the construction of large model technology application data elements become the key

No matter how "gorgeous" the technology is, it is not as practical as the actual application. According to data from CCID Consulting, it is estimated that by 2025, the scale of the domestic artificial intelligence industry will reach 336.93 billion yuan, an increase of 63.85% compared with 2022; the market scale of driving industry application comprehensive solution services will exceed 3 trillion yuan.

For the financial industry, the "Several Measures of Beijing to Promote the Innovation and Development of General Artificial Intelligence (2023-2025) (Draft for Comments)" issued by Beijing clearly supports financial technology companies in financial scenarios with high information load and fast information update. , it is difficult for financial practitioners to quickly and comprehensively obtain accurate information, and explore the application of artificial intelligence technology for in-depth understanding and analysis of financial texts.

picture

And on the basis of focusing on intelligent risk control, intelligent investment advisory, intelligent customer service and other links, promote the accurate analysis of financial professional long texts and the update of model knowledge, break through the fusion technology between complex decision logic and model information processing capabilities, and realize complex The conversion of financial information processing to investment decision-making suggestions supports investment-assisted decision-making in the financial field.

In this regard, the relevant person in charge of the Immediate Consumer Artificial Intelligence Research Institute mentioned that if the general-purpose large model is regarded as a wild horse with outstanding qualifications, creating large-scale model applications that focus on financial vertical fields and subdivided scenarios is equivalent to domesticating the wild horse . First of all, it needs to be "fed" with proprietary processed vertical field data as "grass"; secondly, it is necessary to fine-tune and align the model in the vertical field, which is equivalent to putting a "bridle" on the wild horse; thirdly, Use the reasoning acceleration technology of the large model to add "saddle" and "stirrup" to it to make the horse run faster and more controllable; finally, there must be enough application scenarios for the horse to gallop and iterate, use The more people involved, the more evaluation feedback, and the faster the model iterates, the better it will be. In this regard, large financial institutions have inherent advantages and can produce strong practical effects. In contrast, the first difficulty facing small and medium-sized financial institutions is the resource threshold. Under the influence of strong demand, they will seek assistance from large institutions in the financial industry or financial technology platforms with technological advantages to establish relevant technological capabilities.

According to Qicai Finance, Qifu GPT, a large-scale industry model developed by Qifu Technology, has achieved phased results. As the first general-purpose large-scale model of the financial industry in China, the product-level applications supported by it are expected to be launched within this year and open to financial institutions. Qifu Technology believes that as a large-scale model of the financial industry, it must be the ultimate in accuracy and applicability. Therefore, the quantity and quality of training data and the understanding and insight of financial business have become the core competitiveness of large models in the financial industry.

Qifu GPT relies on a large amount of financial business data accumulated by Qifu Technology over the years, whether it is 5000w+ credit reports and interpretations, in-depth dialogues with monthly users of 350w+, or relying on more than 900 industries, with 3000+ The corporate financial behavior network of 16 million+ enterprises with attributes and the knowledge graph and industry knowledge derived from it are the basis for Qifu GPT to better understand finance, understand users better, and better support various financial businesses in the credit field.

At present, Xinye Technology combines large models to explore the layout of artificial intelligence. On the one hand, it has verified that large models can help improve accuracy in some existing scenarios, such as improving the ability of robot speech and text analysis, understanding and generation, and creating better users. On the other hand, we are also exploring new scenarios based on generative models, including automatic code generation, visual material design, etc., embracing the productivity changes brought about by generative AI.

In the first quarter of 2023, Lexin has accelerated the application of artificial intelligence large models in the financial vertical field in business. At present, the Lexin artificial intelligence large model has been implemented in the fields of research and development code assistance, design idea generation, telemarketing and customer service intelligent services, and has achieved significant efficiency improvements. In the future, Lexin will continue to promote the in-depth exploration of artificial intelligence large models in the fields of risk management and anti-fraud.

In addition, Samoyed Cloud Technology Group mentioned that based on the accumulation of AI decision intelligence, big data and other technologies, the company conducts research on large models in the following areas, and continues to increase technology investment to explore more scenario applications: First, automatic modeling, Using the latest NLP large model technology, explore automatic model building through multiple rounds of dialogue, allowing users to describe the application they want to create through natural language, and then build a model. In addition, users can provide improvement suggestions through continuous natural language, and automatically make modeling adjustments; the second is to introduce ChatGPT technology in the field of cross-border e-commerce, and create a new AI tool for free for Amazon sellers, helping small and medium-sized enterprises reduce costs and increase efficiency.

According to Centaline Consumer Finance, with the explosion of ChatGPT, it once again proves that the era when innovation is king has arrived. Only reformers advance, only innovators are strong, and only reformers and innovators win. With digital "intelligence" management, "empowerment" financial services, and accelerated development, the company has created a market-leading three digital core capability systems of "independent customer acquisition", "intelligent risk control" and "digital operation", providing customers with High-quality, efficient, convenient and warm integrated consumer financial services.