Tian Feilong 天飞龙
Born 1983, Tian Feilong is a child of the Xi era. He swam across our screen some years ago, writing on the ‘project system’ or xiangmu zhi 项目制, a powerful research paradigm emerging from some fine scholars in the Institute of Sociology at Peking University (where I taught as an adjunct professor 2012-17).
Tian’s next appearance, as a champion of the National Security Law and other aberrations in Hong Kong, sapped our interest. Something had induced him, we surmised, to sing in tune, and not a very attractive tune at that.
Remembering his early promise though, we kept an eye on his writings.
The following is recent, reacting to the huge brouhaha about ChatGPT, the natural language AI. The nationalism of much of his ‘in tune’ output is absent; we find him urging surveillance of ‘super AI’ a the level of the United Nations, and lamenting ‘collusion of capital and the state’ (by which he no doubt means the US, not the PRC).Of greatest interest for Beijing Baselines is Tian’s implicit doctrine of ‘baseline rationality’, developed with no discernible Marxism, Maoism or Xi-ism.
Tian Feilong
Challenges of ChatGPT, machine civilisation and super artificial intelligence
Open AI, an American AI research lab, in late 2022 launched a new natural language processing tool, ChatGPT, arousing fierce reactions in business, technology, society, ethics, and law. A milestone in AI development, it is a major test and challenge for human existence and ethics. AI is essentially a kind of machine civilisation, leveraging human rationality. Yet when it at a certain point reaches full build-up and build-out, this directed and anthropomorphic machine civilisation may replace the labour and existence of humanity, disrupting and harming human society’s basic structure, communication methods, and ethical/legal systems.
Not only an inevitable trend of capital seeking profit, AI’s tech development and commercial application is a new arena for inter-state power rivalry. The PRC and the US are both major powers in AI tech and its commercial applications. Their confrontation and rivalry in the high-tech field will inevitably affect this field with the nature of a new Cold War. The US has formulated relevant laws and policies to seize the commanding heights via vigorously developing AI. The emergence of ChatGPT is the result of the combined effect of capital and international rivalry. In the face of AI development and the extensive challenges it brings, Beijing should do a good job of technical reserves, ethical guidance and legal regulation to ensure its own technological advantages and moral legitimacy in this field.
Two-sided machine civilisation
The difference between man and ape, i.e. natural animals, lies in making and using tools, i.e., rational labour. The fundamental purpose of human rationality applied to labour is human survival and happiness, and the birth of tool civilisation and indeed machine civilisation is its practical result. In ancient society, due to the limitations of science and technology and human practical ability, humans could only manufacture and use simple stone/iron tools; for human civilisation, reaching the level of machine civilisation was still difficult. With the advent of capitalist society, human S&T have made breakthroughs, inventions and creations have emerged one after another, factories generally use advanced production machines, capital and the state have maintained a strong interest in and large-scale investment in machine research and development and innovative applications. The concern of capital is profit, and machines can achieve greater efficiency and more profit. The concern of the state is strength, and machines can bring about stronger national strength and competitiveness.
In the era of the industrial revolution, the relationship between man and machine has two sides and contradictory tensions: on the one hand, the advancement of machines brings better productivity, which can liberate and enhance human labour; on the other hand, the widespread use of machines reduces It affects the status and bargaining power of workers, and causes workers to hate machines. There are inherent dualities in machine civilisation: it is both supplementary to human labour and a substitute for human labour. When the function of the machine is relatively simple and mainly reflects auxiliary functions, human beings love and accept the machine. When machine functions are complicated and anthropomorphised to partially replace human labour, the machine itself becomes a kind of ’new labourer’, a competitor of human labour; finding the machines distasteful, humans then reject them.
ChatGPT is of course the new darling of capital and an advantage lever of national competition, but for ordinary workers and their social relations, it causes serious misgivings and panic, holding existential fear of being replaced, inspiring resentment and rejection.
According to humanist logic, man is the ’measure of all things’ and legislator over nature. Man has rights of invention and sovereignty over labour tools, including machines. There is a fictive ’master-slave relationship’ between man and machine, whereby machines are servitors, ’silent slaves’ to human interests. Ethical dominance of human beings is established in this sense; human beings must ensure that their dominance will not be subverted or voided in ethical or legal terms. Given the duality of machine civilisation, people need to give full play to its auxiliary functions and restrain its subversion and countermeasure risks. Making good use of AI is a blessing for mankind that, if not used well, will become a force subversive of human existence.
