对话长江商学院梅丹青：AI时代金融服务的核心特征在于“可规模化的定制化”

Group 1 - The core viewpoint is that the financial system in China is at a historical juncture, with a focus on building a strong financial nation and enhancing the quality and resilience of the financial system during the "14th Five-Year Plan" period [1][17] - AI is expected to have a disruptive long-term impact on the financial industry, shifting the focus from process optimization to cognitive and decision-making intelligence enhancement [1][8] - The concept of the "impossible triangle" in finance—serving a large number of clients, providing highly customized services, and maintaining controllable costs—is becoming feasible in the AI era, characterized by "scalable customization" [1][8][24] Group 2 - The current phase is critical for integrating AI into the financial system, as the capabilities of large models are nearing a bottleneck, and the focus is shifting towards combining engineering capabilities with existing model capabilities for sustainable development [6][22] - Key investments in AI should focus on integrating business processes with AI, assessing AI's risk boundaries, and determining when human intervention is necessary, rather than merely investing in hardware or model training [6][22][23] - The financial industry emphasizes reliability and stability over rapid innovation, leading to a cautious approach in AI adoption, especially in the exploration phase before 2025 [9][25] Group 3 - Data security is a significant concern in the financial industry when integrating AI, as financial data is highly sensitive, necessitating careful handling and innovative solutions for model training [10][26] - A potential solution for data sensitivity is the local deployment of open-source large models, although this may pose higher costs for smaller institutions [10][26] - Financial institutions need to balance data security, model capabilities, and engineering feasibility to leverage AI effectively without compromising core assets [11][27] Group 4 - In the secondary market, AI applications are already widespread, with institutions using machine learning for investment decision-making, while the primary market is slower to adopt due to higher information asymmetry [12][28] - The future trend in the primary market is likely to be a "human-machine combination" model, where AI provides valuation benchmarks, but human judgment remains crucial for investment decisions [13][29] - The financial industry should focus on direct applications of AI rather than excessive investment in retraining large models, as the core task is to effectively utilize existing models [15][30]