中国AI的致命短板与战略转型路径：从追赶者到定义者

破局与定义：论中国AI如何以文化数据与自主创新构建新发展范式

摘要： 埃里克·施密特指出中国AI存在金融支持不足与创新生态薄弱等结构性短板。中国需摒弃单纯追赶思维，立足汉语智慧、系统思维、举国体制及五千年文化数据等独特优势，构建“文化数据与AI融合”“算力自主化”“创新生态重构”及“应用领导与标准制定”四大战略路径。核心在于将文明禀赋转化为定义未来的新范式，推动AI发展从技术竞赛升维为文明竞争，实现从追赶者到定义者的根本转型。

---

中国AI的致命短板与战略转型路径：从追赶者到定义者

埃里克·施密特在GAIC全球人工智能大会上的发言揭示了中美AI发展的结构性差异，直指中国AI发展的致命短板在于金融支持深度不足和创新生态系统薄弱。这一诊断不仅触及表层技术差距，更揭示了中国AI发展深层次的战略思维与制度效率矛盾。在AI大国竞争关乎未来百年国家安全与民族命运的背景下，中国必须正视差距，实事求是，同时发挥自身独特优势，构建"你打你的，我打我的"战略路径，实现从追赶者到定义者的转型。

一、中国AI发展的致命短板

中国AI发展的最大弱点在于金融市场的支持深度不足，以及缺乏美国特有的创新生态系统，这一结构性差异使得中国难以复制美国在AI领域的发展模式。施密特指出，"中国金融市场缺乏足够深度来资助像美国企业家正在做的这些疯狂项目。你不可能随便在中国筹集500亿美元去建一座数据中心"，这反映了中国在AI领域融资能力的局限。

从金融支持深度看，中国AI行业融资呈现"头部垄断+产业资本驱动"特征，单笔融资规模普遍较小。2025年Q3中国AI行业一级市场融资额达370亿元，但最大仅15亿元人民币（如曦智科技），远低于美国500亿美元级数据中心项目规模。全球对比显示，2024年中国AI风险投资仅76亿美元，而美国达808亿美元，差距显著。中国AI融资高度集中于应用层赛道（大模型、机器人、智能驾驶），基础层（芯片、算法框架）融资占比不足20%，存在"倒金字塔"结构风险。

从创新生态系统看，中国AI发展陷入"应用创新陷阱"。七成企业数字化投入不及年销售额的3%，大量AI初创企业缺乏用户基础限制了其开发大模型并进行现有产品改良和模型迭代的能力 。中国虽在应用层领先，但基础研究（如算法框架、芯片）仍受制于人，且存在"大公司主导"倾向，而非0到1的初创团队主导。美国特有的"企业家疯狂投入"文化构成了"美国例外论"的核心，而中国虽拥有"举国体制办大事"优势，但未能与市场活力有效结合，形成了"政产学研用"协同创新机制的断层。

从制度效率看，中国虽在AI领域投入巨大，但存在形式主义、短期逐利等问题。相比之下，欧洲虽拥有顶尖人才但经济体系无法促成突破，民主制度国家普遍存在发展停滞现象。施密特指出，"我花了15年时间试图推动欧洲有所作为。欧洲有很多极其聪明的人，但它们的经济体系并没有促成这种发展"，这反映了制度效率在AI发展中的关键作用。

二、中国AI发展的独特优势

中国拥有汉语智慧优势、系统思维优势、举国体制办大事优势以及五千年文化智慧AI训练大数据优势，这些是西方无法轻易复制的禀赋。中国AI战略不应在旧范式里做"更好的追赶者"，而应用来创造"不同的未来"。

汉语语料库与系统思维方面，中国中文大模型参数规模已达到国际先进水平（如文心一言3.0参数2600亿，星火大模型1700亿，盘古大模型1000亿） ，但训练数据仍以通用互联网文本为主，未充分整合五千年文化智慧。CLUECorpus2020提供了100G中文语料（350亿字符），但规模与西方千亿级语料库仍有差距，且多为学术项目，未体现"汉语智慧"在技术上的独特突破。中文大模型在多模态（图文音三模态OPT模型）和轻量化（"神农""孟子"）方向有突破，但未形成文化特色技术壁垒 。

