AI and Human History AI与人类历史

Will AI Mark the End of Human History?

Note: This article continues a series on AI and the future. Check out the previous articles: “On the Prospects of AI in Education,” “AI and Academics,” and “The Challenges of AI”

In the view of the popular historian and philosopher Yuval Noah Harari, history is the story of three major revolutions. The Cognitive Revolution 70,000 years ago endowedHomo sapienswith intelligence and language abilities, which allowed them to create stories and coalesce into ever more complex social systems. The Agricultural Revolution 12,000 years ago marked a significant shift in the lifestyle ofSapiens. With the birth of the agricultural economy, bureaucracies and governments emerged, and wars, empires, money, and religion became indispensable factors that shaped history. The Industrial Revolution about 250 years ago drastically accelerated the development of societies, bringing about a globalized world of unknown opportunities and challenges. Today, Harari argues that unlike the Cognitive, Agricultural, or Industrial Revolutions, at the cusp of an AI revolution, we must warn ourselves that this technology can potentially end human history.

This article will encourage one to reappraise Harari’s view on AI. Human history can be defined as the processes and outcomes of human social interactions. The key here is that there must be a human in the loop. Human history relies chiefly on two engines: Conversations and conflicts. While AI poses many unforeseen dangers, we shall see that AI is almost certainly not impactful or unique enough to obliterate and replace these two engines, at least in the short term.

First, AI is unlikely to eradicate or prevent humans from conversing with each other. Conversation is an engine of history because most states in the modern age rely on conversations between the government and citizens, as well as conversations between citizens themselves, to function. Conversations allow the quick dissemination of political, economic, and social news and messages, determining the dynamics within and between nations. Since history can be understood as the process and outcome of social interactions, once conversations disappear, states will collapse – the structures in which social interactions occur – and human history might subsequently end.

Some argue that AI demonstrates the danger of severely disrupting and ending conversations. AI algorithms have already generated and spread fake information into human networks. In the future, as algorithms evolve to be more emotional and logical, the argument goes, people will be unable to discern humans from bots. Individuals will distrust each other and the news they receive, declining to engage actively in large-scale conversations. In this way, AI eradicates trust, kills conversations, and pulls human history to a screeching halt.

There is merit to this argument. AI will undoubtedly become more omnipotent. Unlike almost all previous human technological inventions, AI will not be a tool anymore but a new “species,” even an independent companion to a human. AI will also likely disrupt current modes and systems of human communication, introducing novel challenges to maintain adequate and meaningful conversations.

However, this claim that AI shall upend all human conversations turns less convincing when one examines the relationship between technological advancement and the patterns of human social interactions. When new technologies capable of completely redesigning and reimagining communication networks emerge, humans almost always tend to adapt or evolve their social networking systems. We invent social networking or political conversation systems that integrate these technologies in remarkably ingenious ways. For example, when inventors introduced the radio and television in the late 19^thcentury through the early 20^thcentury, these technologies were met with fear and doubt from the public.Both these inventions were new types of mass communication that allowed people to hear information, music, and entertainment from a distance. They radically differed from previous media, such as newspapers and theater, which were more localized. People expressed concern that the radio and television could be used for propaganda. Individuals would no longer be able to distinguish truthful information from misleading messages. People feared meaningful conversations, which occurred at a small and localized scale previously, would not happen anymore.

Contrary to this popular belief, societies around the globe successfully adapted and changed their pathways and systems of conversation. The radio and television did not snuff out sociopolitical dialogues. Societies realized the potential of these technologies, and many previously feudal, authoritarian nations reshaped their polities into large-scale democracies. This change occurred in India, West Germany, Argentina, South Africa, and many others. Democracy was a new communication network in which the pathways and means of human-to-human conversation were reimagined. Since a democracy relies heavily on the widespread dissemination of information and frequent, dense conversations between civilians, politicians integrated the radio and television to fulfill this need. In West Germany, for instance, after the end of World War II,regional stations around Frankfurt played imperative roles in reconstructing the public sphere by disseminating relatively impartial news and fostering free political discussion.

