CM BANK(SH:600036) 第一财经·2025-09-06 12:58Core Viewpoint - The application of large models in the financial industry is entering a phase of accelerated implementation, transitioning from proof of concept to large-scale integration in business processes, customer service, and organizational structures [2][4]. Group 1: Current State of Large Models in Finance - The rapid development of domestic large models has led to significant changes in the financial sector, with a notable shift from concept validation to practical application [4]. - As of August this year, OpenAI released GPT-5, which, despite not fully meeting market expectations, has shown substantial improvements in its foundational model capabilities and reduced hallucination phenomena [4]. - Experts predict three major trends for large models by 2025: enhanced multi-modal deep reasoning capabilities, improved video generation abilities, and increased agentic capabilities for complex multi-step tasks [4][6]. Group 2: Challenges in Implementation - Despite advancements, challenges remain in adapting foundational models to banking logic, suppressing hallucinations, and ensuring that technology departments' developments resonate with business units [5]. - Key strategies for enhancing large models' effectiveness in solving professional problems include context engineering, enterprise-level knowledge management, and post-training [5]. Group 3: Future Development and Investment Opportunities - Future applications of generative models are expected to extend beyond digital content into physical environments, requiring models to possess greater adaptability and generalization capabilities [6]. - The potential for investment in areas such as embodied intelligence, life sciences, industry models, AI agents, and AI hardware is significant, with some sectors already generating revenue [6]. - The concept of "scene cognition" is highlighted as a crucial direction in the AI-first era, with a shift towards proactive AI that can autonomously understand and respond to its environment [7]. Group 4: AI Strategies in Banking - Many banks have initiated AI banking strategies, with examples including WeBank's transition to an AI-native bank and the launch of AI product matrices by MyBank tailored for small and micro enterprises [8]. - China Merchants Bank has adopted an "AI First" philosophy, prioritizing investments in talent, finance, and computing power, with a reported 10,800 R&D personnel, accounting for 9.13% of total employees, and an IT investment of 4.444 billion yuan, representing 2.93% of revenue [8].