Original title: 准确把握人工智能发展前沿与竞争格局
URL: https://www.qstheory.cn/20260515/a9132f0ae2814f3c88eb44300722d4f8/c.html
Author: Yu Xiaohui (余晓晖)
Affiliation: President and Deputy Party Secretary, China Academy of Information and Communications Technology (CAICT)
Source: Qiushi, Issue 10, 2026
Since the new era, the CPC Central Committee has placed great importance on the development of artificial intelligence. General Secretary Xi Jinping has issued a series of important instructions on AI and has twice chaired Politburo group study sessions to deliver important addresses, stressing that AI is a major driver of the new wave of sci-tech revolution and industrial transformation, and that accelerating the development of new-generation AI is a strategic question about whether China can seize the opportunities of this round of revolution. The 4th Plenum of the 20th Central Committee called for comprehensive implementation of the "AI+" Action, the seizure of commanding heights in AI industrial applications, and the empowering of all industries. Accurately grasping the frontiers and competitive landscape of AI, and promoting high-quality, all-around AI empowerment, is an inherent demand for fostering new quality productive forces and meeting people's needs for a better life, and an imperative for winning strategic initiative and realizing the great rejuvenation of the Chinese nation amid the changes unseen in a century.
进入新时代，党中央高度重视人工智能发展，习近平总书记对人工智能作出一系列重要指示批示，两次主持中央政治局集体学习并发表重要讲话，强调人工智能是新一轮科技革命和产业变革的重要驱动力量，加快发展新一代人工智能是事关我国能否抓住新一轮科技革命和产业变革机遇的战略问题。党的二十届四中全会提出全面实施"人工智能+"行动，要求抢占人工智能产业应用制高点，全方位赋能千行百业。准确把握人工智能发展前沿和竞争格局，推动人工智能高质量发展和全方位赋能，是打造新质生产力、满足人民美好生活需要的内在要求，也是在百年变局中赢得战略主动、实现中华民族伟大复兴的必然之举。
The concept of AI has been around for nearly 70 years and has cycled through logic-reasoning, expert-systems and deep-learning phases. With full-chain breakthroughs in algorithms, data and computing power, AI's capabilities have advanced rapidly and in some fields have reached "human-level" or even "superhuman" performance. This wave of AI has shown unprecedented speed, social penetration and cross-domain catalytic effects, expressed in three ways.
人工智能技术构想由来已久，在近70载的跌宕起伏中，已历经逻辑推理、专家系统、深度学习等重要阶段。缘于算法、数据、算力全链条突破，近年来人工智能各方面能力快速提升，在部分领域达到"类人"甚至"超人"的水平。可以说，此轮人工智能发展在速度、社会渗透力与跨界带动作用上均呈现出前所未有的态势，具体表现为三个方面。
First, AI is evolving from "passive training" to "self-evolution," with iteration speed rising sharply. Early on, model parameters were small and computing-power upgrades slow; iteration cycles were measured in years (e.g., the classic object-detection model had a two-year upgrade gap). Algorithmic optimization and surging compute have driven exponential acceleration. Per CAICT monitoring, since 2024 the iteration cycle of top models has compressed from months to weeks; model parameters have leapt from millions and hundreds of millions to trillions; inference and training have improved continuously. This pace reshapes the competitive landscape. Domestic leading tech firms and research teams can catch up and even leapfrog on key technologies within 1–2 iteration cycles — a feature distinguishing this AI revolution from any prior technological shift.
一是人工智能技术创新正从"被动训练"向"自主进化"方向演进，迭代速度快速提升。在发展初期，人工智能模型参数小、算力更新慢，迭代周期以年为单位，经典的目标检测模型升级跨度为两年。随着算法框架的不断优化和算力的大幅提升，人工智能的迭代速度实现指数级跃升。据中国信息通信研究院监测统计，2024年以来，顶尖模型的迭代周期已从数月压缩至数周，模型参数从百万级、亿级飙升至万亿级，推理与训练能力持续大幅提升。这种极速迭代的节奏，彻底重构了行业竞争格局。国内头部科技企业与科研团队的技术追赶速度显著加快，往往能在1—2个迭代周期内完成关键技术的追平与超越，这正是本轮人工智能革命区别于历史上任何一次技术变革的重要特征。
Second, "intelligence-native" is becoming the new core of "AI+," and empowering applications are translating into productive forces. In the 15th Five-Year Plan period, China will enter a new stage of comprehensive AI empowerment for high-quality development. AI is deeply penetrating economy and society, taking on the role of a "symbiotic partner" — reshaping factor allocation, the industrial division of labor and economic efficiency, and emerging as a new growth pole. In industry, our complete UN industrial categories generate vast customized AI demand, while industrial scenes provide AI a "battlefield"; the "two-way rush" drives industrial upgrading and feeds high-quality data back into AI iteration, accelerating new "intelligent emergence." AI is reconstructing enterprise strategy, structure, processes and offerings in "native" ways, letting a few people manage hundreds of agents and replace teams of dozens or more, giving rise to disruptive modes like the "one-person enterprise" and becoming a new engine of the intelligent economy.