文化数据库建设方面，《"十四五"文化发展规划》提出建设国家文化大数据体系，通过"四端一网两翼"架构整合文化遗产标本库、文化基因库等资源，但尚未形成标准化、可直接用于AI训练的国家级文化数据库 。"古籍图典资源库"已上线5万张古籍图像资源，计划扩展至10万张，但需进一步验证是否与AI企业合作开发文化语料库 。文化遗产档案智慧数据资源建设研究显示，中国正积极探索文化数据分类、编目、标签化等技术手段，但尚未形成统一标准 。

举国体制优势方面，"东数西算"工程通过政策推动算力网络建设，但具体成效（如算力利用率提升、基础研究支持）需进一步验证。2025年3月，《AI+2030行动计划》正式落地，中央财政设立1.2万亿元专项资金，重点支持光子芯片、存算一体芯片等前沿领域。国家大基金三期已向第三代半导体材料注资超2000亿元，推动产业链自主化进程。南宁、苏州、深圳等地通过专项基金、税收优惠、应用场景开放等政策支持AI发展，但缺乏国家层面的统一战略文件。

三、构建中国AI新范式的战略路径

中国AI的真正挑战不在于复制一个"美国模式"，而在于能否将自身独特的文明禀赋与制度优势，转化为一种能够定义未来、引领范式的新战略。基于中国独特优势，可构建以下战略路径：

1. 文化数据与AI融合战略

构建国家级文化语料库，将五千年文化智慧转化为AI训练优势。具体措施包括：

• 加快国家文化大数据体系建设，整合文化遗产标本库、中华民族文化基因库、中华文化素材库"三库"资源 ，建立统一的文化数据标准与共享机制 。
• 对企业、高校、科研机构建设包括语料库在内的高质量数据集给予最高100万元奖励 ，推动文化数据转化为AI训练资源。
• 鼓励AI企业与文化机构合作开发文化语料库，如古籍大模型（SikuGPT、GuwenBERT）的训练与应用 ，形成"文化数据—AI模型—场景应用"闭环。
• 推动文化大数据认证体系建设，建立信用认证、内容认证和服务认证三大引擎 ，确保文化数据的真实性、可靠性与安全性。

2. 算力自主化与光子芯片协同战略

利用中国举国体制优势，加速算力基础设施建设与光子芯片研发，实现算力自主化。具体措施包括：

• 深化"东数西算"工程，通过算力调度优化和绿色技术降低算力成本，支持AI基础研究（芯片、算法框架） 。
• 推广南宁"算力券"补贴（结算金额的30%） 和苏州百亿AI芯片母基金模式 ，设立国家级"AI+文化"专项基金，支持早期团队与垂直场景结合。
• 加强光子芯片研发，支持上海交大无锡光子芯片研究院的LightSeek大模型开放接口，降低光子芯片研发门槛 。
• 推动国产AI芯片应用，对采购国产AI芯片的企业给予30%设备补贴，要求新建数据中心国产算力占比不低于50% 。

3. 创新生态重构战略

打破"大公司主导"倾向，支持0到1创新团队，构建健康的创新生态系统。具体措施包括：

• 设立"AI国家实验室"，鼓励高校、科研机构和企业联合攻关，集中力量攻克"卡脖子"问题 。
• 打造"开源基金"，激励开发者贡献，建立软件生态圈，降低AI技术应用门槛 。
• 推动风险投资与产业资本协同，对AI初创企业提供税收优惠（如减免前几年的企业所得税或增值税）。
• 延长创新创业类基金存续期至15年，采用差异化考核机制，容忍长期投入风险 。

4. 应用领导与标准制定战略

依托中国丰富的应用场景，推动AI技术在各行业的深度融合，同时积极参与国际标准制定。具体措施包括：

• 推动AI在国央企场景（能源、制造、交通等）的示范应用，推广"AI Agent+RPA"模式，实现安全可控的流程自动化。
• 在政务、医疗、教育等领域开放应用场景，滚动开放不少于50个应用场景，推动AI技术落地。
• 参与全球人工智能治理倡议，提出"智能向善"原则，推动中国主导的AI伦理与文化数据使用标准 。
• 推动中国AI技术标准国际化，如大模型评估框架、多模态数据协议等，增强中国在AI领域的国际话语权。

四、从追赶者到定义者的转型关键

实现中国AI从追赶者到定义者的转型，关键在于打破"应用驱动"思维，转向"基础创新"与"文化引领"相结合的新范式。

首先，必须摒弃"天天说AI快追赶到美国了"的盲目乐观，正视差距，承认施密特所说差距，实事求是。中国AI在算力、算法、数据、人才等方面仍存在差距，特别是基础研究薄弱，全球高影响力AI论文占比低，美国占5项，中国仅2项。中国AI企业融资规模远低于美国，且资本集中于应用层，基础层投入严重依赖政策性资金，难以形成持续创新的资本生态。

