世纪竞赛:中美人工智能发展战略
本篇编译报道深入分析了中美两国在人工智能(AI)领域截然不同的战略路径。文章指出,当美国政界和科技界将AI视为一场关乎“人工通用智能”(AGI)突破的零和“军备竞赛”时 ,中国正采取一种更为务实和长远的策略。北京的目标并非执着于单一的技术奇点,而是推动“AI+”的广泛应用,使其像电力一样赋能千行百业,并借此构建一个庞大的应用生态系统 。通过开源模型、聚焦实际应用等方式 ,中国试图在另一条赛道上实现“弯道超车”,这种策略的深远影响,已开始在美国科技巨头的全球布局中显现 。

新闻原文
Artificial intelligence: Race of the century China is doing AI differently, but still thinks it can overtake America
ON MAY 21st J.D. Vance, America’s vice-president, described the development of artificial intelligence as an “arms race” with China. If America paused because of concerns over AI safety, he said, it might find itself “enslaved to PRC-mediated AI”. The idea of a superpower showdown that will culminate in a moment of triumph or defeat circulates relentlessly in Washington and beyond. This month the bosses of OpenAI, AMD, CoreWeave and Microsoft lobbied for lighter regulation, casting AI as central to America remaining the global hegemon. On May 15th President Donald Trump brokered an AI deal with the United Arab Emirates that he said would ensure American “dominance in AI”. America plans to spend over $1trn by 2030 on data centres for AI models. The “DeepSeek moment” in January, when the Chinese company unveiled a large language model (LLM) almost matching the capabilities of an OpenAI model, confirmed that China is snapping at the heels of America.
Yet a recent meeting of the Communist Party’s leadership suggests it is preparing for a different kind of strategic race. “American firms focus on the model, but Chinese players emphasise practically applying AI,” says Zhang Yaqin, a former boss of Baidu, a tech giant, now at Tsinghua University. This focus on practical applications—in factories and for consumers—is how China stole a lead in e-commerce and e-payments. On May 19th Jensen Huang, the boss of Nvidia, a chip firm, warned that America is in danger of being left behind again. If American firms do not compete in China as it builds a “rich ecosystem”, Chinese technology and leadership “will diffuse all around the world”, he told Stratechery, a newsletter.
America’s view of AI is often abstract and hyperbolic. LLMs are expected to match humans’ cognitive abilities. Boosters believe this Rubicon of artificial general intelligence (AGI) will be crossed quite soon. Sam Altman, the boss of OpenAI, reckons the next step could be superintelligent systems that actually surpass human abilities in cognitive tasks. Being the first to develop a model that can recursively improve itself (some call this “take-off”) may create a decisive advantage comparable to a nuclear bomb. Barath Harithas of CSIS, a think-tank, notes that American planners believe “the first country to secure the AGI laurel will usher in the 100-year dynasty.” America’s export controls on semiconductors are there to ensure China comes second.
It is true some Chinese entrepreneurs are also believers in the arms race. Liang Wenfeng, DeepSeek’s founder, has made developing AGI his firm’s mission and reckons it may arrive in as little as two years time. Less noticed is that the government is betting on a different approach. Mr Liang’s exploits won him a meeting with Li Qiang, the prime minister, in January. But days later a vice-premier in charge of the party’s science effort seemed to rebuke the American approach, stating: “China will not blindly follow trends or engage in unrestrained international competition.” Last month Qiushi, an authoritative party journal, described AGI as a tool “to promote human understanding and transformation of the world”. In China the term for AGI, tongyong rengong zhineng, typically refers to a “general-purpose AI” that is applied and has multiple uses, rather than to the Western concept of a superhuman, or self-improving, system.
In April the party’s Politburo met for its second-ever study session on AI (the first was in 2018). At the meeting, President Xi Jinping told his lieutenants they should focus on how it can be applied to everyday uses: more like electricity than nuclear weapons. At least a dozen prominent researchers and government officials have aired scepticism over the reasoning ability of LLMs. Wu Zhaohui, a former minister of science and the current vice-president of a state think-tank, suggests China needs to explore different paths to AGI. Chinese experts generally expect AGI to take longer to arrive than do their American counterparts, notes Mr Zhang of Tsinghua.
“While American tech leaders often frame AI with Utopian aspirations, China’s government appears more focused on using it to solve concrete problems like economic growth and industrial upgrading,” says Karson Elmgren of RAND, an American think-tank. The government’s annual work report in March mentions a new campaign called “AI+”, which prioritises firms adopting AI, including in physical facilities using automated robots.
