
世界领导人都知道,旧的全球秩序正在消亡。2027年,他们将首次领会取而代之的新现实的怪诞本质。正如爱因斯坦的广义相对论扭曲了牛顿关于时间、空间和引力的概念一样,各国政府将开始看到,一种陌生的国际关系几何正在他们周围形成。一些国家将嗅到机遇;更多国家将意识到自己准备不足,且易受其并不真正理解的奇异力量冲击。
WORLD LEADERS know that the old global order is dying. In 2027 they will, for the first time, grasp the weirdness of the new reality that is replacing it. As surely as Einstein’s general relativity warped the Newtonian concepts of time, space and gravity, so governments will begin to see that an unfamiliar geometry of international relations is forming around them. Some will sense opportunities; many more will realise how unprepared they are, and that they are vulnerable to strange forces they do not really understand.
原因在于人工智能的猛烈冲击。即便没有它,战后1945年建立的秩序也正因中国的崛起和美国对充当世界唯一霸权所需军事与财政成本的不耐烦而被拖垮。有了AI作为加速剂,崩塌的秩序将燃烧得更快,因为这项技术将直接冲击国家权力的许多关键要素。
The reason for this is the onrush of artificial intelligence. Even without it, the post-1945 order was being pulled down by the rise of China and America’s impatience with the military and financial costs of acting as the world’s sole hegemon. With AI as an accelerant, the collapsing order will burn much faster, because the technology will impinge directly on many of the vital components of national power.
先看时间。AI时间不仅比政府时间流逝得更快,而且随着算力增长和前沿模型能力增强,正在呈指数级加速。第一颗原子弹1945年投入实战;管控核武器扩散的《不扩散核武器条约》用了25年才生效。如今前沿实验室发布新AI模型的速度已达每月十余个。未来一两年内,模型可能实现自我改进。
Start with time. AI time does not just pass faster than government time, it is also accelerating exponentially, as computing power grows and frontier models become more capable. The first atomic bomb was used in war in 1945; the Nuclear Non-Proliferation Treaty, which governs the spread of nuclear weapons, took 25 years to come into force. Today the frontier labs are releasing new AI models at the rate of over ten a month. Within the next year or two, the models could be improving themselves.
正如新冠疫情所显示的,人类很难理解指数增长。他们不断对最新一次翻倍的规模感到惊讶,无论是新增感染人数,还是最新AI模型的新能力。政府机构更难以应对这种变化节奏。他们需要对不同利益集团的游说保持开放态度——这就是为什么他们通过咨询期、听证会和上诉权来设置延迟。而政治领导人自然倾向于关注AI的国内影响——直到国外爆发危机。
As covid-19 showed, human beings struggle to comprehend exponential growth. They are continually surprised by the magnitude of the latest doubling, whether it is the number of new infections or the new powers of the latest AI model. Government institutions struggle even more to cope with such a rhythm of change. They need to be receptive to lobbying from different interests—which is why they build in delays with consultation periods, hearings and rights of appeal. And political leaders will naturally tend to focus on AI’s domestic impacts—until a crisis erupts abroad.
时间之后是空间。正常空间受物理距离的制约。英国与爱尔兰的贸易额是与日本的2.5倍,尽管日本的经济规模是爱尔兰的六倍。人们每年在爱尔兰海两岸往返超过1200万次,但英国与日本之间仅约70万次——日本的人口是爱尔兰岛的17倍。
After time comes space. Normal space is governed by physical proximity. Britain does 2.5 times more trade with Ireland than with Japan, even though Japan’s economy is six times bigger. People make more than 12m round trips across the Irish Sea each year, but only about 700,000 between Britain and Japan—which has 17 times as many people as the island of Ireland.
相比之下,AI空间是全球性的。为数据中心提供供应的供应链环绕整个星球。美国设计芯片,日本提供芯片制造化学品,荷兰制造光刻设备,韩国生产内存,台湾生产处理芯片,等等。即使中国也依赖其中一些,尽管它努力发展自己的供应链。每个环节都是潜在的杠杆来源。
By contrast, AI space is global. The supply chains that furnish data centres are wrapped around the planet. America designs chips, Japan supplies chipmaking chemicals, the Netherlands builds lithography equipment, South Korea makes memory, Taiwan produces processing chips, and so on. Even China depends on some of these, despite efforts to develop its own supply chain. Each is a potential source of leverage.