其次，必须放弃逐利短视思维，政客必须放弃狭隘自私、形式主义、掩耳盗铃、不懂装懂、到处揽功劳，动辄拿定力说成战略可笑叙事。政府、企业、资本圈要敢于在AI领域砸钱，否则以后想砸也没机会了！中国必须放弃"成为另一个美国"的思维，而是"创造一个不同的未来"——一个融合了五千年文明智慧的AI新纪元。

第三，必须聚焦0到1创新团队，而非寄托于或押宝于大公司大厂身上。大公司大厂往往追求短期利益和市场回报，难以承担长期、高风险的基础研究任务。中国应设立专项基金支持0到1团队在汉语系统性思维、文化数据融合等方向的探索，鼓励"算法-硬件"协同创新。

最后，必须将汉语智慧、系统思维、举国体制优势、中国独有的五千年文化智慧AI训练大数据，转化为能够定义未来、引领范式的新战略。中国AI不应在"旧范式里做'更好的追赶者'"，而应利用汉语的系统性思维（如整体性、辩证逻辑）开发独特架构，如多模态文化表达模型、轻量化框架等，同时依托举国体制优势，整合算力资源，支持AI基础研究。

五、中国AI新范式的实践路径

基于中国独特优势，可构建以下实践路径：

1. 建立"国家文化AI实验室"

整合中科院、高校与AI企业资源，建立国家级文化AI实验室，负责文化数据的采集、清洗、标注、关联和解构工作 ，形成标准化的文化语料库。实验室应具备以下功能：

• 数据整合：整合古籍、传统艺术、历史文献等文化资源，形成统一的文化数据标准与共享机制。
• 模型开发：开发融合汉语系统性思维的AI模型，如多模态文化表达模型、古籍理解模型等。
• 技术转化：将文化数据与AI技术结合，开发文化IP生成、文物修复、文化传播等应用场景。
• 人才培养：培养跨学科的AI+文化复合型人才，为文化AI发展提供智力支持。

2. 推动"文化数据—AI模型—场景应用"闭环

依托国家文化大数据体系，构建"文化数据—AI模型—场景应用"闭环，实现文化数据的价值转化：

• 文化数据采集与处理：建立全国文化专网，实现文化资源数据的采集、清洗、标注、关联和解构 。
• AI模型训练与优化：利用文化数据训练AI模型，开发文化IP生成、文物修复、文化传播等应用场景 。
• 场景应用与反馈：在国央企、政府、文化机构等场景中应用AI模型，收集反馈数据，优化模型性能。
• 数据共享与生态建设：建立文化数据共享机制，推动文化数据在AI领域的应用，构建文化AI生态体系。

3. 加强光子芯片与AI的协同创新

利用光子芯片的高速低能耗特性，推动光子芯片与AI的协同创新，实现算力自主化：

• 光子芯片研发支持：设立专项基金支持光子芯片研发，如上海交大无锡光子芯片研究院的LightSeek大模型。
• AI优化光子芯片研发：推广LightSeek等AI工具开放接口，降低光子芯片研发门槛，加速算力自主化。
• 光子芯片与AI融合应用：在数据中心、智能终端等领域应用光子芯片，提升AI算力效率与安全性。
• 产学研协同创新：建立光子芯片与AI融合的产学研协同创新机制，推动技术成果转化。

4. 构建"AI+文化"创新生态

依托中国丰富的文化资源和应用场景，构建"AI+文化"创新生态，推动文化与科技深度融合：

• 应用场景开放：在政务、医疗、教育等领域开放应用场景，推动AI技术落地 。
• 文化IP开发：利用AI技术开发文化IP，如数字人、虚拟偶像、文化产品等，提升文化影响力 。
• 文化传播创新：利用AI技术创新文化传播方式，如AIGC、元宇宙等，扩大文化覆盖面 。
• 国际合作交流：加强与国际组织、企业、研究机构的合作交流，推动中国AI文化成果走向世界。

