At the Global Wealth Management Forum·2025 Shanghai Suhewan Conference held in Jing'an District, Shanghai, on October 18-19, Xiao Jing, Chief Scientist of Ping An Group, delivered a keynote speech on the transformative role of artificial intelligence (AI) in investment risk management.

Xiao Jing highlighted that AI has entered a new phase characterized by learning, deep reasoning, and generalization capabilities. To enhance market competitiveness, Ping An Group will focus on three strategic directions: leveraging multi-scenario advantages to accumulate domain-specific data, building foundational AI and model platforms to empower business teams with large-scale models, and establishing an intelligent agent platform to drive diversified AI applications.

**1. Next-Gen AI: A Four-Step "Perception-Learning-Memory-Thinking" Closed Loop** AI development has evolved through three stages: - **Stage 1 ("Manual + AI")**: Reliant on exhaustive human input (e.g., early AI customer service), with high labeling costs and limited scalability. - **Stage 2 ("Memorization Without Reasoning")**: AI gained learning capacity but lacked interpretability and error-tracing capabilities. - **Stage 3 ("Learning, Reasoning, and Generalization")**: Modern models like DeepSeek enable self-improvement, transparent logic, and error correction, forming a closed loop of perception, learning, memory, and reasoning.

**2. AI-Driven Transformation: Three Key Shifts** - **Client Evolution**: Rising trust in AI narrows knowledge gaps, expanding services from professionals to mass users. Future models may shift to "expert-guided AI agent" operations. - **Model Advancement**: Scaling Law propels large models from generalist to specialist capabilities. Ping An’s proprietary "Ping An Brain" system, trained on billions of high-quality financial and healthcare data points, achieves near-expert accuracy in medical diagnostics after reinforcement learning iterations. - **Ecosystem Growth**: AI-native applications (e.g., Ping An Good Doctor’s AI diagnosis, robo-advisors) emerge as interconnected agents, sharing capabilities via a unified model base. This "specialization-spillover-collaboration" flywheel elevates industry-wide intelligence.

**3. Sustaining Competitive Edge** Ping An’s trifecta for AI leadership: - **Deepened Models**: Weekly iterations of domain-specific models via proprietary data and reinforcement learning. - **Expanded Scenarios**: A unified platform supports 100,000+ intelligent agents annually, accelerating deployment across insurance, banking, and investment. - **Controlled Safety**: Explainable AI design, dynamic risk factors, and ethical governance ensure compliance and traceability.

Through these strategies, Ping An aims to fortify its AI-driven moat in the financial sector.