Hundsun(SH:600570) Jing Ji Guan Cha Wang·2025-12-29 04:24Core Insights - The co-founder of Hengsheng Electronics, Liu Shufeng, stated that substantial breakthroughs in the industrial application of large models are expected by 2025, particularly in the financial sector [1] - Key applications include interactive entry points, document information processing, and customer service, with effective production in code generation also being validated [1] - The development of intelligent agents is anticipated to make significant progress within the next year [1] Group 1 - The hybrid use of large and small models is suggested as an effective approach, as traditional small models still hold value in processing structured data [1] - The issue of "hallucination" in large models cannot be fundamentally eliminated, necessitating the exploration of effective boundaries and acceptance of coexistence with this phenomenon [1] - In finance, this implies effective utilization of AI in high-value areas such as risk management and investment decision-making [1] Group 2 - Liu noted that the financial application of large models is still in its early stages, with key limiting factors including the effectiveness of private deployment, compliance risks, and budget constraints [1] - For many institutions, a wait-and-see approach may be a prudent strategy, alleviating anxiety about immediate implementation [1] - From a long-term strategic perspective, advancements in underlying technology are expected to transform business models and the foundational infrastructure of industries [1] Group 3 - Recent observations highlight the integration of data platforms and AI platforms, with the "ontology" business logic model being crucial [1] - The depth of industry know-how and abstraction capabilities remain the source of core competitiveness [1]