Group 1 - The core viewpoint emphasizes the rapid transformation of the automotive industry towards intelligence, driven by AI technology, with L2 driver assistance penetration exceeding 50% and automatic parking technology surpassing 20% in China [2] - The automotive industry is urged to accelerate the development of intelligent driving technologies, with a focus on popularizing driver assistance by 2030 and advancing L3 and above autonomous driving commercialization [3] - The current development of autonomous driving systems is lagging behind expectations, facing challenges in progressing from L3 conditional automation to L4 and L5 full automation [3] Group 2 - The development of automotive intelligence relies on a robust AI operating system (AIOS) as a foundational support, with many companies building multi-agent systems that require a unified technology base [4] - There is a growing demand for AI computing power across automotive enterprises, necessitating the rapid construction of high-quality intelligent computing infrastructure [4] - The industry is in a rapid iteration phase, and focusing solely on cost reduction could negatively impact user experience and safety [4] Group 3 - The relationship between vehicle manufacturers and component suppliers is evolving into a deep binding model, forming strategic alliances to enhance AI competitiveness and achieve win-win cooperation [5] - Successful collaboration examples include Jianghuai Automobile's deep integration with Huawei, resulting in significant sales for their new model [5] - Chinese automotive companies are increasingly engaging in reverse technology transfer through joint ventures and investments with foreign firms, enhancing their global competitiveness [6] Group 4 - The rise of intelligent vehicles presents new opportunities for Chinese automotive companies to expand globally, with L2 and L3 features leading the global average by nearly 20% [7] - Challenges for intelligent vehicles "going global" include addressing cybersecurity, data security, and privacy compliance issues [7] - The industry is encouraged to view automotive development in a broader context, integrating advancements in robotics, low-altitude technology, and shipping into a cohesive intelligent industrial chain [8]

地平线余凯：L4对算力的需求会达到5000TOPS 金十数据4月21日讯，近日，地平线创始人兼CEO余凯在演讲中表示，智能驾驶的L3/L4/L5前提都是足 够好的全场景辅助驾驶。L3是受限但清晰ODD边界的L4，加上全场景L2辅助驾驶。要实现L3，需要海 量L2全场景辅助驾驶系统车辆部署和实际道路数据的充分验证。在这个发展过程中，算力非常重要。 今年越来越多车企提出智驾平权，普及L2辅助驾驶，他预测，L3可能会在2-3年以后出现，L3的算力需 求在500-1000TOPS，2030-2035年可能会发展到L4，届时对算力的需求会达到5000TOPS。所以不管是 车企自研还是和第三方合作，想求生存一定要追赶更高的技术高度。 ...