[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fqrNfTV3uNSYGE-hiwT6Gd41bkZnlNqCxGOxBkUqZVNE":3},{"code":4,"msg":5,"data":6},200,"操作成功",{"id":7,"title":8,"content":9,"digest":10,"source":10,"coverPath":11,"thumbsCoverPath":12,"isTop":13,"isShow":14,"baseClick":13,"clickCount":15,"createTime":16,"typeId":17,"isNewest":18,"newsInfoTypeRespVo":19,"voiceUrl":22,"voiceSize":23,"taskId":24,"releaseTime":25,"titleEn":26,"contentEn":27,"voiceUrlEn":28,"taskIdEn":29,"voiceSizeEn":30},1525,"王江平详解如何破除AI科学发现“堰塞湖”","\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">中新网北京12月16日电 (记者 刘文文)中国新闻社16日在北京主办以“新格局·新动能”为主题的“国是论坛：2025年会”。围绕如何破解AI科学发现的“堰塞湖”困境，工业和信息化部原副部长、工业和信息化部电子科技委主任王江平在会上展开深入分析。\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002F8f2645c2825749e8ae3ed06f0a833e58.png\" width=\"739\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp class=\"ql-align-center\">\u003Cspan style=\"color: rgb(187, 187, 187);\">12月16日，工业和信息化部原副部长、工业和信息化部电子科技委主任王江平在北京参加由中国新闻社举办的“国是论坛：2025 年会”。 中新社记者 蒋启明 摄\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">王江平表示，当前AI的预测成果呈现指数级迅猛增长，但人类的验证能力和产业化能力却呈线性增长，两者之间差距巨大。AI一天的预测结果，人类需要10年甚至更长的时间来验证，这种矛盾就像“堰塞湖”一样堵塞了科学发现转化为实际应用的通道，不仅导致海量预测成果无法及时得到实验验证和产业化应用，还占用了大量科研资源和算力资源。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">为什么会出现“堰塞湖”？他分析，主要有以下原因。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">第一，预测模型具有局限性。比如，逻辑推理和知识深度不足、存在黑箱困境与幻觉风险、目标推导能力有限等。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">第二，缺乏标准和评估体系。由于缺乏评估标准，海量预测结果的准确率和可合成性难以确定。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">第三，实验验证能力普遍不足。环境适配性与灵活性较差，当前实验室多为人类操作设计，难以满足AI自主验证需求；跨平台互操作性偏低，存在数据孤岛、设备孤岛等问题；感知与分析衔接不够，自主实验需实现“感知-决策-执行”闭环，但目前这一环节仍存在脱节。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">针对上述难题，王江平提出一些对策和建议。一是加强数据集、高价值知识中心和AI预测结果评估标准体系的建设。他分析，当前重点行业的高精度、长序列、多模态的数据集仍然欠缺，亟须建立公共的高价值数据中心，减少重复工作，并构建权威性的预测结果评估体系。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">二是加快AI自主实验室的建设。王江平认为当前，AI自主实验室建设仍有诸多工作待推进。要倡导开源与模块化发展，降低自主实验室建设门槛。要探索“人在回路中”的混合增强智能，当前完全无人化的“AI科学家”尚难实现，仍需人类参与，因此“人在回路中”的增强智能在现阶段不可或缺。要发展数字孪生与通用知识模型。要探索多智能体协作的“联合科学家”模式。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">三是加强中试平台的建设，发挥我国应用场景的优势，推动工程化的创新。此外，还要推动学术界和产业界合作等。(完)\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">【新闻来源】今日头条 \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.toutiao.com\u002Farticle\u002F7584365822689280538\u002F?upstream_biz=doubao&source=m_redirect\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1RTDQy?ocid=BingHp01&cvid=6936317f054647a2afcd53fafcde084a&ei\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\">（本网转发此文章，旨在为读者提供更多的信息资讯，所涉内容不构成投资、消费建议。文章事实如有疑问，请与有关方核实，文章观点非本网观点，仅供读者参考。）\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>","","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F12\u002F58d3477f62c34b6d93286ab055102de0\u002FAI领域.jpg","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F12\u002Fthumbs\u002F58d3477f62c34b6d93286ab055102de0\u002FAI领域.jpg",0,1,64,"2025-12-19 09:06",2,false,{"id":17,"name":20,"enName":21},"芯位视野","Xinwei Vision","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A2b2fecd2-fba0-417f-9243-a34d06224488%3A0.wav?Expires=1766157242&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=oBKDQg7WkHiVcynfGycaNRfsH%2Bc%3D",5346678,"2b2fecd2-fba0-417f-9243-a34d06224488","2025-12-19 09:03","Wang Jiangping explains how to break through the \"dam lake\" of AI scientific discovery","\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">XinHua News Agency, Beijing, December 16th (Reporter Liu Wenwen) XinHua News Agency held a forum titled \"New Pattern · New Momentum\" on December 16th in Beijing. Focusing on how to solve the \"dam lake\" dilemma of AI scientific discovery, Wang Jiangping, former vice minister of the Ministry of Industry and Information Technology and director of the Electronic Science and Technology Committee of the Ministry of Industry and Information Technology, gave an in-depth analysis at the meeting.\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002F8f2645c2825749e8ae3ed06f0a833e58.png\" width=\"739\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp class=\"ql-align-center\">\u003Cspan style=\"color: rgb(187, 187, 187);\">On December 16th, Wang Jiangping, former vice minister of the Ministry of Industry and Information Technology and director of the Electronic Science and Technology Committee of the Ministry of Industry and Information Technology, participated in the \"Guoshi Forum: 2025 Annual Meeting\" hosted by XinHua News Agency in Beijing. Photo by Jiang Qiming, XinHua News Agency\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">Wang Jiangping said that the current AI prediction results are growing exponentially, but human verification and industrialization capabilities are only increasing linearly, with a huge gap between them. The results predicted by AI in one day would take humans 10 years or even longer to verify. This contradiction is like a \"dam lake,\" blocking the channel for scientific discoveries to be transformed into practical applications, not only leading to massive prediction results failing to be experimentally verified and applied industrially in time, but also consuming a large amount of research resources and computing power resources.