二是智能原生成为"AI+"新内核，赋能应用加快向现实生产力转化。"十五五"时期，我国将进入人工智能全面赋能高质量发展新阶段，人工智能深度渗透经济社会各领域，在生产生活中将扮演"共生伙伴"的重要角色，进而重构生产要素配置模式、重塑产业分工体系、优化经济运行效率，成为我国经济发展的重要增长极。以工业为例，我国拥有联合国产业分类中的所有工业门类，催生大量对人工智能的个性化、场景化应用需求，为人工智能提供了"用武之地"。两者"双向奔赴"，既推动我国工业的转型升级，也通过工业场景的大规模深度应用和高质量数据沉淀反哺人工智能技术的迭代创新，加速人工智能新的"智能涌现"。人工智能正在通过"原生"方式重构企业战略规划、运行架构、业务流程和产品服务，使少数人管理监督成百上千的智能体，完成原来数十人甚至上百人团队的工作，不断催生"一人企业"等颠覆式的新模式新业态，成为智能经济新的增长引擎。
Third, AI is reconstructing human production and life across the board, but its dual-use nature is increasingly evident. From smart glasses doing real-time translation and digital humans giving consultations, to humanoid robots "interning" in factories — assembling, carrying, sorting and inspecting — AI is moving from "chatting" to "working," entering every facet of production and life. From production to consumption, from physical to digital, AI is reshaping production and lifestyles, and profoundly changing social structure, values and ethics. Globally, AI has become a core topic in multilateral and regional mechanisms; the UN, BRICS, ASEAN and SCO have issued declarations or joint statements to build consensus on AI development and governance. At the same time, AI's safety and governance issues are not trivial. The viral agent app "Crayfish" early this year, while offering novel experiences, also exposed risks of unclear permission boundaries, sensitive-information leakage, prompt-injection susceptibility and supply-chain poisoning — and amplified social concerns over job displacement.
三是人工智能技术正在全维度重构人类生产生活，但双刃剑属性日益凸显。从智能眼镜实时翻译、数字人提供咨询服务，到工业领域人形机器人进厂"实习"，从事装配、搬运、分拣、巡检劳动，人工智能正从"会聊天"转向"会干活"，走进生产生活的方方面面。从生产端到消费端，从物理空间到数字空间，人工智能不仅正在重塑生产方式和生活方式，更深刻改变着社会结构、价值观念与伦理准则。放眼世界，人工智能已成为全球与区域多边机制的核心议题，联合国、金砖、东盟、上合等国际组织均通过发布领导人宣言或联合声明，凝聚人工智能发展与治理共识。与此同时，人工智能带来的安全治理问题不容忽视。今年年初风靡全球的智能体应用"小龙虾"，在带来全新体验的同时，也暴露出可能存在的权限边界模糊、敏感信息泄露、易遭受提示词注入以及供应链投毒攻击等问题，并进一步加剧了就业替代等社会性担忧。
Overall, this wave of AI breaks traditional linear development and reveals four inherent rules different from traditional industries, carrying a wholly new logic of technological renewal.
总的来看，此轮人工智能打破了传统线性发展模式，形成了有别于传统产业的四个内在发展规律，承载着技术革新的全新逻辑。
(1) A continuing tension between scale expansion and efficiency breakthroughs. Overall the "scaling law" of high input and high return still holds, but model performance shows diminishing marginal returns with rising input, making efficiency innovation more important. DeepSeek used open-source engineering innovation to achieve top-model performance at low cost. Meanwhile, the cost reductions from efficiency breakthroughs spawn a paradox: mainstream model inference token cost has fallen 99 percent in three years; extreme cost convergence has stimulated broader use cases and higher revenue, producing a positive cycle of technological iteration spurring application diffusion.