六、中国AI新范式的政策保障

实现中国AI新范式的转型，需要强有力的政策保障：

1. 财税金融支持政策

• 对企业、高校、科研机构建设包括语料库在内的高质量数据集给予最高100万元奖励。
• 对通过公共数据授权运营平台开展数据融合应用，形成高质量数据产品和优质数据服务的，给予最高100万元奖励。
• 设立总规模50亿元的人工智能产业基金，重点支持初创期、成长型人工智能企业和项目。
• 对符合中国—东盟产业合作区等现行相关税收优惠政策的企业，可自取得第一笔生产经营收入所属年度起，第1年至第5年免征、第6年至第10年减半征收企业所得税地方分享部分。
• 对科技型规上人工智能企业新招用的高端科技人才，按税后薪酬比例给予薪酬奖励。
• 研发费用加计扣除比例提高至100%-200%，形成无形资产的按成本200%在税前摊销。

2. 人才队伍建设政策

• 实施人工智能领域人才专项行动，发挥人才链、教育链、产业链、创新链"四链"融合机制作用。
• 支持行业企业联合高校、科研院所开展人才共引共用、产教融合协同育人、关键技术联合攻关等行动。
• 对掌握关键核心技术，能够带来重大影响、重大突破的海内外人才，给予3000万元至1亿元项目资助和300万元至1000万元购房补贴 。

3. 开放交流合作政策

• 加强对国内外龙头企业招引合作，落地建设一批研发机构、区域总部和生产基地 。
• 支持企业优化国际营销服务体系，提高产品竞争力和附加值，全力开拓外部市场。
• 支持举办国际虚拟现实创新大会、消费电子博览会等高水平产业对接交流活动和创业创新赛事。
• 参与全球人工智能治理倡议，提出"智能向善"原则，推动中国主导的AI伦理与文化数据使用标准 。

七、结语：AI竞争是"文明竞争"，而非"技术竞赛"

埃里克·施密特的警示并非危言耸听，而是中国AI战略必须正视的现实。中国AI的真正出路，不在于"成为另一个美国"，而在于"创造一个不同的未来"——一个融合了五千年文明智慧的AI新纪元。这需要政企学研协同发力，需要国家战略定力，更需要一种"敢为天下先"的文明自信。

中国拥有汉语智慧、系统思维、举国体制优势和五千年文化智慧AI训练大数据，这些是西方无法轻易复制的禀赋。中国AI战略应当立足这些独特优势，构建不同于美国的新范式，实现从追赶者到定义者的转型。只有这样，中国才能在AI大国竞争中占据主动，为未来百年国家安全与民族命运奠定坚实基础。

未来的AI竞争，不仅是算法与芯片的较量，更是国家战略智慧与制度韧性的终极比拼。中国必须摒弃猛抓西方中心论所谓权威数据进行AI训练的方向性错误，充分利用汉语智慧和五千年文化数据库，构建具有中国特色的AI发展道路，牢牢掌握人工智能发展和治理主动权 。只有这样，中国才能在AI时代实现真正的引领，为人类文明进步贡献中国智慧和中国方案。

说明：报告内容仅供参考。

---

Breaking Through and Defining: On How China's AI Constructs a New Development Paradigm with Cultural Data and Independent Innovation

Abstract: Eric Schmidt has pointed out that China's AI sector faces structural shortcomings such as insufficient financial support and a weak innovation ecosystem. China must abandon the mindset of mere catch-up and leverage its unique advantages—including Chinese linguistic wisdom, systems thinking, the whole-of-nation system, and five millennia of cultural data—to build four strategic pathways: "Integration of Cultural Data and AI," "Computing Power Autonomy," "Innovation Ecosystem Reconstruction," and "Application Leadership and Standard Setting." The core lies in transforming civilizational endowments into a new paradigm that defines the future, elevating AI development from a technological race to a civilizational competition, and achieving a fundamental shift from a follower to a definer.

The Fatal Shortcomings and Strategic Transformation Pathways of China's AI: From Follower to Definer

Eric Schmidt's remarks at the GAIC Global Artificial Intelligence Conference revealed the structural differences in AI development between China and the United States, pointing directly to China's fatal shortcomings: insufficient depth of financial support and a weak innovation ecosystem. This diagnosis not only touches on surface-level technological gaps but also exposes deeper contradictions in China's AI development strategy, particularly in strategic thinking and institutional efficiency. Against the backdrop of AI great-power competition, which concerns national security and ethnic destiny for the next century, China must face its gaps squarely, seek truth from facts, leverage its unique advantages, and construct a strategic path of "You fight your way, I fight mine" to transform from a follower to a definer.