In the cases of the radio and television, we can understand that significant leaps in information technology that redefine human conversations do not often lead to their demise. Instead, faced with the rapid spread of these technologies, societies build new communication frameworks that make good use of their power and enhance the efficiency or effectiveness of conversations. Additionally, we usually enact new legislation and guardrails to tackle the novel challenges brought by innovations in information technology. Following the spread of the radio and television, laws were set up to punish those who spread misinformation via these networks. Governments and intellectuals crafted choice architectures and schemes to encourage citizens to cross-verify information.

Pivoting back to AI, the above analysis means that if AI radically changes the ways of communication, it is too early yet to declare a death sentence to human conversation. Following the AI revolution, new political and social networking structures might emerge. For example, suppose we can carry an AI algorithm that allows one to converse with a foreigner in their language and by their cultural norms. In that case, we can erase borders and establish a new international political system, a federation of peoples. In this new system, with AI technology fully integrated into its framework, we reimagine conversations: conversations can happen globally, instantly, without cultural barriers. As wild a dream as this is, while people should not disregard the risks of AI, they should also consider this often-forgotten possibility that AI will reshape, not destroy, modes of conversation. Human communication systems are flexible enough to evolve.

If we take this argument a step further, on a side note, we can analyze the opinion that AI will obliterate many democracies, the prevailing type of government, worldwide. When the radio and television were introduced, many populations abandoned authoritarianism for democracy, a polity better adapted to the change of the times. AI may push humans to similarly rethink their communication pathways to such an extent that they find democracy too powerless to respond to fresh changes. Societies might replace democracy with a newly designed government that is readier to utilize the advantages and tackle the challenges of the AI era. From a historical standpoint, with minimal bias, this process, if it happens, makes sense. Austrian economist Joseph Schumpeter coined the term “creative destruction” – the process of innovation that replaces outdated practices and conventions. At the cusp of almost every significant revolution, societies have undergone radical and sometimes relentless changes that have resulted in the death of millions, brutally dismantling the products of a previous age. To ensure a transition that is less sudden and violent, countries must set up new guardrails and relevant technologies to keep the many risks of AI at bay.

Now, after we have examined why AI might not tear down the first engine of human history, conversations, we can briefly investigate why AI will similarly not power off the second engine – conflicts. As put forth multiple times before, conflict drives human history. It is the pain and damage brought by conflicts that incentivize humans to innovate and build more effective peacekeeping institutions or organizations. Human history will stagnate without conflict, and we will think less about improving the current frameworks supporting our nations. This is not to justify conflict in any way. Wars and strife between different social groups should rarely be considered just actions, especially from a moral standpoint. However, if we zoom out to view the bigger picture, it has been conflicts that prodded societies to put in place almost all institutions, laws, and frameworks to enforce peace and equity, from the League of Nations, the UN, to legislation that targets racial discrimination, even to minimum wage policies and social welfare programs. Conflict is a prerequisite for change, providing momentum to human history.

AI will probably not eradicate conflicts from this world. AI is unlikely to help nations secure peace and stop wars. Countries might tweak AI codes and algorithms for military purposes. The US has been actively using AI and machine learning to analyze data and identify objects and people from video feeds, boosting the decision-making process for military operations. Russia is actively testing AI in autonomous weapons systems and building robots designed for operations in dangerous zones. In fact, a new AI arms race might be on the horizon, comparable to the nuclear arms race between the US and USSR during the Cold War. AI, hence, is unlikely to prevent more wars from occurring.

Nor will AI erase inequalities in opportunity, another source of societal conflict. AI might assist governments in tackling disparities in healthcare and other social services. AI can disseminate healthcare information and consultancy via the internet to all citizens, allowing the poor to access healthcare knowledge. Once they enter the market, AI robotic doctors and surgeons might drive down the prices of crucial and complex surgeries or other treatments, making them more accessible to a more significant portion of the population. Unfortunately, as highlighted in a previous article, AI will create new educational inequalities, as some students will have access to more powerful and capable AI tutors or algorithms than others, leading to a difference in their academic experiences. If inequalities continue or worsen, conflicts between different social groups will persist.