A protracted war
This application-oriented approach reflects a shortage of AI talent and chips, or “basic theory and key core technologies”, Mr Xi said in April. “We must face up to the gap.” Liu Zhiyuan of Tsinghua University compared China’s approach to an argument in “On Protracted War”, a series of lectures by Mao Zedong in 1938: a weak opponent can tire, and outlast, a strong one. On May 8th Qiushi published an article by Tang Jie, also of Tsinghua, urging China to follow American innovation and create applications cheaper and faster.
China’s emerging AI strategy seems to have two parts. One is to undercut the monopoly America has over advanced AI, by replicating Western innovations and making the model weights freely available, or “open source”, as DeepSeek has done. The idea is that the value AI generates will accrue to those who apply it, not to the model-makers. By the time AGI arrives, China will be better placed than America, with “robust social applications, search engines, agents and hardware in place,” notes Kai-Fu Lee, a prominent Beijing-based entrepreneur. He argues that by amassing users and data early, Chinese applications can build a moat that Western competitors will struggle to cross (see Business section), just as TikTok, a video app, has done.
Alongside the push to deploy AI faster and more cheaply, there is an effort to create moonshots that bypass America’s trillion-dollar bet on LLMS. “If we merely follow the well-worn American path—computing power, algorithms, deployment—we will always remain followers,” said Zhu Songchun, boss of the Beijing Institute for General Artificial Intelligence, a state-run laboratory dedicated to advanced AI, in a speech last month. In April the Shanghai government offered funding for researchers advancing towards AGI using new kinds of architectures, such as models that interact with the real world through imagery, others that can control computers with the mind, or algorithms to emulate the human brain.
Will China’s approach work? A new IMF study concludes AI could boost America’s economy by 5.6% in ten years’ time, compared with 3.5% for China, largely because China’s relatively small services sector means that, even if AI diffuses fast in manufacturing, the productivity gains are capped. Yet what is clear is that China is accelerating down a different track. One sign of this is Apple: in order to reverse a decline in its revenue in China, it desperately needs a local partner to provide AI services which customers now expect. But reports suggest America may block it from doing so. Without local AI applications, American tech products, such as the iPhone, risk becoming also-rans in China, and perhaps, in time, elsewhere.
