wo[Table_Title] Research Report 28 Jul 2025 电子 Technology WAIC 2025 前沿聚焦（4）：从模型驱动向意图驱动的重大范式跃迁 WAIC 2025 Frontier Focus (4): A Paradigm Shift from Model-Driven to Intent-Driven 姚书桥 Barney Yao 邓雅文 Linda Deng barney.sq.yao@htisec.com linda.yw.deng@htisec.com [Table_yemei1] Flash Analysis [Table_summary] (Please see APPENDIX 1 for English summary) 事件： 2025 年 7 月 27 日，在上海举行的 2025 世界人工智能大会明确指出，人工智能正从"模型驱动范式向"意图驱动"范 式加速跃迁。模型驱动依赖于大规模统计相关性和预定义规则，将输入映射为输出；而意图驱动则强调将人类目 标、价值体系与 AI 的信息处理深度融合，使系统能在缺乏明确指令时自主识别并分解目标、执行闭环反馈 ...

Core Insights - The article emphasizes the importance of AI applications and tailored products for key task networks in maintaining business success [2] - Ericsson is increasing its innovative collaboration with Chinese operators and industry partners, highlighting China's potential in the global market, especially in 5G capabilities [2][3] - Challenges in the Chinese market include low monetization ability and user value (ARPU), with many commercial software solutions failing to capture the potential value from data growth [2] Group 1 - Ericsson's global network business strategy head, Hans Hammar, identifies numerous opportunities in the Chinese market, leveraging advanced technology and global best practices [3] - The competitive assessment by Omdia shows that Nokia, ZTE, and Ericsson are leading in the end-to-end 5G private network infrastructure market [3] - Ericsson has deployed approximately 40 to 45 independent networking and 5G Advanced networks globally, with Chinese operators like China Mobile at the forefront [3] Group 2 - Future service delivery based on quality and dynamic network slicing cannot be achieved through traditional deployment models, requiring breakthroughs in service perception and intent-driven network management [3] - Over 85% of current network functions possess service recognition capabilities, allowing for autonomous adaptation to business needs [3] - Ericsson has validated intent-driven energy management solutions in mainland China, enabling networks to optimize resource allocation while reducing energy consumption [3]

本文转载自「深思圈」 未来，对于开发者来说，AI 不再是工具，而是构建软件的全新基础。 基于 AI Agent 驱动下，版本控 制、模板、文档，甚至用户概念正在被重新定义。 近日，a16z 发文提出了 9 个未来开发者趋势，虽然还处于早期阶段，但都是基于真实的痛点，非常具 备前瞻性。这些趋势包括重新思考 AI 生成代码的版本控制，到大语言模型驱动的用户界面和文档。 TLDR： Founder Park 正在搭建「 AI 产品市集」社群，邀请从业者、开发人员和创业者，扫码加群： AI Agent 编写或修改大量代码，开发者更关注代码输出是否符合预期，而不是具体的代码行。这 就导致「真相的上移」，prompt 和测试组合成为新的「真相」，进而促使意图驱动的版本控制出 现，未来可能将 prompt + 测试包作为可版本化的单元来跟踪。 传统仪表板是静态的，展示固定的指标，用固定的方式。但 AI 驱动的仪表板可以根据用户当前 的任务、角色、甚至过去的行为模式来重新配置。 文档正在逐步演变为交互式知识系统，这些系统具备语义搜索能力，可以作为编码 Agent 的上下 文来源。未来的文档可能会有三个层次：人类阅读层（有故事 ...