I. The Fatal Shortcomings of China's AI Development

The greatest weaknesses of China's AI development lie in the insufficient depth of financial market support and the lack of an innovation ecosystem unique to the United States. This structural difference makes it difficult for China to replicate the U.S. development model in AI. Schmidt noted, "The Chinese financial market lacks sufficient depth to fund the crazy projects that American entrepreneurs are doing. You can't just raise $50 billion in China to build a data center," reflecting the limitations of China's financing capacity in AI.

From the Perspective of Financial Support Depth: China's AI industry financing is characterized by "head monopoly + industrial capital drive," with generally small single financing amounts. In Q3 2025, China's AI industry raised 37 billion yuan in the primary market, but the largest single deal was only 1.5 billion yuan (e.g., Lightelligence), far below the $50 billion scale of U.S. data center projects. Global comparisons show that in 2024, China's AI venture capital was only $7.6 billion, while the U.S. reached $80.8 billion, a significant gap. China's AI financing is highly concentrated in application-layer tracks (large models, robotics, autonomous driving), with basic-layer (chips, algorithm frameworks) financing accounting for less than 20%, posing a risk of an "inverted pyramid" structure.

From the Perspective of Innovation Ecosystem: China's AI development has fallen into an "application innovation trap." Seventy percent of enterprises invest less than 3% of annual sales in digitalization, and many AI startups lack a user base, limiting their ability to develop large models and iterate on existing products. Although China leads in the application layer, basic research (e.g., algorithm frameworks, chips) is still constrained by others, and there is a tendency toward "large company dominance" rather than leadership by 0-to-1 startup teams. The unique U.S. culture of "entrepreneurial crazy investment" constitutes the core of "American exceptionalism," while China, despite its advantage of the "whole-of-nation system to accomplish great things," has failed to effectively combine it with market vitality, resulting in a disconnect in the "government-industry-university-research-application" collaborative innovation mechanism.

From the Perspective of Institutional Efficiency: Although China has invested heavily in AI, there are issues such as formalism and short-term profit-seeking. In contrast, Europe, while possessing top talent, has an economic system that fails to foster breakthroughs, and democratic countries generally experience development stagnation. Schmidt noted, "I spent 15 years trying to get Europe to do something. Europe has many extremely smart people, but their economic system hasn't facilitated this kind of development," reflecting the critical role of institutional efficiency in AI development.

II. The Unique Advantages of China's AI Development

China possesses advantages in Chinese linguistic wisdom, systems thinking, the whole-of-nation system to accomplish great things, and five millennia of cultural wisdom as AI training big data—endowments that the West cannot easily replicate. China's AI strategy should not aim to be a "better follower" in the old paradigm but should instead be used to create a "different future."

In Terms of Chinese Language Corpus and Systems Thinking: The parameter scale of China's Chinese large models has reached international advanced levels (e.g., Wenxin Yiyan 3.0 with 260 billion parameters, Xinghuo with 170 billion, Pangu with 100 billion), but training data still mainly consists of general internet text, without fully integrating five millennia of cultural wisdom. CLUECorpus2020 provides 100G of Chinese corpus (35 billion characters), but its scale still lags behind Western trillion-level corpora, and it is mostly an academic project that does not reflect unique technological breakthroughs in "Chinese linguistic wisdom." Chinese large models have made breakthroughs in multimodality (OPT model for text-image-audio) and lightweighting ("Shennong," "Mencius"), but have not formed a culturally distinctive technological barrier.

In Terms of Cultural Database Construction: The "14th Five-Year Plan for Cultural Development" proposes building a national cultural big data system, integrating resources such as cultural heritage specimen libraries and cultural gene banks through a "four terminals, one network, two wings" architecture, but a standardized national cultural database directly usable for AI training has not yet been formed. The "Ancient Books and Classics Resource Library" has launched 50,000 ancient book image resources, with plans to expand to 100,000, but further verification is needed on whether to cooperate with AI enterprises to develop cultural corpora. Research on the construction of intelligent data resources for cultural heritage archives shows that China is actively exploring technical means such as cultural data classification, cataloging, and tagging, but a unified standard has not yet been formed.