单篇编译版
AI争霸:中美貌合神离的两条赛道
当华盛顿将人工智能(AI)描绘成一场争夺技术制高点的“军备竞赛”时,北京正悄然布局一场不同的世纪对决。中国的策略并非执着于创造一个超越人类的“超智能”,而是要让AI像电力一样渗透到经济的每一个角落——这或许是其后来居上的真正王牌。

豪赌“奇点”:美国的AGI执念
在美国,一场关于AI的“超级大国摊牌”叙事正甚嚣尘上 。上月,美国副总统J.D. Vance将AI发展形容为与中国的“军备竞赛” ,并警告称任何在AI安全问题上的犹豫都可能使美国“被中国主导的AI奴役” 。这种观点的核心,是将AI的终极目标锁定在“人工通用智能”(AGI)的实现上。
在美国科技领袖和政策制定者看来,AGI是能与人类匹敌甚至超越人类认知能力的系统 。率先研发出能自我递归改进的AGI,被比作拥有“核弹”般的决定性优势 。美国智库CSIS的学者指出,华盛顿的规划者相信,“第一个摘得AGI桂冠的国家将开启一个百年王朝” 。正是在这种“赢者通吃”的思维下,美国计划到2030年投入超万亿美元建设AI数据中心 ,并动用半导体出口管制等手段,以确保中国在这场竞赛中屈居第二 。
实干为王:中国的“AI+”主义
然而,正当美国全神贯注于AGI这一终极目标时,中国领导层似乎正准备进行一场不同类型的战略竞赛 。清华大学教授、百度前总裁张亚勤一针见血地指出:“美国公司关注模型,但中国玩家强调AI的实际应用。”
这种对实际应用的侧重,根植于中国高层的战略考量。在今年四月的一次政治局学习会议上,中国国家主席习近平告诉下属,应关注AI如何应用于日常生产生活,使其更像“电力”而非“核武器” 。中国官方对AGI的定义也与西方不同,其术语“通用人工智能”通常指具备多种用途的“通用目的AI”,而非超越人类的系统 。
“美国科技领袖常用乌托邦式的愿景来描绘AI,而中国政府显然更专注于利用它解决经济增长和产业升级等具体问题,”美国兰德公司(RAND)的分析师Karson Elmgren表示。今年三月的政府工作报告中提到的“AI+”行动,就优先支持企业在实体设施中采用AI和自动化机器人 。这种务实的态度,被清华大学学者刘知远比作毛泽东在《论持久战》中的策略:弱小的对手可以通过消耗战最终拖垮并战胜强大的对手 。
生态之战:应用、开源与标准
中国的AI战略分为两部分。其一,是通过复刻西方的创新并将其开源,来削弱美国在先进AI领域的垄断,正如中国公司深度求索(DeepSeek)所做的那样 。其核心思想是,AI创造的价值最终将流向应用者,而非模型制造者 。知名投资人李开复认为,通过早期积累海量用户和数据,中国的应用程序可以构建起西方竞争者难以逾越的护城河,就像TikTok的成功一样 。
芯片巨头英伟达的CEO黄仁勋对此发出了警告。他指出,如果美国公司在中国构建“丰富生态系统”的过程中缺席,中国的技术和标准“将扩散到世界各地” 。
其二,中国在推动AI快速、廉价部署的同时,也在尝试“登月计划”,以绕开美国在LLMs上的万亿美元豪赌 。北京通用人工智能研究院的负责人朱松纯表示,“如果我们只是追随美国的老路……我们将永远是追随者。”
这场竞赛的结局远未确定。但一个清晰的信号是,苹果公司正迫切需要在中国寻找本地AI合作伙伴,以扭转其销售颓势 。若被美国政府阻止,那么像iPhone这样的顶尖美国科技产品,在中国市场恐将沦为配角,并随着时间的推移,可能在世界其他地方也面临同样的命运 。【全文完】
文章分析
这篇题为《世纪竞赛》(Race of the century)的报道,是对中美人工智能(AI)发展战略的一篇深度新闻分析(News Analysis)。它没有停留在简单报道事件的表面,而是深入探讨了两国在AI领域不同路径选择背后的逻辑、策略及其潜在影响。
1. 体裁、文体与结构
- 体裁 (Genre): 本文属于解释性报道(Explanatory Report)或新闻分析(News Analysis)。它旨在回答“为什么”和“这意味着什么”,而不仅仅是“发生了什么”。
- 文体 (Style): 秉承了《经济学人》一贯的分析性、权威性风格。语言精炼、逻辑严谨,通过引用多方信源(政府官员、科技领袖、学者、智库报告)来支撑其论点 ,整体基调客观、冷静,符合新闻专业主义中的准确性和客观性原则 。
- 叙事结构 (Narrative Structure): 这篇文章的结构非常经典,近似于**“华尔街日报体”(Wall Street Journal Structure)** ,其特点是“以小见大”,通过具体案例引出宏大主题:
- 开头 (Hook): 以美国副总统J.D. Vance极具冲击力的“军备竞赛”言论开篇 ,迅速抓住读者眼球,设定了文章的竞争基调。
- 核心段落 (Nut Graf): 随即笔锋一转,在文章前部通过“Yet a recent meeting of the Communist Party's leadership suggests it is preparing for a different kind of strategic race” 这句话,明确点出了全文的核心论点——中国正在进行一场不同类型的竞赛。
- 主体 (Body): 文章主体部分围绕这一核心论点,通过对比(美国追求“人工通用智能”AGI的颠覆性突破 vs. 中国强调“AI+”的实际应用)、举例(DeepSeek的开源 、百度的前总裁的观点 )、引述(英伟达CEO黄仁勋的警告 )等方式,层层深入地论证中美AI战略的差异。
- 结尾 (Kicker): 结尾用苹果公司在华业务的困境作为案例 ,巧妙地呼应了开头的论点,即在应用生态的竞赛中,即便拥有顶尖技术的美国公司也可能落于人后,使得文章结构完整,论点得到有力收束。
2. 信源、可能的偏向和政治立场
