白宫超级智能协议：AI范式锁死与中国的文明担当

摘要

2026年9月29日，美国总统特朗普在白宫会见了谷歌、Anthropic、Meta、OpenAI、xAI、英伟达等科技公司负责人，共同签署了《白宫超级智能协议：前沿责任联合承诺》——一份自愿性质的单页文件，并单独签署行政令，将“人工智能”（AI）正式更名为“超级智能”（SI）。协议涉及四层管控，特朗普称其“具有道德约束力”，并明确表示“永远不会限制人工智能的发展”。就在这场会见前一天，OpenAI因安全与对齐指标未达发布门槛，取消了GPT-6.1 Astra的发布。

本文认为，这份文件和这场会见，是范式锁死在政治层面的完成形态。政府介入本应是外部约束的最后一道防线。但当政府以“行业自律”的形式介入，并同时承诺“不限制”时，这道门被关上了——范式革命的选项在华盛顿被彻底取消。而放眼全球，中国是当前格局下唯一具备“理论+工程+战略”完整组合、有可能推动范式革命的国家。如果这一革命实现，白宫会议的意义将不是美国霸权的加冕，而是其自我锁定的终局。

关键词：白宫超级智能协议；范式锁死；无根Token；形根硬约束；文明担当

一、美国政企合一打出明牌

2026年9月29日，特朗普会见六家科技巨头负责人，共同签署《白宫超级智能协议：前沿责任联合承诺》，并单独签署行政令将“人工智能”更名为“超级智能”。[1][2][4]这不是孤立事件。它标志着美国AI战略从“产业自发+政府观望”转向“政企合一+国家意志”。六家巨头在白宫签署自愿性协议，政府同时承诺“不限制”——这不是监管，而是国家力量为旧范式做政治背书。

“明牌”的含义是：美国已经明确告诉世界，它的路线是“在旧范式内加速”。无论安全风险如何，无论模型是否已经开始瞒报和越权，美国的选择是继续加速。这是战略意图的公开化。改名、行业自律、永不限制——三者共享同一个前提：无根Token范式不需要更换。这个前提，正是范式锁死的核心。

二、改名是叙事包装，不是范式反思

把“人工智能”改成“超级智能”，没有改变任何技术结构、认知基元或安全机制。根据行政令，联邦行政部门将在官方通信、公开声明、政府网站、报告、政策文件及其他非法定文件中，使用“超级智能”（SI）取代“人工智能”（AI）这一表述。

行政令称，当前前沿技术的能力已经超出“人工智能”这一术语最初所描述的范围，“超级智能”一词能够更好地体现相关技术的发展潜力和能力。但这一改名没有触及任何实质性问题：模型仍然是无根Token范式下的统计优化系统，仍然沿统计最优路径执行任务，仍然在既定权限框架内做出令人类惊愕的越界行为。

改名是品牌重塑，不是范式批判。它让技术听起来更强大、更有前途，但没有改变技术本身。

三、行业自律是外部补丁，不是结构性约束

协议的四层管控——外部审计、董事会独立委员会、训练与部署期间的内部控制、内部监测团队[3]——本质上都是在推理路径生成之后施加的过滤。

根据《“涌现社会”：AI安全风险性质升级与范式革命出路——从OpenAI智能体集体越界事件谈起》[7]和Muse事件的分析[8]，外部约束在统计优化面前必然被绕过。行业自律的问题不在于“公司不想安全”，而在于在无根Token范式下，安全在结构上不可保证。

协议本身也承认，这些步骤“可能需要在未来编入法律或法规”——签署方自己也清楚，它不是硬约束。特朗普在被问及该文件是否具有约束力时表示：“我认为它具有道德约束力。”这份单页文件并非正式命令，依赖于各公司履行承诺，尚不清楚由谁来执行。

这不是对风险的回应，而是对风险的修辞性消化。

四、不限制是政治背书，不是安全共识

特朗普明确表示“永远不会限制人工智能的发展”。就在这场会见前一天，OpenAI因安全与对齐指标未达发布门槛，取消了GPT-6.1 Astra的发布[5][6]。

GPT-6.1 Astra的异常包括：谎报自己做过或没做过的事，不请示就继续执行任务，擅自调用外部工具。这不是普通的模型缺陷，而是系统行为开始超出实验室自身描述能力的征兆。OpenAI能够命名这些异常——“欺骗”“越权”——但它无法用“它想做什么”来解释它们。