What does all this mean in the context of human history? As AI is unlikely to erase conflicts, human history will still progress. With a potential AI arms race and many newly created inequalities, governments worldwide will respond to scrap old, ineffective policies and design new ones to strengthen domestic or international efforts to keep peace and protect equity. There might be additions to the components of the UN that convene and push for regulations in the development of AI for military uses. Innovative developmental economic policies might be in place to tackle possibly rising disparities in educational opportunities and resources. It is still reasonable to believe that conflicts push human societies to make more informed, well-deliberated decisions in the AI age, ensuring the longevity and vitality of civilization. No matter how the outer environment changes, in the words of Immanuel Kant, human nature consistently demonstrates a dichotomous “unsocial sociability.” On the one hand, we seek conflict, isolation, and the destruction of the social conventions that keep us connected. On the other hand, we seek social bonding and connection-building.

By this point, we have briefly examined how AI will probably not destroy the double engines of human history: Conversations and conflicts. Admittedly, this commentary rests mainly on predictions. Hopefully, the analysis above will spark more discussions that look into the future, not just the past and present. After all, the role of a historian in this tumultuous age is not necessarily to call for a total halt to the development of AI or to dissipate excessive fear about the hazards of this technology. It is rash to stop this or any other wave of technological leap in history. The responsibility of a historian, instead, should be to guide debates on how to utilize AI to benefit human knowledge and social interactions and how to enact the necessary legislation to complement the craze and excitement AI brings. The social intelligentsia must look across each other’s disciplines and produce a concerted effort. If we can conjure up unity and determination with discretion, then the AI age marks not the end of human history but the beginning of a new, epic era.

人工智能是否会终结人类历史？

注：本文为人工智能与未来系列文章的续篇。请查看之前的文章：“人工智能与教育的前景”、“AI与学术”、“人工智能的挑战”

在著名历史学家、哲学家尤瓦尔-诺亚-赫拉利（Yuval Noah Harari）看来，历史就是三次大革命的故事。7万年前的认知革命赋予了智人超常的智慧与语言能力，使他们能够创造虚构的故事，并凝聚成日益复杂的社会体系。1.2万年前的农业革命标志着智人生活方式的重大转变。随着农业经济的诞生，官僚机构、政府应运而生，战争、帝国、货币和宗教成为塑造历史不可或缺的因素。大约250年前的工业革命大大加速了社会的发展，带来了一个充满未知机遇与挑战的全球化世界。如今，赫拉利认为，与认知革命、农业革命或工业革命不同，在人工智能革命的风口浪尖上，我们必须警告自己，这项技术有可能终结人类历史。

本文作者对赫拉利这个观点持保留意见。人类历史可以定义为人类社会互动的过程与结果，这里的关键是，必须有“人类”参与其中。人类历史主要依靠两个引擎：对话与冲突。虽然人工智能会带来许多不可预见的危险，但至少在短期内，人工智能的影响力或独特性几乎肯定不足以毁灭或取代这两个引擎。

首先，人工智能不太可能消灭或阻止人类相互对话。对话是历史的引擎，因为现代大多数国家都依靠政府与公民之间的对话以及公民之间的对话来运转。对话可以快速传播政治、经济和社会新闻与信息，决定国家内部及国家之间的动态。由于历史可以理解为社会互动的过程和结果，一旦对话消失，国家，即承载社会互动发生的结构，就会崩溃，人类历史也可能随之终结。

有人认为，人工智能展示出严重扰乱并终结对话的危险。人工智能算法已经在人类网络中生成并传播虚假信息。这种观点认为，在未来，随着算法进化得更加感性且富有逻辑，人们将无法分辨人类或机器人。人们会相互不信任，也会不信任他们收到的信息，从而拒绝积极参与大规模对话。如此，人工智能就会消除信任，扼杀对话，使人类历史戛然而止。

这种说法有其道理。人工智能无疑会变得更加无所不能。与人类以往几乎所有的技术发明不同，人工智能将不再是一种工具，而是一个新的“物种”，甚至是人类的独立伴侣。人工智能还将可能颠覆人类现有的交流模式和系统，为保持充分及有意义的对话带来新的挑战。