In Terms of Whole-of-Nation System Advantages: The "East Data West Computing" project promotes computing power network construction through policies, but specific results (e.g., improved computing power utilization, support for basic research) need further verification. In March 2025, the "AI+2030 Action Plan" was officially launched, with the central government allocating 1.2 trillion yuan in special funds to focus on cutting-edge fields such as photonic chips and in-memory computing chips. The National Integrated Circuit Industry Investment Fund Phase III has injected over 200 billion yuan into third-generation semiconductor materials, promoting the industrial chain autonomy process. Nanning, Suzhou, Shenzhen, and other places support AI development through policies such as special funds, tax incentives, and open application scenarios, but there is a lack of a unified national strategic document.

III. Strategic Pathways to Construct a New Paradigm for China's AI

The real challenge for China's AI is not to replicate an "American model" but to transform its unique civilizational endowments and institutional advantages into a new strategy that can define the future and lead the paradigm. Based on China's unique advantages, the following strategic pathways can be constructed:

1. Integration of Cultural Data and AI StrategyBuild a national cultural corpus to transform five millennia of cultural wisdom into AI training advantages. Specific measures include:

• Accelerate the construction of the national cultural big data system, integrate resources from the "three libraries" (cultural heritage specimen library, Chinese nation cultural gene bank, Chinese cultural material library), and establish a unified cultural data standard and sharing mechanism.
• Award up to 1 million yuan to enterprises, universities, and research institutions for building high-quality datasets including corpora, promoting the transformation of cultural data into AI training resources.
• Encourage AI enterprises to cooperate with cultural institutions to develop cultural corpora, such as the training and application of ancient book large models (SikuGPT, GuwenBERT), forming a closed loop of "cultural data—AI models—scenario applications."
• Promote the construction of a cultural big data certification system, establishing three engines: credit certification, content certification, and service certification, to ensure the authenticity, reliability, and security of cultural data.

2. Computing Power Autonomy and Photonic Chip Synergy StrategyLeverage China's whole-of-nation system advantages to accelerate computing power infrastructure construction and photonic chip R&D, achieving computing power autonomy. Specific measures include:

• Deepen the "East Data West Computing" project, reduce computing power costs through computing power scheduling optimization and green technologies, and support AI basic research (chips, algorithm frameworks).
• Promote Nanning's "computing power vouchers" subsidy (30% of settlement amount) and Suzhou's 10 billion yuan AI chip mother fund model, and establish a national "AI + Culture" special fund to support early-stage teams in combining with vertical scenarios.
• Strengthen photonic chip R&D, support the open interface of the LightSeek large model from the Shanghai Jiao Tong University Wuxi Photonic Chip Research Institute, and lower the threshold for photonic chip R&D.
• Promote the application of domestic AI chips, provide a 30% equipment subsidy to enterprises purchasing domestic AI chips, and require that the proportion of domestic computing power in newly built data centers is not less than 50%.

3. Innovation Ecosystem Reconstruction StrategyBreak the tendency of "large company dominance," support 0-to-1 innovation teams, and build a healthy innovation ecosystem. Specific measures include:

• Establish "AI National Laboratories," encourage universities, research institutions, and enterprises to jointly tackle key problems, and concentrate efforts on overcoming "bottleneck" issues.
• Create an "open-source fund" to incentivize developer contributions, establish a software ecosystem, and lower the threshold for AI technology application.
• Promote synergy between venture capital and industrial capital, and provide tax incentives to AI startups (e.g., exemption or reduction of enterprise income tax or value-added tax in the first few years).
• Extend the duration of innovation and entrepreneurship funds to 15 years, adopt a differentiated assessment mechanism, and tolerate long-term investment risks.

4. Application Leadership and Standard Setting StrategyRely on China's rich application scenarios to promote the deep integration of AI technology in various industries, while actively participating in international standard setting. Specific measures include:

• Promote the demonstration application of AI in state-owned enterprise scenarios (energy, manufacturing, transportation, etc.), popularize the "AI Agent + RPA" model, and achieve safe and controllable process automation.
• Open application scenarios in government affairs, healthcare, education, and other fields, rolling out no less than 50 application scenarios to promote AI technology landing.
• Participate in the Global Initiative on AI Governance, propose the principle of "Intelligence for Good," and promote China-led AI ethics and cultural data usage standards.
• Promote the internationalization of China's AI technical standards, such as large model evaluation frameworks and multimodal data protocols, to enhance China's international voice in AI.

IV. Key to Transformation from Follower to DefinerTo achieve China's AI transformation from a follower to a definer, the key is to break the "application-driven" mindset and shift to a new paradigm combining "basic innovation" and "cultural leadership."