- 信源 (Source): 文章来源于英国的《经济学人》,这是一家在全球享有盛誉的媒体,其编辑立场通常被认为是古典自由主义(Classical Liberalism),在美国政治光谱中可被视为中间派或在经济议题上中间偏右 。它关注全球化、自由贸易和市场竞争,这篇文章的分析框架也体现了这一点。
- 潜在偏向 (Potential Bias):
- 框架偏向 (Framing Bias): 文章从一开始就采用了“竞赛”和“摊牌”(showdown)的框架来定义中美AI关系 ,这本身就是一种源于西方的、以竞争为核心的视角。
- 西方中心视角: 尽管文章试图客观呈现中国的策略,但其叙述的基点是西方的,不断将中国的做法与美国进行比较(如“snapping at the heels of America” ,“always remain followers” )。这在某种程度上也符合“东方主义”的某些特征,即在“我们”(西方)与“他们”(东方)的二元对比中来定义“他们” 。
3. 跨文化和国际传播特征
- 文化语境 (Cultural Context): 这是一篇典型的由西方媒体(低语境文化 )解读中国(高语境文化 )战略的报道。文章的写作风格直接、明确,依赖事实和逻辑论证,是典型的低语境沟通方式。
- 文化维度 (Cultural Dimensions): 报道内容无意中反映了霍夫斯泰德的文化维度理论 。它将美国的方式描绘为更具个人主义色彩(强调Sam Altman等天才人物对AGI的追求 ),而将中国的方式描绘为更具集体主义和国家主义色彩(由党和政府主导,服务于经济增长和产业升级的国家战略 )。
4. 背景知识介绍
- 人工通用智能 (AGI - Artificial General Intelligence): 这是文章的核心概念之一。在西方语境下,AGI通常指能与人类一样、甚至超越人类执行任何认知任务的超级智能 。文章特意指出,中国的对应术语“通用人工智能”更多指“通用目的”的人工智能,强调其应用性而非超人性 。
- 大语言模型 (LLM - Large Language Model): 这是当前AI发展的主流技术,如OpenAI的GPT系列。文章中提到的DeepSeek模型就是中国在这一领域的有力竞争者 。
- 美国出口管制 (America's export controls): 美国对向中国出口高端半导体(芯片)实施严格管制 ,目的是延缓中国在需要强大算力的AI模型训练方面的发展速度,这是理解两国AI竞赛背景的关键。
编译分析
经过上述分析,我们对这篇文章有了深入的理解。现在,我们可以开始头脑风暴,讨论如何编译这篇稿件了。
1. 编译方式建议
考虑到这篇文章结构严谨、论证有力、信息量大,我认为最能发挥其价值的编译方式是单篇编译(Single-article Compilation) 。
- 为什么不是全译 (Full Translation)? 全译虽然能保留全部信息,但篇幅较长,且部分针对西方读者的背景细节(如对美国多位政客和公司游说的详细描述 )对国际读者来说可能关联度稍低,可能影响阅读效率。
- 为什么不是节译 (Abridged Translation)? 简单的删减(节译)可能会破坏原文严谨的“华尔街日报体”论证结构 ,仅仅摘取片段会使其失去深度和说服力。
因此,单篇编译允许我们在保留核心论点和证据的同时,对结构进行优化,对次要信息进行精简,使其更适合目标读者的阅读习惯。
2. 具体编译步骤
我们可以按照单篇编译的流程,策划一次“再创作” 。
第一步:确立编译主旨与规划结构
- 核心主旨: 揭示中美两国在AI领域的战略分歧——美国追求革命性的技术奇点,而中国则采取更务实的、以应用为导向的持久战策略,并分析后一种策略的潜在优势和对全球科技格局的影响。
- 规划新结构: 我们可以保留原文的“华尔街日报体”核心逻辑,但为了增强网络传播的可读性,可以增加小标题,并重写导语。
第二步:编译初稿——结构重塑与内容取舍
重写导语 (New Lead): 原文的导语戏剧性强,但我们可以尝试一个更直接点明核心矛盾的导语,比如:
“当华盛顿将人工智能(AI)描绘成一场争夺技术制高点的‘军备竞赛’时,北京正悄然布局一场不同的世纪对决。中国的策略并非执着于创造一个超越人类的‘超智能’,而是要让AI像电力一样渗透到经济的每一个角落——这或许是其后来居上的真正王牌。”
增加小标题 (Add Subheadings): 将原文内容重新组织在清晰的小标题下,例如:
- 豪赌“奇点”:美国的AGI执念 (The American Gamble on 'Singularity') 整合原文中关于美国视AGI为核弹、追求超级智能、出口管制等内容
- 实干为王:中国的“AI+”主义 (Pragmatism is King: China's 'AI+' Doctrine) 整合中国强调实际应用、官方文件导向、产业升级等内容
- 生态之战:应用、开源与标准 (The War of Ecosystems: Applications, Open Source, and Standards) 整合黄仁勋的警告、DeepSeek的开源策略、以及未来应用生态可能决定胜负的观点
内容删减与精炼 (Abridgement & Refinement): 可以适当精简美国国内政治的细节,比如将“J.D. Vance”、“Donald Trump”等人物的言论进行概括性转述,重点突出其观点即可。对于专家引语,保留最核心、最精辟的部分。
第三步:细节处理与语言润色
- 文化适配: 在翻译AGI等概念时,要清晰地解释中美语境下的不同侧重。对于《经济学人》这种带有鲜明风格的文本,译文要力求在保持准确的同时,也能体现出其语言的精炼和逻辑的严密。
- 撰写新标题: 我们可以构思几个更具吸引力的中文标题:
- 《AI争霸:中美貌合神离的两条赛道》
- 《超越“军备竞赛”:中国AI的“长征”新策略》
通过以上步骤,我们大致就能将一篇优秀的英文深度报道,编译成一篇既忠实于原文核心精神,又更符合目标读者阅读习惯的、高质量的中文新闻作品。