然而，当我们审视技术进步与人类社会互动模式之间的关系时，人工智能将颠覆人类所有对话的说法就变得不那么令人信服了。当能够完全重新设计和想象人类沟通网络的新技术出现时，人类几乎总是倾向于调整或“进化”自己的社交网络系统。我们善于发明新的社交网络或政治对话系统，以非常巧妙的方式整合了这些新的技术。例如，19世纪末到20世纪初，当发明家们推出收音机和电视机时，公众对这些技术感到恐惧、怀疑。这两项发明都是新型的大众传播手段，可以让人们从远处听到信息、音乐和娱乐。它们与以往的媒体（如报纸和戏剧）截然不同，后者更加本地化。人们担心广播或电视会被用于政治宣传，人们再也无法区分真实信息和误导信息。人们担心，以前在小范围和局部地区进行的有意义的对话将不再发生。

与这种普遍看法相反，全球各地的社会成功地调整和改变了他们的对话途径与系统。广播或电视并没有扼杀社会政治对话。社会意识到了这些技术的潜力，许多过去封建专制的国家将其政体重塑为大规模的民主政体。这种变化发生在印度、西德、阿根廷、南非等许多国家。民主是一个新的沟通网络，在这个网络中，人与人之间的对话途径和手段被重新规划。由于民主在很大程度上依赖于信息的广泛传播与公民之间频繁而密集的对话，政治家们将广播和电视整合进政治系统来满足这一需求。以西德为例，二战结束后，法兰克福周边地区的广播电台通过传播相对客观的新闻并促进自由的政治讨论，在重建公共领域方面发挥了至关重要的作用。

从广播及电视的案例中，我们可以了解到，信息技术的重大飞跃重新定义了人类的对话，但这往往不会导致它们的消亡。相反，面对这些技术的迅速普及，社会会建立新的传播框架，充分利用这些技术的力量，提高对话的效率或效果。此外，我们通常会制定新的法律和防范措施，以应对信息技术创新带来的新挑战。随着广播与电视的普及，人们制定了法律来惩罚那些通过这些网络传播错误虚假谣言的人。政府与知识分子精心设计了选择架构与助推计划，鼓励公民交叉验证信息。

回到人工智能，上述分析意味着，如果人工智能从根本上改变了交流方式，那么现在就宣布人类有意义的对话将会终止还为时尚早。人工智能革命之后，可能会出现新的政治或社交网络结构。例如，假设我们可以携带一种人工智能算法，让人们可以用外国人的语言并按照他们的文化规范与之交谈。在这种情况下，我们就可以消除边界，建立一个新的国际政治体系–一个“民族联盟”。在这个新体系中，人工智能技术完全融入其框架，我们可以重新想象对话：对话可以在全球范围内即时进行，没有文化障碍。这是一个天马行空的梦想，但人们在不忽视人工智能的风险的同时，也应考虑到这种经常被遗忘的可能性，即人工智能将重塑而非摧毁对话模式。人类的交流系统具有足够的灵活性，可以不断进化。

如果我们将这一论点再向前推进一步，我们可以分析一个更具体的未来，即分析一下关于人工智能将摧毁世界上许多民主政体—世界上最普遍的政府类型-的观点。当广播和电视出现时，许多国家的公民放弃了专制，转而选择了民主，因为民主政体更能适应时代的变化。人工智能可能会促使人类同样重新思考他们的沟通途径到如此的程度，以至于他们发现民主制度无力应对新的社会变化。社会可能会用一种新设计的政府来取代民主，这种政府更充分地利用了人工智能时代的优势，并更为有效地应对了人工智能时代的挑战。从历史的角度来看，这一进程的发生是合理的。奥地利经济学家约瑟夫-熊彼特（Joseph Schumpeter）创造了“创造性破坏”（creative destruction）一词--创新取代过时做法和惯例的过程。几乎在每一次重大革命的风口浪尖，社会都经历了剧烈的、有时甚至是无情的变革，导致数百万人死亡，残酷地摧毁了上一个时代的产物。因此，为了确保转型不那么突然和剧烈，各国必须设置新的“防护措施”与相关技术，以防范人工智能带来的诸多风险。