First, we must abandon the blind optimism of "saying every day that AI is catching up to the United States," face the gaps squarely, acknowledge the gaps mentioned by Schmidt, and seek truth from facts. China's AI still lags in computing power, algorithms, data, talent, etc., especially in weak basic research. The proportion of globally high-impact AI papers is low—5 for the U.S., only 2 for China. The financing scale of Chinese AI enterprises is far lower than that of the U.S., and capital is concentrated in the application layer. Basic layer investment relies heavily on policy funds, making it difficult to form a sustainable innovation capital ecosystem.

Second, we must abandon profit-seeking and short-sighted thinking. Politicians must abandon narrow self-interest, formalism, self-deception, pretending to know, taking credit everywhere, and the ridiculous narrative of equating perseverance with strategy. The government, enterprises, and capital circles must dare to invest heavily in AI; otherwise, there will be no chance to invest in the future! China must abandon the mindset of "becoming another America" and instead "create a different future"—a new AI era integrating five millennia of civilizational wisdom.

Third, we must focus on 0-to-1 innovation teams rather than relying on or betting on large companies. Large companies often pursue short-term interests and market returns, making it difficult for them to undertake long-term, high-risk basic research tasks. China should establish special funds to support 0-to-1 teams in exploring Chinese systemic thinking and cultural data integration, and encourage "algorithm-hardware" collaborative innovation.

Finally, we must transform Chinese linguistic wisdom, systems thinking, the advantages of the whole-of-nation system, and China's unique five millennia of cultural wisdom as AI training big data into a new strategy that can define the future and lead the paradigm. China's AI should not aim to be a "better follower in the old paradigm" but should use the systemic thinking of Chinese (e.g., holism, dialectical logic) to develop unique architectures, such as multimodal cultural expression models and lightweight frameworks. At the same time, relying on the advantages of the whole-of-nation system, it should integrate computing power resources and support AI basic research.

V. Practical Pathways for China's New AI ParadigmBased on China's unique advantages, the following practical pathways can be constructed:

1. Establish a "National Cultural AI Laboratory"Integrate resources from the Chinese Academy of Sciences, universities, and AI enterprises to establish a national cultural AI laboratory responsible for the collection, cleaning, annotation, association, and deconstruction of cultural data, forming a standardized cultural corpus. The laboratory should have the following functions:

• Data Integration: Integrate cultural resources such as ancient books, traditional art, and historical documents to form a unified cultural data standard and sharing mechanism.
• Model Development: Develop AI models integrating Chinese systemic thinking, such as multimodal cultural expression models and ancient book understanding models.
• Technology Transformation: Combine cultural data with AI technology to develop application scenarios such as cultural IP generation, cultural relic restoration, and cultural communication.
• Talent Training: Cultivate interdisciplinary AI + culture compound talents to provide intellectual support for cultural AI development.

2. Promote a Closed Loop of "Cultural Data—AI Models—Scenario Applications"Rely on the national cultural big data system to build a closed loop of "cultural data—AI models—scenario applications" to realize the value transformation of cultural data:

• Cultural Data Collection and Processing: Establish a national cultural private network to realize the collection, cleaning, annotation, association, and deconstruction of cultural resource data.
• AI Model Training and Optimization: Use cultural data to train AI models and develop application scenarios such as cultural IP generation, cultural relic restoration, and cultural communication.
• Scenario Application and Feedback: Apply AI models in scenarios such as state-owned enterprises, government, and cultural institutions, collect feedback data, and optimize model performance.
• Data Sharing and Ecosystem Construction: Establish a cultural data sharing mechanism, promote the application of cultural data in AI, and build a cultural AI ecosystem.

3. Strengthen Collaborative Innovation between Photonic Chips and AILeverage the high-speed and low-energy consumption characteristics of photonic chips to promote collaborative innovation between photonic chips and AI, achieving computing power autonomy:

• Photonic Chip R&D Support: Establish special funds to support photonic chip R&D, such as the LightSeek large model from the Shanghai Jiao Tong University Wuxi Photonic Chip Research Institute.
• AI-Optimized Photonic Chip R&D: Promote open interfaces for AI tools such as LightSeek to lower the threshold for photonic chip R&D and accelerate computing power autonomy.
• Integration Application of Photonic Chips and AI: Apply photonic chips in data centers, smart terminals, and other fields to improve AI computing power efficiency and security.
• Industry-University-Research Collaborative Innovation: Establish an industry-university-research collaborative innovation mechanism for the integration of photonic chips and AI to promote the transformation of technological achievements.

