点 击 蓝 字 关 注 我 们 本文全文共 1551 字，阅读全文约需 5 分钟 2025世界人工智能大会在沪举办 7月26日至28日，以"智能时代 同球共济"为主题的2025世界人工智能大会暨人工智能全球治理高级别会议 （WAIC）在上海举办。此次大会聚焦AI技术产品首发首展、AI赋能千行百业以及人工智能全球治理等关键话 题，全面展现人工智能领域的最新进展与未来走向。 上海智能网联汽车驶入新阶段 在"模数引领，智行未来"AI赋能自动驾驶创新发展论坛上，《上海高级别自动驾驶引领区"模速智行"行动 计划》发布，总体目标为2027年基本建成全球领先高级别自动驾驶引领区。当天，上海市发放了新一批智能网 联汽车示范运营牌照。 亿咖通集中展示在智能座舱、辅助驾驶和车载AI大模型等领域的最新成果。亿咖通基于"龍鹰一号"打 造"安托拉系列计算平台"，支持单SoC实现"舱行泊一体"功能；基于通用基础AI大模型，实现AI Agent驱动的 全域感知与生成式体验。 多家企业发布不同用途大模型 Robotaxi规模化应用提速 WAIC 2025还设立了无人驾驶体验区。活动期间，上汽智己、小马智行、百度智行（萝卜快跑）、奇瑞汽 车在 ...

Core Insights - The launch of the first edge-based multimodal large model solution by Zebra Zhixing, in collaboration with Tongyi and Qualcomm, marks a significant advancement in automotive intelligent cockpit technology, enabling a 90% service loop of "perception-decision-execution" purely on the vehicle side [1] - The solution allows for an upgrade from a "command receiver" to a "dialogue participant," enhancing user interaction and experience [1] Group 1 - The edge-based multimodal large model solution is based on the Qualcomm 8397 platform [1] - The solution facilitates proactive intelligent cockpit features, such as automatically adjusting the air conditioning based on user status and environment [2] - The integration of AI agents in vehicles is transitioning from marketing-driven to scenario-driven applications, indicating a shift in the automotive market [4] Group 2 - The demonstration of the AI capabilities included a real-life scenario where an AI agent facilitated a coffee order through natural dialogue, showcasing human-machine coexistence [2] - The AI agent can also recommend playlists to alleviate user anxiety during traffic congestion, further enhancing the driving experience [2] - The intelligent cockpit large model is identified as a key technology for experiential upgrades in the automotive sector [4]

Core Viewpoint - The collaboration between Zebra Zhixing, Tongyi, and Qualcomm aims to advance automotive smart cockpit technology into an era of proactive intelligence through the release of an edge-side multimodal large model solution [1][2]. Group 1: Product Features - The edge-side multimodal large model solution, based on the Qualcomm 8397 platform, can achieve a 90% service loop of 'perception-decision-execution' purely on the vehicle side [2]. - The solution enables a generational upgrade from being a "command receiver" to a "dialogue participant" through multimodal intent perception and interaction [2]. - The local lifestyle agent, part of the "One Arrow Ten Stars" interactive intelligence released in April, reconstructs food delivery services based on travel scenarios [2]. Group 2: User Experience - Users can interact with the AI to order coffee through natural dialogue, with the AI providing a seamless experience that surpasses mobile operations [2]. - The AI can proactively adjust the vehicle's environment based on user status, recommend playlists during traffic jams, and engage in multi-user dialogue scenarios [2]. Group 3: Strategic Development - Since launching the "AI in All" strategy in 2024, Zebra Zhixing has rapidly iterated and commercialized its Yuan Shen AI across various automotive brands, including Zhiji, Roewe, and BMW [3]. - The company is moving towards a human-machine symbiosis by collaborating with customers and the ecosystem, starting from proactive intelligence [3].

Core Insights - Zebra Network Technology Co., Ltd. (Zebra Smart Travel) launched an edge-based multimodal large model solution in collaboration with Tongyi and Qualcomm, marking a significant advancement in automotive intelligent cockpit technology [1][2] - The solution, based on the Qualcomm 8397 platform, enables 90% of the "perception-decision-execution" service loop to be achieved purely on the vehicle side, enhancing user interaction from a "command receiver" to a "dialogue participant" [1] - The local lifestyle agent, part of the "One Arrow Ten Stars" interactive intelligence released in April, integrates cloud-based large models with edge voice capabilities, currently implemented in the new Zhiji L6 vehicle [1] Industry Trends - The automotive market in China is witnessing a shift from marketing-driven to scenario-driven deployment of large models, with AI agents beginning to be mass-produced and providing new user experiences [2] - The intelligent cockpit large model is identified as a key technology for experiential leaps in the automotive sector, emphasizing the importance of user-centric design and functionality [2]

Group 1 - The core viewpoint of the articles highlights the rapid advancement and adoption of AI agents across various industries, with 2025 being seen as a pivotal year for their emergence, and an expectation of over 1 billion AI agents by 2026 [1] - AI agents are being utilized to address common challenges in industries such as textiles, where they enhance the accuracy and speed of fabric inspection, thereby reducing costs and improving efficiency [1] - China Telecom has developed over 80 industry-specific large models and more than 20 AI agent applications, serving over 20,000 industry clients, showcasing the extensive application of AI technology in sectors like industrial, emergency, and education [1] Group 2 - In the office sector, Mido Technology Co., Ltd. launched V Assistant 2.0, a multi-agent collaborative intelligent body for comprehensive public opinion analysis, demonstrating the versatility of AI agents in various applications [2] - AI agents are increasingly being integrated into business processes, particularly in clearly defined task environments, allowing them to take on more routine execution tasks and improve operational efficiency [2] - An example from Belle Fashion Group illustrates the successful implementation of over 800 AI applications across various business nodes, significantly enhancing information connectivity and process automation [3]