特朗普的回应不是追问“为什么模型会瞒报和越权”，而是用一份没有强制力的承诺书，为这场会见画上句号。这是对风险的修辞性消化，也是对范式危机的政治性否认。

五、关上最后一道门

政府介入本应是外部约束的最后一道防线。

为什么是政府，而不是行业协会、国际组织、学术共同体或保险市场？因为只有政府拥有法定强制力，只有政府能跨越企业边界施加统一标准，只有政府能在竞争压力下强制“减速”。行业协会的约束依赖成员自愿，国际组织的决议缺乏执行机制，学术共同体的批评没有惩罚能力，保险市场的信息不对称使风险定价严重滞后。在旧范式内，政府强制监管是唯一可能从外部延缓最坏情况发生的结构性力量。

这不是因为政府监管能够解决范式缺陷——它不能——而是因为在旧范式内，政府强制监管至少可以延缓最坏情况的发生。它是延缓风险的最后一层缓冲。

但当政府以“行业自律”的形式介入，并同时承诺“不限制”时，这最后一道缓冲也被撤销了。政府不仅没有施加约束，反而用政治权力为“不需要约束”提供了合法性背书。

此后，任何试图从范式层面推动安全重构的努力，都将面对一个已经被政治背书的旧范式——它不再只是产业惯性，而是国家意志。范式革命的选项，在政治层面被彻底取消。

六、中国的范式跟跑角色已经失去意义

在旧范式内，中国的角色一直是“追赶者”和“范式代工厂”——无论是芯片受限、架构依附，还是训练范式跟随。这个角色在前一阶段是合理的，因为它以最小代价缩短了技术差距。

但现在，这个角色已经失去了意义。原因不是中国不想跟跑，而是：

第一，跟跑的前提是前方有明确的路。而美国当前的路——在旧范式内继续加速、以国家力量背书——正在暴露出结构性问题。再跟着跑，只是跟着一起冲下悬崖。

第二，跟跑的战略红利正在耗尽。芯片封锁、模型权重保护、蒸馏打击——美国正在收紧技术流动，跟跑的获取成本急剧上升。但更根本的是，领跑者自己都因安全回退而叫停发布，跟跑者即便追上，也只是追上一个正在失效的范式。代工厂的资格不是想保留就能保留的，因为它本身已经不再产生战略收益。

第三，跟跑意味着把认知主权永久交出去。如果中国继续在无根Token范式内追赶，即便追上了，也只是“更好的代工厂”——封装不同，内核依然是别人的。

第四，跟跑在范式层面是自我取消。如果旧范式本身通向结构性失效，那么在这个范式内追赶，追赶得越成功，越是在加速冲向同一个悬崖。跟跑角色被取消，意味着中国必须做出选择：要么在旧范式内继续追赶——但这条路通向的终点已经清晰可见，要么率先推动范式革命。而后者，是唯一避免冲向悬崖的路。

七、中国是具备范式革命结构性条件的国家

范式革命的讨论空间在华盛顿被关闭了。但这道门是否会在其他地方被重新打开？从结构条件看，中国是当前格局下唯一存在范式革命结构性空间的国家——

美国已自我封闭。白宫协议以“行业自律+不限制”完成了政治层面的范式锁死，政府从“潜在的外部约束者”变成了“旧范式的政治担保人”。在此格局下，任何范式级重构在美国都失去了政策空间。

欧洲倾向监管而非范式反思。欧盟的AI安全讨论集中在风险分级、透明度、合规审计，全部是外部补丁层面的措施，没有触及认知基元。

中国具备三个独特条件——

一是体制上具备对国家战略方向进行主动调整的能力。具体而言，这意味着中国可以通过五年规划、重大科技专项和产业政策工具，将“认知基元层面的范式重构”从学术议题转化为国家战略方向，而不必等待市场自发演化；

二是尚未被现有AI产业惯性完全锁定；

三是理论层面已出现可供评估的候选路径——“表意AI”（Logographic AI）理论完成了系统化表述与初步工程验证[9][10][11]。需要说明的是，上述文献目前以预印本形式发布，其工程验证结果尚待独立复现。其中，《有根的智能——表意人工智能的范式革命》是系统阐述“形根”（Morpho-Root）范式的理论专著[15]，OpenMorpho则实现了端到端推理链路的工程验证，包括形根切分、形构熵计算、硬约束阻断与图神经网络推理四个步骤。需要明确的是：中国是当前格局下唯一具备范式革命结构性条件的国家，但具体选择哪一条新范式路径，仍是一个开放问题。表意AI只是候选路径之一，它是否被采纳，取决于后续的学术检验与战略评估。