4. Build an "AI + Culture" Innovation EcosystemRely on China's rich cultural resources and application scenarios to build an "AI + Culture" innovation ecosystem, promoting the deep integration of culture and technology:

• Open Application Scenarios: Open application scenarios in government affairs, healthcare, education, and other fields to promote AI technology landing.
• Cultural IP Development: Use AI technology to develop cultural IP, such as digital humans, virtual idols, and cultural products, to enhance cultural influence.
• Cultural Communication Innovation: Use AI technology to innovate cultural communication methods, such as AIGC and metaverse, to expand cultural coverage.
• International Cooperation and Exchange: Strengthen cooperation and exchange with international organizations, enterprises, and research institutions to promote the global dissemination of China's AI cultural achievements.

VI. Policy Guarantees for China's New AI ParadigmAchieving the transformation of China's new AI paradigm requires strong policy guarantees:

1. Fiscal and Financial Support Policies

• Award up to 1 million yuan to enterprises, universities, and research institutions for building high-quality datasets including corpora.
• Award up to 1 million yuan to those who carry out data fusion applications through public data authorized operation platforms to form high-quality data products and services.
• Establish an artificial intelligence industry fund with a total scale of 5 billion yuan, focusing on supporting early-stage and growth-stage artificial intelligence enterprises and projects.
• Enterprises that meet the current relevant tax preferential policies such as the China-ASEAN Industrial Cooperation Zone may be exempt from enterprise income tax 地方分享部分 for the first to fifth years and halved for the sixth to tenth years starting from the year in which they obtain their first production and operation income.
• Provide salary rewards to high-end scientific and technological talents newly recruited by science and technology-based 规上 artificial intelligence enterprises according to the proportion of after-tax salaries.
• Increase the R&D expense additional deduction ratio to 100%-200%, and amortize intangible assets at 200% of cost before tax.

2. Talent Team Construction Policies

• Implement a special talent action plan in the field of artificial intelligence, giving play to the role of the "four chains" integration mechanism of talent chain, education chain, industrial chain, and innovation chain.
• Support industry enterprises to jointly carry out talent co-recruitment and sharing, industry-education integration collaborative education, and key technology joint research with universities and research institutes.
• Provide 30 million to 100 million yuan in project funding and 3 million to 10 million yuan in housing subsidies to domestic and foreign talents who master key core technologies and can bring significant impact and breakthroughs.

3. Open Exchange and Cooperation Policies

• Strengthen the recruitment and cooperation of leading domestic and foreign enterprises, and land a number of R&D institutions, regional headquarters, and production bases.
• Support enterprises to optimize their international marketing service systems, improve product competitiveness and added value, and fully explore external markets.
• Support the holding of high-level industry docking and exchange activities and entrepreneurship and innovation competitions such as the International Virtual Reality Innovation Conference and the Consumer Electronics Expo.
• Participate in the Global Initiative on AI Governance, propose the principle of "Intelligence for Good," and promote China-led AI ethics and cultural data usage standards.

VII. Conclusion: AI Competition is a "Civilizational Competition," Not a "Technological Race"

Eric Schmidt's warning is not alarmist but a reality that China's AI strategy must face squarely. The real way out for China's AI is not to "become another America" but to "create a different future"—a new AI era integrating five millennia of civilizational wisdom. This requires collaborative efforts from government, industry, academia, and research, national strategic perseverance, and, more importantly, a civilizational confidence to "dare to be the first in the world."

China possesses Chinese linguistic wisdom, systems thinking, the advantages of the whole-of-nation system, and five millennia of cultural wisdom as AI training big data—endowments that the West cannot easily replicate. China's AI strategy should be based on these unique advantages to build a new paradigm different from the United States and achieve transformation from a follower to a definer. Only in this way can China take the initiative in AI great-power competition and lay a solid foundation for national security and ethnic destiny in the next century.

The future AI competition is not only a contest of algorithms and chips but also the ultimate competition of national strategic wisdom and institutional resilience. China must abandon the directional error of frantically grasping so-called authoritative data from Western centrism for AI training, fully utilize Chinese linguistic wisdom and the five-millennium cultural database, build a path of AI development with Chinese characteristics, and firmly grasp the initiative in artificial intelligence development and governance. Only in this way can China achieve true leadership in the AI era and contribute Chinese wisdom and solutions to the progress of human civilization.

Note: The content of the report is for reference only.