2 1 Shi Ji Jing Ji Bao Dao·2025-09-25 08:03Core Viewpoint - Alibaba Group's CEO emphasizes that AI will be the next generation operating system, with aspirations towards Artificial Superintelligence (ASI) [1] Group 1: AI in Automotive Industry - The automotive industry is increasingly integrating AI agents into smart cockpits, with companies like Li Auto, BYD, and NIO leading the charge [1][2] - Li Auto's AI agent can perform tasks such as ordering food through voice commands, showcasing the practical applications of AI in vehicles [1] - The goal is to create a closed-loop system for users that connects "people, cars, and life" through the vehicle's AI capabilities [2] Group 2: Functionality of AI Agents - AI agents in vehicles primarily assist with navigation, food ordering, and ride-hailing, but require external ecosystem connectivity to function effectively [3] - Li Auto's AI agent operates on two frameworks: CUA (Cockpit Using Agent) and MCP/A2A (Multi-Channel Processing/Agent-to-Agent), allowing for task execution through third-party applications [3][5] - CUA is the prevalent technical route, enabling multi-modal understanding and task execution via apps like Alipay [3] Group 3: Challenges and Solutions - CUA faces limitations, particularly in complex tasks where accuracy is low, with only about 30% success in multi-step tasks [4] - Li Auto proposes a solution by breaking down tasks into manageable steps, improving prediction accuracy for user actions [4] Group 4: Future Developments - The vision for Li Auto's AI agent includes "full memory" and "environmental perception," allowing the agent to remember user actions and recognize environmental cues [7] - Full memory encompasses user interactions and relationships, enabling the AI to proactively assist rather than just react [7] - Environmental perception is crucial for the AI to interact with real-world data, enhancing its ability to perform tasks effectively [8]