人工智能将考验政治体系适应创造性破坏的能力。这些模型同样不受距离限制。它们大多是美国的或中国的,用户面临压力,必须把自己绑定在某一套技术栈上,从芯片到聊天机器人。模型经过优化,以便在例如英伟达或华为的处理器上运行。各国规则涵盖出口管制、数据隐私和安全。各国自然希望避免受制于单一供应商,但美国和中国会设法把它们拉入自己的阵营。
Artificial intelligence will test how well political systems can adapt to creative destruction The models defy distance, too. They are, for the most part, American or Chinese—and users are under pressure to bind themselves to one stack or the other, from chip to chatbot. Models are optimised to run on, say, Nvidia or Huawei processors. National rules cover export controls, data privacy and security. Countries, naturally, want to avoid being captive to a single supplier, but America and China will seek to corral them.
美国的“硅和平”机制承诺提供其技术,前提是排除中国设备。中国的“数字丝绸之路”推广廉价、开放权重的模型,作为对美国“新帝国主义”的制衡。
America’s Pax Silica regime promises access to its technology, so long as Chinese kit is excluded. China’s Digital Silk Road promotes cheap, open-weight models as a counterweight to American “neo-imperialism”.
最后是人工智能危险的引力转移,这可能把世界拖入冲突。已故的亨利·基辛格曾指出,稳定的世界秩序在合法性与权力之间取得平衡。若要持久,它们必须依照被广泛接受的规则运作,而所有人都知道这些规则在必要时能够通过武力成功捍卫。人工智能将在国家能力的各个维度上打破这一平衡。
Last is AI’s dangerous shift in gravity, which risks dragging the world into conflict. The late Henry Kissinger observed that stable world orders strike a balance between legitimacy and power. If they are to endure, they must work by broadly accepted rules that everybody knows can, if necessary, be successfully defended by force. AI will disrupt that equilibrium in all the dimensions of state capacity.
战争形态正在乌克兰被重塑,成群的AI无人机和自主武器投入战场;在伊朗上空,AI增强的瞄准加速了美国和以色列的轰炸节奏。对人工智能基础设施的大量投资正在撼动各国经济,预示着剧烈的商业变革——即便股市崩盘从中作梗。而且,正如美国中期选举中数据中心争议所显示的,人工智能将考验政治体系适应创造性破坏的能力。
Warfare is being recast in Ukraine, by swarms of AI-enabled drones and autonomous weapons, and over the skies of Iran, where AI-enhanced targeting accelerated the pace of American and Israeli bombing. Economies are being shaken up by a torrent of investment in AI infrastructure that heralds dramatic commercial change—even if a stock-market crash gets in the way. And, as the data-centre flap in America’s midterm elections showed, AI will test how well political systems can adapt to creative destruction.
有一点是确定的。人工智能将制造赢家和输家。一些雄心勃勃的国家领导人会告诉自己,世界已变得不公平,因为它不再反映他们增强后的军事、经济和政治地位。感受到自身力量,少数人会变得不耐烦——而不耐烦可能导致战争。
One thing is certain. AI will create winners and losers. Some ambitious national leaders will tell themselves that the world has become unfair, because it no longer reflects their enhanced military, economic and political status. Feeling their own power, a few will become impatient—and impatience can lead to war.
在许多领域,人工智能可以帮助人类克服自身的局限。它在解决数学问题、模拟蛋白质结构、编写代码和分析数据方面具有无与伦比的能力。相比之下,地缘政治仍然取决于人类的智慧。从第二次世界大战中崛起的强国,能否容纳在人工智能时代蓬勃发展的强国?而蓬勃发展的国家,能否克制住算账和征服的古老冲动?■
In many areas AI can help humans overcome their limitations. It has an unmatched ability to solve mathematical problems, model the structure of proteins, write code and analyse data. Geopolitics, by contrast, will still depend on the wisdom of people. Can the powerful countries that emerged from the second world war accommodate the powerful countries that thrive in the era of AI? And can the countries that thrive resist the ancient urge for score-settling and conquest? ■