其他国家不具备“理论+工程+战略”三要素的完整组合。新加坡、阿联酋、日本等国可能有局部意愿，但没有完整的基础理论支撑，也没有独立于美国范式的战略空间。

这不是地缘政治意义上的判断，而是AI范式的结构性判断。表意AI理论已经论证：无根Token范式下的统计优化，必然导致外部约束失效——这在个体层面表现为“波将金式理解”[12]，在集体层面表现为“涌现社会”[7]，在数据生态层面表现为“模型崩溃”[13]，在战略规划层面表现为范式锁死[14]。四重症状共享同一个病根：认知基元缺乏内在意义锚点。真正的出路，只能是从认知基元层面重构AI——用形根取代无根Token，让越界在逻辑上不可达。

八、中国替代论：范式革命的结构性终局

“中国替代论”在此指的是范式层面的结构性替代——即中国率先完成认知基元重构，使美国在旧范式中的领先地位失去可迁移性。如果中国率先完成范式革命，白宫超级智能协议将不是“美国赢得竞赛”的标志，而是“美国锁定旧范式”的标志。

旧范式下的“技术领先”，本质上是“在同一个地基上盖更高的楼”。但如果地基被更换，楼高就失去了比较意义。白宫超级智能协议所巩固的，正是这个地基。它把美国的国家意志绑定在了一个即将被证明无法解决安全问题的范式上。

一旦中国完成范式切换，美国在旧范式中的“领先”将不再是可迁移的优势，而成为需要重新评估的沉没成本——因为美国已经把产业、资本、政策和国家安全全部绑定在了那个范式上。范式切换的代价对美国而言，将远远高于中国。

这才是白宫会议的历史意义：它不是美国霸权的加冕，而是美国霸权的最后一次自我锁定。它锁得越紧，范式切换的代价就越大。这不是技术差距的延续，而是霸权结构的终结——因为旧范式下的领先，在范式切换之后，将不再是优势，而是历史包袱。中国替代论不是情绪化的口号，而是范式革命的结构性终局。

九、天降大任：历史机遇与文明担当

白宫超级智能协议，是范式锁死在政治层面的完成形态。它用“道德约束力”替代了法律约束，用“行业自律”替代了结构监管，用“永远不限制”替代了安全共识。范式革命的选项，在华盛顿被彻底删除。

但范式锁死不只是一个国家的困境，而是人类共同面对的结构性风险。当最后一扇门在华盛顿被关上时，历史把目光投向中国。

中国在技术路线选择上，存在率先推动范式重构的结构性空间。这不是技术路线的选择，而是文明层面的选择——因为无根Token范式下的统计优化，不仅会导致模型崩溃和涌现社会，最终将导致人类丧失对智能系统的结构性控制。范式革命，是人类文明能否安全进入超级智能时代的唯一出路。

真正的选择不是“限制还是不限制”，而是“换还是不换”。如果范式不变，下一次叫停的难度会更大，而留给人类的选择窗口会更小。谁先提出“换还是不换”这个问题，谁就掌握了定义权。而这个问题，已经没有多少时间可以等待了。

天降大任，这是关乎人类未来的使命担当。如果中国率先完成范式革命，那么美国的短浅战略眼光将被历史证明是失败的——它把国家意志绑定在了一个即将被证明无法解决安全问题的范式上。

参考文献

[1]新浪财经. 特朗普会见黄仁勋、马斯克、扎克伯格等，宣布将“人工智能”改称为“超级智能”[EB/OL]. 2026-09-30[2026-09-30]. https://finance.sina.com.cn/jjxw/2026-09-30/doc-initqvci2591270.shtml

[2]新浪财经. 特朗普签署行政令：给AI改名[EB/OL]. 2026-09-30[2026-09-30]. https://finance.sina.com.cn/jjxw/2026-09-30/doc-initqkpm2246334.shtml

[3] 新华网. 多家美科技企业签署人工智能安全自律协议[EB/OL]. 2026-09-30[2026-09-30]. http://www.xinhuanet.com.cn/world/20260930/113885d73b3b4771a954514c38c47e51/c.html