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">Why does this \"dam lake\" occur? He analyzed that there are mainly the following reasons.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">First, the prediction models have limitations. For example, insufficient logical reasoning and knowledge depth, the existence of the black box dilemma and hallucination risks, limited goal derivation capability, etc.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">Second, there is a lack of standards and evaluation systems. Due to the lack of evaluation standards, it is difficult to determine the accuracy and synthetizability of the massive prediction results.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">Third, the experimental verification capability is generally insufficient. The adaptability and flexibility of the environment are poor. Current laboratories are mostly designed for human operation, making it difficult to meet the needs of AI autonomous verification; the interoperability across platforms is low, with issues such as data islands and equipment islands; the connection between perception and analysis is inadequate. Autonomous experiments need to achieve a \"perception-decision-execution\" loop, but this link still has disconnection at present.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">In response to these challenges, Wang Jiangping proposed some countermeasures and suggestions. First, strengthen the construction of data sets, high-value knowledge centers, and AI prediction result evaluation standard systems. He analyzed that currently, high-precision, long-sequence, multi-modal data sets in key industries are still lacking. It is urgently needed to establish public high-value data centers to reduce redundant work and build authoritative prediction result evaluation systems.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">Second, accelerate the construction of AI autonomous laboratories. Wang Jiangping believes that there are still many tasks to be promoted in the construction of AI autonomous laboratories. We should advocate open source and modular development to lower the threshold for building autonomous laboratories. We should explore hybrid augmented intelligence with people in the loop. At present, completely automated \"AI scientists\" are difficult to achieve, so human participation is still needed. Therefore, augmented intelligence with people in the loop is indispensable at this stage. We should develop digital twins and general knowledge models. We should explore the \"joint scientist\" model of multi-agent collaboration.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"font-size: 18px;\">Third, strengthen the construction of pilot production platforms, leverage China's application scenario advantages, and promote engineering innovation. In addition, we should also promote cooperation between academia and industry. (end)\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">【News Source】 Toutiao \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.toutiao.com\u002Farticle\u002F7584365822689280538\u002F?upstream_biz=doubao&source=m_redirect\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1RTDQy?ocid=BingHp01&cvid=6936317f054647a2afcd53fafcde084a&ei\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\"> (This article is reprinted by this site to provide readers with more information and news. The content involved does not constitute investment or consumption advice. If there are any questions about the facts of the article, please verify with the relevant parties. The views of the article are not the views of this site, and are for reference only.)\u003C\u002Fspan>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cbr>\u003C\u002Fp>","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A93336842-2359-41b8-ae00-a9f0b63cacdb%3A0.wav?Expires=1774838440&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=HrWa1oktDc1jbXEWo%2FTOOGBgEdQ%3D","93336842-2359-41b8-ae00-a9f0b63cacdb",8133868]