Workflow
上汽乘用车张栋林:大模型与汽车的融合仍面临多重挑战

Core Insights - The forum on "How Intelligent Connected Vehicles Reshape New Ecosystems" highlighted the systemic and comprehensive transformation brought by large models in automotive intelligence, while also acknowledging multiple technical and non-technical challenges faced in this integration [1][6][10] Group 1: Intelligent Driving and Safety - Zhang Donglin rated the safety level of current L2-level assisted driving at 7 out of 10, emphasizing that while the system can control steering and acceleration, the driver must monitor the environment at all times [3][4] - The automotive industry must accurately define and communicate the capabilities and limitations of assisted driving features to users, ensuring a clear understanding of the system's boundaries [3][5] Group 2: Industry Trends and Developments - The rise in popularity of intelligent assisted driving in 2025 is driven by advancements in domestic SOC chip development and sensor industrialization, alongside stricter regulatory requirements for safety design and testing [5] - SAIC Group is actively promoting intelligent features, with the MG7 being the first global model equipped with advanced driving assistance systems, and the upcoming MG4 set to enhance cabin intelligence with innovative connectivity features [5] Group 3: Large Model Integration Challenges - The integration of large models in automotive applications presents challenges such as balancing decision-making speed with real-time response, and managing the trade-offs between low power consumption and high computational demands [8][9] - The automotive industry faces uncertainties and quality risks associated with new technology applications, including the interpretability of large model decisions and the complexities of engineering deployment [8][9] Group 4: Future Outlook - Despite the challenges, the automotive sector will continue to balance safety, user experience, and cost in the development and application of new technologies [10]