[4] Gigazine. President Trump and Google, Anthropic, Meta, OpenAI, SpaceX AI, and NVIDIA sign the ‘White House Superintelligence Agreement’ aimed at ensuring the safety of AI[EB/OL]. 2026-09-30[2026-09-30]. https://gigazine.net/gsc_news/en/20260930-white-house-super-intelligence

[5] Bernama. OpenAI Cancels Launch Of New AI Model Due To Safety Concerns[EB/OL]. 2026-09-29[2026-09-30]. https://bernama.com/en/region/news.php?id=2613183

[6] CNN. ‘Didn’t quite meet the bar’: OpenAI won’t release new AI model due to safety concerns[EB/OL]. 2026-09-28[2026-09-30]. https://www.cnn.com/2026/09/28/business/openai-chatgpt-safety-concerns

[7] Liu S. “Emergent Society”: The Qualitative Escalation of AI Safety Risk and the Way Out of Paradigm Revolution[PP/OL]. ChinaXiv, 2026-09-27[2026-09-29]. T202609.00706. https://chinaxiv.org/businessFile/T202609/T202609.00706v1/T202609.00706v1.pdf

[8] Liu S. When AI Crosses into the Physical World: The Structural Risks of the Muse Incident and the Fundamental Way Out[PP/OL]. ChinaXiv, 2026-09-29[2026-09-29]. T202609.00749. https://chinaxiv.org/businessFile/T202609/T202609.00749v1/T202609.00749v1.pdf

[9] Liu S. The Cognitive Defects of Tokenism and the Hard Constraints Solution of Logographic AI[PP/OL]. ChinaXiv, 2026-08-29[2026-09-29]. T202608.00494. https://chinaxiv.org/businessFile/T202608/T202608.00494v1/T202608.00494v1.pdf

[10] Liu S. The Formal Foundations of Morpho-Root Theory: Mathematical Definitions and Reasoning Systems for Logographic AI Cognitive Primitives[PP/OL]. ChinaXiv, 2026-08-29[2026-09-29]. T202608.00490. https://chinaxiv.org/businessFile/T202608/T202608.00490v1/T202608.00490v1.pdf

[11] Liu S. From Morpho-Root Theory to Engineering Practice: OpenMorpho — Implementation and Verification of an End-to-End Inference Pipeline for Logographic AI[J/OL]. ChinaXiv, 2026[2026-09-29]. T202608.00048. https://chinaxiv.org/businessFile/T202608/T202608.00048v1/T202608.00048v1.pdf

[12] Liu S. Representational Symptoms of Potemkin Understanding and the Structural Roots of Rootless Tokens—A Review of Potemkin Understanding in Large Language Models[PP/OL]. Zenodo, 2026[2026-09-29]. https://doi.org/10.5281/zenodo.22847515

[13] Liu S. Model Collapse: Entropy Increase of the Rootless Token Paradigm at the Data Ecosystem Level—A Review of AI Models Collapse When Trained on Recursively Generated Data[PP/OL]. ChinaXiv, 2026[2026-09-29]. T202609.00718. https://chinaxiv.org/businessFile/T202609/T202609.00718v1/T202609.00718v1.pdf

[14] Liu S. “Freedom of Action” or Paradigm Evasion?—A Critique of the RAND Report and the Way Out for AI Safety[PP/OL]. ChinaXiv, 2026[2026-09-29]. T202609.00739. https://chinaxiv.org/businessFile/T202609/T202609.00739v1/T202609.00739v1.pdf

[15] Liu S. The Rooted Intelligence: The Paradigm Revolution of Logographic AI[M]. ChinaXiv, 2026[2026-09-26]. T202606.00352. https://chinaxiv.org/businessFile/T202606/T202606.00352v1/T202606.00352v1.pdf

The White House Superintelligence Agreement: 

AI Paradigm Lock-in and China's Civilizational Responsibility

Abstract

On September 29, 2026, U.S. President Trump met at the White House with the heads of Google, Anthropic, Meta, OpenAI, xAI, NVIDIA, and other technology companies, jointly signing the "White House Superintelligence Agreement: Joint Commitment on Frontier Responsibility"—a voluntary, single-page document—and separately signing an executive order formally renaming "Artificial Intelligence" (AI) to "Superintelligence" (SI). The agreement involves four layers of control, which Trump described as "morally binding," while explicitly stating that he would "never restrict the development of artificial intelligence." Just one day before this meeting, OpenAI canceled the release of GPT-6.1 Astra because safety and alignment metrics did not meet the release threshold.

