In recent discussions surrounding the digital transformation of financial institutions, a plethora of measures have emerged to not only leverage digital technologies to enhance financial services but also to lay a robust foundation for the development of digital financeThis shift is pivotal in a world increasingly reliant on technology, necessitating a comprehensive approach toward modernization across various financial realms.

The shift towards what is being termed "new infrastructure" in finance is driven by the advent of advanced technologies such as AI models, data asset integration, supply chain finance, and blockchainThis transformation signifies a crucial step forward in realizing an innovative financial ecosystemAs part of this ongoing narrative, our publication is excited to unveil a series entitled “The Progress of Financial New Infrastructure,” showcasing the latest achievements in this domain while contemplating the future landscape of finance.

The financial sector has embraced the capabilities of large AI models with vigor

In the current year alone, numerous financial service providers, ranging from data terminals to fintech giants, have introduced applications based on these advanced modelsNotably, the focus has been on areas like human-computer interaction, intelligent investment analysis, and smart advisory services— all areas experiencing increased engagement from investors eager to utilize artificial intelligence to augment their investment capabilities.

Industry experts indicate that the integration of AI technology empowers average investors to conduct analysis and make decisions comparable to those made by professionalsThis rapid incorporation hints at a potentially transformative shift toward democratizing professional investing, setting the stage for what is being described as an imminent era of inclusion in investment strategies.

Pioneering Applications of AI Models

Since the start of this year, notable players in the financial data services sector, such as Wind Information, Hang Seng Qiyuan, and Dongfang Caifang, have accelerated development in AI, unveiling numerous applications powered by large models

• US November CPI Seals the Deal on December Rate Cut
• ETFs Leading the Tech Charge
• Nikkei 225 ETFs See Wild Swings
• Leading the Way in New Energy Vehicle Production
• Evergrande's Losses Exceed 800 Billion: Debt Restructuring Crucial

Ant Group and SenseTime are also notable participants, with Ant Group launching "Zhixiang" earlier this year, an AI-driven investment research assistant available for free to users to bolster their data processing and analytical skillsThis innovation is pivotal for investors, refining how they manage data and perform analyses, dramatically enhancing their capabilities.

The burgeoning field of human-computer interaction has seen advancements with AI tools such as the Alice robot from Wind and the WarrenQ robot by Hang Seng QiyuanThese bots signify a transformative approach to investment analytics.

“AI robots enable professional-level investment research and analysis,” stated Bai Xue, Deputy General Manager at Hang Seng QiyuanThe comprehensive capabilities of these large models are manifested in conversational chatbots, enabling users to seamlessly gather and organize information by simply posing a question.

For instance, a user can prompt a chatbot to generate a research report on a publicly traded company

The bot will then categorically distill basic company information and financial data, facilitating continuous interaction for deeper data processingTasks such as sourcing economic data, constructing visual aids, or crafting sophisticated models that previously consumed substantial time and effort can now be accomplished with ease.

Historically, asset management firms would require analysts to condense daily research report insights swiftly, a process often referred to as creating “dehydrated reports.” Nowadays, this service is almost universally available on large model platforms, with enhanced capabilities to compare perspectives among various corporate and sector organizations, alongside customized services under continuous development, as highlighted by Cao Hui, a partner at Bohui Cloud Technology.

At present, large models have found widespread application within capital markets, influencing investment research, advisory services, intelligent customer service, marketing, and risk control.

According to Bai Xue, the primary avenues for large model application involve knowledge-base querying, smart investment research, and advisory systems

These innovations are poised to significantly enhance the efficiency of human-machine interactions in capital markets, with projections suggesting an explosive growth period for large model applications, potentially achieving an average annual growth rate of 60% over the next decade.

Simplifying Professional Investment

Insiders believe the swift deployment of large model applications could enable investors to execute analyses and make decisions on par with professional standards.

As highlighted by Zhang Qilong, co-founder of Jiuyan Technology, technological advancements mean large models have greatly improved the handling of unstructured data—like text, images, audio, and video—leading to swifter acquisition, processing, and analysis of informationThe financial market, as a result, can operate more efficiently due to enhanced information processing and fluidity.

For instance, investors frequently receive multiple video conference invitations simultaneously

Today, provided by large models, the functionality of a "conference avatar" allows a robot to automatically attend meetingsFollowing attendance, the robot can record the session, perform speech-to-text conversion, conduct real-time translation, and compile a summary of themes and key takeaways.

Another example can be observed with central bank monetary policy execution reportsPresently, large models have the capacity to compare different periods' reports, facilitating a thorough analytical overviewThis capability similarly extends to the many meetings conducted by the Federal Reserve.

Zhang Qilong asserts that the application of large models in finance conveys a process of making professional investment more accessibleIt is expected to substantially enhance investment research efficiency, condensing hours of work into mere minutes.

Additional applications of large models extend to institutional sectors

As noted by Cao Hui, collaborations between technology firms and brokerages, funds, and investment institutions are advancing the interconnectedness of internal and public databases, particularly enriching AI multimodal search scenarios tailored for English-language databasesMore complex operational workflows are also being developed from AI search systems, featuring capabilities such as automated reporting and industry chain monitoring.

Leitao, CEO of Tianyun Data, recognizes that deploying large models in regulatory contexts also presents a viable solutionAI technology can identify risks such as off-market financing while building comprehensive risk relationship diagrams for asset management trust products across the entire market.

Continued Room for Enhancement

Presently, the most substantial challenge facing large model development is "AI hallucination." This phenomenon occurs when generated content is inconsistent with factual information or unrelated to user inputs.

Zhang Qilong explains that the finance industry demands exceptionally high precision levels, emphasizing interpretability and risk management

Therefore, financial institutions exercise caution in many scenarios of large model application.

Cao Hui believes the issue of "AI hallucination" arises primarily from data pollution during training, underscoring the necessity of advancing precision from a software engineering perspectiveIt also requires industry standards and regulations to better guide large model technology, ensuring its efficacy in supporting societal and economic development.

"To mitigate 'AI hallucination' challenges, two strategies can be implemented: regularly updating financial databases and reengineering processes to ensure that large models operate within compliant data retrieval frameworks," Bai Xue advocates.

Moreover, the models exhibit limitations in reasoning capabilitiesMin Min, Vice Dean of the Shanghai International Financial Center Research Institute, notes that longer logical chains in large models often expose vulnerabilities

They are also constrained within their dataset frameworks, limiting the ability to transcend boundaries or engage in self-reflection.

However, the potential of large models should not be underestimatedLeitao emphasizes that the current phase of applications is focused on service implementation, predicting that a major developmental pathway will be to establish comprehensive modeling and computational infrastructure for capital markets, simultaneously lowering barriers for utilizing AI capabilities across the sector while facilitating the scaled production of synthetic data through large models.

Regarding future applications of large models, Zhang Qilong identifies two primary directions: first, enhancing the processing of financial news, public sentiment, and various textual data to quantify the correlation and causation between market movements and media narratives using advanced semantic understanding and sentiment analysis