新京报贝壳财经讯（记者韦英姿 罗亦丹）北京两会期间，市政协委员，百望股份创始人、董事长陈杰建议，北京市进一步扩大"算力券"的发放规模和覆盖 范围，调整支持重点，降低微调成本。"算力应当像水电煤一样，成为企业'用得起、用得好'的公共资源。" 市政协委员，百望股份创始人、董事长陈杰（右）。受访者供图 对此，她建议政府设立专项基金，支持行业头部企业与在京高校深度合作，例如共建实训实验室、进行数据与场景共享，让学生在校期间就能接触到真实 的"税务逻辑"或"供应链难题"，在"战壕"中练兵，毕业即有机会成为既懂技术又懂业务的复合型人才。 "还应打破高校与企业之间的人才围墙，鼓励高校科研人员到企业挂职'首席科学家'，直接参与产品研发；同时鼓励企业技术高管进入高校担任'产业导师'， 传授实战经验。政府可以在职称评定、科研经费配套、人才落户等方面，对参与此类跨界交流的人员给予专项加分或政策倾斜，提升人才跨界流动的积极 性。"陈杰表示。 校对 王心 她指出，应当将政策支持重心从基础大模型研发，向"垂直行业应用"延伸，例如重点支持财税大模型、供应链风控大模型等"垂直行业大模型"研发。另外， 应当针对企业利用国产算力芯片进行模型适配 ...

Group 1 - The core viewpoint is that the integration of AI with key industries presents a trillion RMB-level growth opportunity, evolving from "tool empowerment" to "business integration" and ultimately "ecosystem reshaping" [1] - The application of AI is shifting from isolated attempts to systematic integration within core business processes, indicating a deeper fusion with IT systems [1] - Companies are increasingly focusing on smaller, specialized models (SLM) due to their lower costs, faster response times, and better data privacy, rather than solely pursuing large models [1] Group 2 - Key industries for AI integration include finance, healthcare, and manufacturing, with manufacturing focusing on smart upgrades to drive automation and improve yield rates [1][2] - Vertical industry large models are becoming focal points for commercialization, particularly in AI-assisted diagnostics in healthcare and intelligent risk control in finance and law [2] - AI applications are moving from concept validation to production, with AI customer service, scheduling, and operational services becoming integral to core business functions [2] Group 3 - Development bottlenecks include data quality issues, high costs of AI research and computing power, and the challenge of adapting general large models to specific industry needs [2][3] - There is a scarcity of composite talent that understands both AI and specific industries, which poses a significant challenge for AI adoption [3] - The clarity of business models for revenue generation through AI remains uncertain, alongside increasing regulatory requirements regarding data privacy and algorithm fairness [3]