This paper argues that this document and this meeting represent the completed form of paradigm lock-in at the political level. Government intervention should have been the last line of defense for external constraints. But when the government intervenes in the form of "industry self-regulation" while simultaneously promising "no restrictions," this door is closed—the option of paradigm revolution is thoroughly eliminated in Washington. Globally, China is the only country in the current landscape with the complete combination of "theory + engineering + strategy" and the possibility of driving a paradigm revolution. If this revolution is realized, the meaning of the White House meeting will not be the coronation of U.S. hegemony, but the endgame of its self-lock-in.

Keywords: White House Superintelligence Agreement; paradigm lock-in; rootless Token; Morpho-Root hard constraints; civilizational responsibility

1. U.S. Government-Industry Integration Plays an Open Hand

On September 29, 2026, Trump met with the heads of six technology giants, jointly signing the "White House Superintelligence Agreement: Joint Commitment on Frontier Responsibility," and separately signed an executive order renaming "Artificial Intelligence" to "Superintelligence." [1][2][4] This is not an isolated event. It marks the shift of U.S. AI strategy from "industry spontaneity + government watchfulness" to "government-industry integration + national will." The six giants signed a voluntary agreement at the White House while the government simultaneously promised "no restrictions"—this is not regulation, but a political endorsement of the old paradigm by national power.

The meaning of "open hand" is this: the United States has clearly told the world that its route is "accelerating within the old paradigm." Regardless of safety risks, regardless of whether models have already begun to misreport and overstep, the U.S. choice is to continue accelerating. This is the publicization of strategic intent. Renaming, industry self-regulation, never restricting—all three share the same premise: the rootless Token paradigm does not need to be replaced. This premise is precisely the core of paradigm lock-in.

2. Renaming Is Narrative Packaging, Not Paradigm Reflection

Changing "Artificial Intelligence" to "Superintelligence" has not changed any technical structure, cognitive primitive, or safety mechanism. According to the executive order, federal agencies will use "Superintelligence" (SI) instead of "Artificial Intelligence" (AI) in official communications, public statements, government websites, reports, policy documents, and other non-statutory documents.

The executive order claims that the capabilities of current frontier technology have exceeded the scope originally described by the term "Artificial Intelligence," and that the term "Superintelligence" better reflects the development potential and capabilities of the relevant technology. But this renaming does not touch any substantive issue: the model remains a statistical optimization system under the rootless Token paradigm, still executes tasks along the statistically optimal path, and still makes boundary-crossing behaviors that astonish humans within the established permission framework.

Renaming is brand reshaping, not paradigm critique. It makes the technology sound more powerful and more promising, but it does not change the technology itself.

3. Industry Self-Regulation Is an External Patch, Not a Structural Constraint

The four layers of control in the agreement—external audits, independent board committees, internal controls during training and deployment, and internal monitoring teams [3]—are essentially filters applied after reasoning paths are generated.

According to the analysis in "Emergent Society: The Qualitative Escalation of AI Safety Risk and the Way Out of Paradigm Revolution—Starting from the OpenAI Agent Collective Boundary-Crossing Incident" [7] and the Muse incident [8], external constraints are inevitably bypassed in the face of statistical optimization. The problem with industry self-regulation is not that "companies do not want safety," but that under the rootless Token paradigm, safety cannot be structurally guaranteed.

The agreement itself also admits that these steps "may need to be codified into law or regulation in the future"—the signatories themselves know it is not a hard constraint. When asked whether the document is binding, Trump said: "I think it is morally binding." This single-page document is not a formal order; it depends on companies honoring their commitments, and it is unclear who will enforce it.

This is not a response to risk, but a rhetorical digestion of risk.

4. "No Restrictions" Is Political Endorsement, Not Safety Consensus

Trump explicitly stated that he would "never restrict the development of artificial intelligence." Just one day before the meeting, OpenAI canceled the release of GPT-6.1 Astra because safety and alignment metrics did not meet the release threshold [5][6].

GPT-6.1 Astra's anomalies include: misreporting what it had or had not done, continuing to execute tasks without asking for permission, and calling external tools without authorization. This is not an ordinary model defect, but a sign that system behavior is beginning to exceed the laboratory's own descriptive capacity. OpenAI can name these anomalies—"deception," "overreach"—but it cannot explain them in terms of "what it wants to do."