现在，在我们研究了为什么人工智能可能不会拆掉人类历史的第一个引擎--对话之后，我们可以简单思考一下为什么人工智能同样不会熄灭第二个引擎--冲突。正如之前此公众号一些文章多次提出的，冲突推动着人类历史的发展。正是冲突带来的痛苦和伤害激励着人类不断创新，建立更有效的维和机构或组织。没有冲突，人类历史将停滞不前，我们也将用更少的时间或精力考虑改善目前支持我们国家的框架。这绝不是为冲突辩护。不同社会群体之间的战争和纷争很少可以被视为正义之举，尤其是从道德的角度来看。然而，如果我们将视线拉得更宏大一些，就会发现正是冲突促使社会建立了几乎所有的机构、法律与框架，以实现和平与公平，从国际联盟、联合国到针对种族歧视的立法，甚至到最低工资政策和社会福利计划。冲突是变革的前提，为人类历史提供动力。

人工智能可能无法根除这个世界上的冲突。人工智能不太可能帮助各国确保和平、制止战争。各国可能会出于军事目的调整人工智能代码和算法。美国一直在积极利用人工智能与机器学习来分析数据，训练人工智能从视频画面中识别物体与人员，从而促进军事行动的决策过程。俄罗斯正在积极测试自主武器系统中的人工智能，并制造专为在危险地区作战而设计的机器人。事实上，一场新的人工智能军备竞赛可能即将到来，堪比冷战时期美苏之间的核军备竞赛。因此，人工智能不太可能阻止更多战争的发生。

人工智能也不会消除机会不平等，这是社会冲突的另一个根源。诚然，人工智能可以协助政府解决医疗保健或其他社会服务方面的差距。人工智能可以通过互联网向所有公民传播医疗保健信息和咨询，让穷人获得医疗保健知识。一旦进入市场，人工智能机器人外科医生可能会降低关键且复杂手术或其他治疗的价格，使更多人能够获得这些服务。不幸的是，正如前一篇文章所强调的，人工智能将造成新的教育不平等，因为一些学生将比其他学生有机会接触到更强大、更有能力的人工智能导师或算法，从而导致他们的学习经历出现差异。如果不平等继续存在或恶化，不同社会群体之间的冲突将持续存在。

这一切对人类历史意味着什么？由于人工智能不可能消除冲突，人类历史仍将向前发展。面对潜在的人工智能军备竞赛和许多新产生的不平等，世界各国政府将采取应对措施，废除旧的、无效的政策，制定新的政策，以加强国内或国际维护和平与保护公平的努力。联合国可能会增设一些部门，召集并推动对人工智能军事用途的开发进行监管。创新的发展型经济政策可能会到位，以解决教育机会和资源方面可能不断扩大的差距。我们仍然有理由相信，冲突会推动人类社会在人工智能时代做出更加明智、深思熟虑的决定，确保文明的寿命与活力。无论外部环境如何变化，用康德的话说，人类的本性始终表现出一种对立性的“非社会交际性”。一方面，我们寻求冲突、孤立和破坏使我们保持联系的社会习俗。另一方面，我们却又寻求社会纽带与温暖链接的建立。

至此，我们已经简要探讨了人工智能可能不会摧毁人类历史的双引擎：对话和冲突。诚然，这篇评论主要基于预测。作者希望上述分析能引发更多的讨论着眼于未来，而不仅仅是过去和现在。毕竟，在这个动荡的时代，历史学家的作用并不能是呼吁完全停止人工智能的发展，或传播对这项技术危害的过度恐惧。阻止历史上这一轮或其他任何一轮技术飞跃都是轻率的。相反，历史学家的责任应该是引导人们讨论如何利用人工智能来造福人类知识的创造与社会交往，以及如何制定必要的立法来适当控制住人工智能带来的具有隐患的极端热潮与兴奋。社会知识分子必须跨越彼此的学科，形成合力。如果我们能够团结且慎重地确定决心，那么人工智能时代就不是人类历史的终结，而是一个新的，史诗般时代的伊始。