Analysys: ChatGPT and GPT-4 have been released one after another, triggering our imagination and exploration of the possible application directions of large models in various industries, including the financial industry we highlighted today. In the financial industry, large models have many application scenarios, including investment research, product development, credit review, etc. For example, using Copilot to help programmers write code, or to do code inspection and testing, etc., can also use large models Multi-round dialogue capability, maximizing the quality of customer service, thereby improving user experience, etc.
In the process of truly promoting the implementation of the large-scale model in the financial industry, the following four challenges need to be considered. The last challenge is the most important and has the longest impact.
The first one is the challenge of credibility . While we are admiring the ability of ChatGPT, we should all notice that it has exposed some problems in the process of talking to us, such as the answer is not accurate enough, or even wrong, such as For the same question, there may be diametrically opposite answers, and the financial industry is an industry that has very high requirements for model interpretability and robustness. The interpretability of the current output results of large language models is currently relatively closed and opaque. At the same time, its stability is still disturbed by data, algorithms, training, etc., and non-robust features appear. These are very important challenges for the financial industry to implement large-scale model applications in terms of credibility.
Second, the challenge of business understanding . The current large language model is actually trained based on a general knowledge base, but when it enters the financial business scenario, whether it is credit, wealth management, etc., it still needs to target the business attributes of the financial industry Incremental training with business logic can truly solve business problems and realize intelligent decision-making.
Third, the cost input challenge . The application cost of the current large model is still relatively high. On the one hand, it is a huge cost of computing power consumption; The cost of corpus, data labeling, and model training. Although the financial industry has relatively high technology investment, it still needs to further reduce the application cost through model compression, small sample training, etc., before it can be put into production environment in a real sense.
Fourth, and the most important challenge we just mentioned is the challenge of organizational capabilities . Of course, the financial industry can replace manpower through the application of large models to do a lot of mechanical, process, and even some creative work. However, what financial institutions need to consider at the same time is how to create an organic synergy and cooperation relationship between humans and machines or humans and artificial intelligence. On the one hand, how to better empower employees and humans, and improve people’s ability to use AI tools On the other hand, it also includes continuously adjusting and optimizing the functional boundaries between human and digital employees.
Only when we seriously discuss the risks and challenges, it means that we are not far from entering the era of financial intelligence opened by the big language model. If you have any questions about the digitalization of the financial industry, please leave me a message and let's discuss together .
Disclaimer: The third-party data and other information quoted by Analysys in this article are all from public channels, and Analysys does not assume any responsibility for this. In any case, this article is for reference only, not as any basis. The copyright of this article belongs to the publisher. Without the authorization of Analysys Analysis, it is strictly forbidden to reproduce, quote or use any content released by Analysys Analysis. Any media, website or personal use after authorization should quote the original text and indicate the source, and the analysis point of view is subject to the official content released by Analysys Analysis, and any form of deletion, addition, splicing, deduction, distortion, etc. . Analysys does not assume any responsibility for disputes arising from improper use, and reserves the right to pursue responsibility against the relevant responsible parties.