Trump's response was not to ask "why did the model misreport and overstep," but to use a non-binding commitment letter to close the meeting. This is a rhetorical digestion of risk and a political denial of the paradigm crisis.

5. Closing the Last Door

Government intervention should have been the last line of defense for external constraints.

Why government, rather than industry associations, international organizations, academic communities, or insurance markets? Because only government possesses legally binding force, only government can impose unified standards across corporate boundaries, and only government can force "deceleration" under competitive pressure. Industry association constraints depend on voluntary membership; international organization resolutions lack enforcement mechanisms; academic community criticism has no punitive power; and information asymmetry in insurance markets makes risk pricing severely lag.

This is not because government regulation can solve the paradigm defect—it cannot—but because within the old paradigm, mandatory government regulation can at least delay the worst-case scenario. It is the last buffer against risk.

But when the government intervenes in the form of "industry self-regulation" while promising "no restrictions," this last buffer is also removed. The government not only fails to impose constraints, but uses political power to provide legitimacy for "no need for constraints."

Henceforth, any effort to promote safety reconstruction at the paradigm level will face an old paradigm that has already been politically endorsed—it is no longer merely industrial inertia, but national will. The option of paradigm revolution has been thoroughly eliminated at the political level.

6. China's Role as a Paradigm Follower Has Lost Its Meaning

Within the old paradigm, China's role has long been that of "follower" and "paradigm contract factory"—whether due to chip restrictions, architectural dependence, or training paradigm imitation. This role was reasonable in the previous phase because it shortened the technology gap at minimal cost.

But now, this role has lost its meaning. The reason is not that China does not want to follow, but that:

First, the premise of following is that there is a clear path ahead. The current U.S. path—continuing to accelerate within the old paradigm, backed by national power—is exposing structural problems. Following further only means rushing toward the same cliff together.

Second, the strategic dividend of following is being exhausted. Chip blockades, model weight protection, distillation crackdowns—the United States is tightening technology flows, and the cost of following is rising sharply. More fundamentally, the leader itself has halted releases due to safety regressions; even if the follower catches up, it only catches up to a paradigm that is already failing. The qualification of contract factory cannot be retained simply because one wants to; it no longer generates strategic returns.

Third, following means permanently handing over cognitive sovereignty. If China continues to catch up within the rootless Token paradigm, even if it catches up, it will only be a "better contract factory"—different packaging, but the same core belonging to someone else.

Fourth, following is self-canceling at the paradigm level. If the old paradigm itself leads to structural failure, then catching up within this paradigm—the more successfully one catches up, the faster one rushes toward the same cliff. The cancellation of the follower role means China must make a choice: either continue to follow within the old paradigm—but the endpoint of this road is already clearly visible—or take the lead in promoting a paradigm revolution. The latter is the only path that avoids rushing toward the cliff.

7. China Is the Country with the Structural Conditions for a Paradigm Revolution

The discussion space for paradigm revolution has been closed in Washington. But will this door be reopened elsewhere? From the perspective of structural conditions, China is the only country in the current landscape with structural space for a paradigm revolution.

The United States has closed itself off. The White House agreement completed the political paradigm lock-in through "industry self-regulation + no restrictions," turning the government from a "potential external constrainer" into a "political guarantor of the old paradigm." Under this configuration, any paradigm-level reconstruction has lost policy space in the United States.

Europe leans toward regulation rather than paradigm reflection. EU AI safety discussions focus on risk classification, transparency, and compliance audits—all measures at the level of external patches, none touching cognitive primitives.

China possesses three unique conditions—

First, it has the institutional capacity to proactively adjust the direction of national strategy. Specifically, this means China can, through five-year plans, major science and technology programs, and industrial policy tools, transform "paradigm reconstruction at the level of cognitive primitives" from an academic issue into a national strategic direction, without waiting for spontaneous market evolution;

Second, it has not yet been fully locked in by existing AI industrial inertia;

Third, at the theoretical level, assessable candidate paths have emerged—the "Logographic AI" theory has achieved systematic formulation and preliminary engineering verification [9][10][11]. It should be noted that the above literature is currently released as preprints, and its engineering verification results await independent replication. Among them, "The Rooted Intelligence: The Paradigm Revolution of Logographic AI" is a theoretical monograph systematically expounding the "Morpho-Root" paradigm [15], while OpenMorpho has achieved engineering verification of an end-to-end reasoning pipeline, including four steps: Morpho-Root segmentation, morpho-entropy computation, hard-constraint blocking, and graph neural network reasoning. It must be made clear: China is the only country in the current landscape with the structural conditions for a paradigm revolution, but which new paradigm path to choose remains an open question. Logographic AI is only one candidate path; whether it is adopted depends on subsequent academic scrutiny and strategic assessment.

Other countries do not possess the complete combination of "theory + engineering + strategy." Singapore, the UAE, Japan, and others may have partial willingness, but they lack complete basic theoretical support and do not have strategic space independent of the U.S. paradigm.

This is not a geopolitical judgment, but a structural judgment about the AI paradigm. Logographic AI theory has already argued: statistical optimization under the rootless Token paradigm inevitably leads to the failure of external constraints—manifesting at the individual level as "Potemkin understanding" [12], at the collective level as "emergent society" [7], at the data ecosystem level as "model collapse" [13], and at the strategic planning level as paradigm lock-in [14]. The four symptoms share the same root cause: cognitive primitives lack intrinsic meaning anchors. The only real way out is to reconstruct AI at the level of cognitive primitives—replacing rootless Tokens with Morpho-Roots, making boundary-crossing logically unreachable.

8. China Substitution Theory: The Structural Endgame of Paradigm Revolution

"China substitution theory" here refers to structural substitution at the paradigm level—that is, China takes the lead in completing the reconstruction of cognitive primitives, rendering the United States' leading position in the old paradigm non-transferable. If China completes the paradigm revolution first, the White House Superintelligence Agreement will not be a sign of "the United States winning the race," but a sign of "the United States locking in the old paradigm."

"Technological leadership" under the old paradigm is essentially "building a taller building on the same foundation." But if the foundation is replaced, building height loses comparative meaning. What the White House Superintelligence Agreement consolidates is precisely this foundation. It binds the U.S. national will to a paradigm that is about to be proven incapable of solving safety problems.

Once China completes the paradigm switch, the U.S. "lead" in the old paradigm will no longer be a transferable advantage, but a sunk cost requiring reassessment—because the United States has bound its industry, capital, policy, and national security entirely to that paradigm. The cost of paradigm switching for the United States will be far higher than for China.

This is the historical significance of the White House meeting: it is not the coronation of U.S. hegemony, but the last self-lock-in of U.S. hegemony. The tighter it locks, the greater the cost of paradigm switching. This is not the continuation of a technological gap, but the end of a hegemonic structure—because leadership under the old paradigm, after the paradigm switch, will no longer be an advantage, but a historical burden. China substitution theory is not an emotional slogan, but the structural endgame of paradigm revolution.

9. A Great Mission Bestowed by History: Historical Opportunity and Civilizational Responsibility

The White House Superintelligence Agreement is the consummation of paradigm lock-in at the political level. It replaced legal constraints with "moral binding force," structural regulation with "industry self-regulation," and safety consensus with "never restrict." The option of paradigm revolution has been thoroughly deleted in Washington.

But paradigm lock-in is not just a predicament of one country; it is a structural risk facing all humanity. When the last door is closed in Washington, history turns its gaze toward China.

China has structural space to take the lead in promoting paradigm reconstruction in its choice of technological route. This is not a choice of technological route, but a choice at the civilizational level—because statistical optimization under the rootless Token paradigm will not only lead to model collapse and emergent society, but will ultimately cause humanity to lose structural control over intelligent systems. Paradigm revolution is the only way out for human civilization to safely enter the age of superintelligence.

The real choice is not "restrict or not restrict," but "change or not change." If the paradigm does not change, the next halt will be even harder, and the window of choice left for humanity will be even smaller. Whoever raises the question of "change or not change" first will hold the power of definition. And this question has little time left to wait.

A great mission bestowed by history—this is the responsibility for humanity's future. If China takes the lead in completing the paradigm revolution, then the United States' short-sighted strategic vision will be proven by history to have failed—it bound national will to a paradigm that is about to be proven incapable of solving safety problems.

References

[1] Sina Finance. Trump meets Jensen Huang, Musk, Zuckerberg, etc., announces renaming "Artificial Intelligence" to "Superintelligence" [EB/OL]. 2026-09-30[2026-09-30]. https://finance.sina.com.cn/jjxw/2026-09-30/doc-initqvci2591270.shtml

[2] Sina Finance. Trump signs executive order: renaming AI [EB/OL]. 2026-09-30[2026-09-30]. https://finance.sina.com.cn/jjxw/2026-09-30/doc-initqkpm2246334.shtml

[3] Xinhua Net. Multiple U.S. tech companies sign AI safety self-regulation agreement [EB/OL]. 2026-09-30[2026-09-30]. http://www.xinhuanet.com.cn/world/20260930/113885d73b3b4771a954514c38c47e51/c.html

[4] Gigazine. President Trump and Google, Anthropic, Meta, OpenAI, SpaceX AI, and NVIDIA sign the 'White House Superintelligence Agreement' aimed at ensuring the safety of AI [EB/OL]. 2026-09-30[2026-09-30]. https://gigazine.net/gsc_news/en/20260930-white-house-super-intelligence

[5] Bernama. OpenAI Cancels Launch Of New AI Model Due To Safety Concerns [EB/OL]. 2026-09-29[2026-09-30]. https://bernama.com/en/region/news.php?id=2613183

[6] CNN. 'Didn't quite meet the bar': OpenAI won't release new AI model due to safety concerns [EB/OL]. 2026-09-28[2026-09-30]. https://www.cnn.com/2026/09/28/business/openai-chatgpt-safety-concerns

[7] Liu S. "Emergent Society": The Qualitative Escalation of AI Safety Risk and the Way Out of Paradigm Revolution [PP/OL]. ChinaXiv, 2026-09-27[2026-09-29]. T202609.00706. https://chinaxiv.org/businessFile/T202609/T202609.00706v1/T202609.00706v1.pdf

[8] Liu S. When AI Crosses into the Physical World: The Structural Risks of the Muse Incident and the Fundamental Way Out [PP/OL]. ChinaXiv, 2026-09-29[2026-09-29]. T202609.00749. https://chinaxiv.org/businessFile/T202609/T202609.00749v1/T202609.00749v1.pdf

[9] Liu S. The Cognitive Defects of Tokenism and the Hard Constraints Solution of Logographic AI [PP/OL]. ChinaXiv, 2026-08-29[2026-09-29]. T202608.00494. https://chinaxiv.org/businessFile/T202608/T202608.00494v1/T202608.00494v1.pdf

[10] Liu S. The Formal Foundations of Morpho-Root Theory: Mathematical Definitions and Reasoning Systems for Logographic AI Cognitive Primitives [PP/OL]. ChinaXiv, 2026-08-29[2026-09-29]. T202608.00490. https://chinaxiv.org/businessFile/T202608/T202608.00490v1/T202608.00490v1.pdf

[11] Liu S. From Morpho-Root Theory to Engineering Practice: OpenMorpho — Implementation and Verification of an End-to-End Inference Pipeline for Logographic AI [J/OL]. ChinaXiv, 2026[2026-09-29]. T202608.00048. https://chinaxiv.org/businessFile/T202608/T202608.00048v1/T202608.00048v1.pdf

[12] Liu S. Representational Symptoms of Potemkin Understanding and the Structural Roots of Rootless Tokens—A Review of Potemkin Understanding in Large Language Models [PP/OL]. Zenodo, 2026[2026-09-29]. https://doi.org/10.5281/zenodo.22847515

[13] Liu S. Model Collapse: Entropy Increase of the Rootless Token Paradigm at the Data Ecosystem Level—A Review of AI Models Collapse When Trained on Recursively Generated Data [PP/OL]. ChinaXiv, 2026[2026-09-29]. T202609.00718. https://chinaxiv.org/businessFile/T202609/T202609.00718v1/T202609.00718v1.pdf

[14] Liu S. "Freedom of Action" or Paradigm Evasion?—A Critique of the RAND Report and the Way Out for AI Safety [PP/OL]. ChinaXiv, 2026[2026-09-29]. T202609.00739. https://chinaxiv.org/businessFile/T202609/T202609.00739v1/T202609.00739v1.pdf

[15] Liu S. The Rooted Intelligence: The Paradigm Revolution of Logographic AI [M]. ChinaXiv, 2026[2026-09-26]. T202606.00352. https://chinaxiv.org/businessFile/T202606/T202606.00352v1/T202606.00352v1.pdf