AI’s threats to humanity
Behind all machine civilisation is emulation of human rationality, different degrees of which will create AI of different strengths and perfections. According to the degree of imitation, AI can be categorised as ‘weak’, ‘strong’ or ‘super’. In this pedigree, the machines’ emulation of human rationality gradually escalates and surpasses it, the threat to human existence gradually rising:
weak AI is just a simple, fragmented imitation and application of human rationality by machines; its threat to human society is completely controllable.
strong AI involves multi-functional integration of human rationality, and is the mainstream of AI development’s current stage;
ChatGPT roughly belongs to this stage, already integrating human rationality and a degree of independent learning and practical ability, but insufficient to completely replace humans.
super AI belongs to the perfected, extreme form of AI development. It was first proposed by philosopher and AI ethics expert Nick Bostrom, who defined it as ‘almost surpassing the intelligence of the human brain in every field’.
This definition subsumes some key points:
super AI is a general transcendence of human rationality, and it surpasses the superiority of human rationality. The overall limit is a kind of ’superhuman’ intelligence;
it ’almost’ suggests that superhuman intelligence may not have achieved superhuman advantages in some specific fields exclusive to humans;
with advantages in all aspects from labour to social management, the status and dominance of human legislators may be subverted and shaken
Judging from current technical forms of AI, the era of weak AI has passed, that of strong AI is unfolding, and that of super AI constitutes the future picture. ChatGPT itself is a technical system of strong AI thay has serious technical advantages and application prospects in intelligent dialogue, text creation and social management. Via directed accumulation, iteration and optimisation in this direction, it can move step by step towards super AI. What people fear is its future trend towards the development of super AI. Due to the collusion of capital and the state, this will inevitably speed up, and the threat of AI to humans will inevitably become more realistic and challenging.
Restricting ourselves to the application of ChatGPT, its threats and negativity are predictable and confirmable:
it replaces a wide range of jobs: intelligent customer service, coding, image production, text scripting, manuscript drafting, news writing, legal consulting, basic teaching, etc.;
it induces academic plagiarism, such as using this tool for paper writing and plagiarism;
the teaching field’s dependence on machines and degradation of its practical ability will cause serious damage and weakening to students' learning ability, practical ability and moral practical ability
cultural security and moral security at the social level, including factors such as algorithmic dictatorship, radical thought penetration, and the proliferation of politically correct culture.
These threats are real and will inevitably become more prominent with the wide application of this procedure.
AI development must uphold the status of human legislators and human dominance, and insist on the premise and foundation of human existence and human value. This baseline has the normative status of fundamental law and should bind all investors, tech developers and social managers.
Are Intelligent Robots a ’New Man’?
In the theocratic age, human reason was, in the name of faith, subjected to the reason of God. In a democratic age, human reason legislates for self and nature, creating modern civilisation in the process. As product of and assistant in the liberation of human rationality, however, the development of machine civilisation is gradually embarking on a road of human alienation: constructing a new form of rationality that emulates and even surpasses human rationality beyond human existence.
AI’s development and evolution from weak to strong, thence to super strong, is a remarkable trajectory of human rational self-liberation and self-alienation. Can humans, as rational lawmakers, control AI forever? Will the algorithm of intelligent robots be ’smart enough’ to break through human rational defences? How to break the technical tyranny over the ’majority’ of the ’minority’ behind the algorithm? Do algorithmic discrimination and platform hegemony constitute new sources and foundations of human inequality? Will intelligent robots have a disruptive impact on human labour, existence, and social relationships? Will intelligent robots eventually become the rulers of mankind and completely subvert the ’master-slave relationship’ between humans and machines? Can robots dominate the world as a kind of ’new man’? These ultimate questions have strong ontological significance and philosophical challenges.
For the ethical and legal challenges of super AI, human society must have a sufficient sense of crisis and a dialectical philosophy of technology, guide AI to always serve human beings with controllable developmentalism, and comprehensively regulate the threat:
at the level of international law, the UN and relevant international organisations should lead the formulation of ethical guidelines or conventions for the global development of AI to provide normative constraInts and guidance;
at the level of national law, countries should formulate their own ethical guidelines and policy frameworks for the development of AI and legal mechanisms, and pay attention to international coordination;
At the level of industry self-regulation, companies with industry influence should strengthen the formulation of ethical standards and the construction of compliance systems, and promote the formation of international industrial standards for AI;
the level of family and individual self-regulation , every family and everyone should understand the pros and cons of AI, pay attention to balance and self-discipline in the application of work and life, and maintain the baseline of human beings and the basic ability of social practice.
A rational product of human beings, AI must serve human interests and be legislated and led by humans in order to ensure that it conforms to human ethics and promotes better survival and life for human beings. In the face of the extensive and profound challenges brought by ChatGPT, human society, countries, groups and everyone e\have the responsibility to think about how to adapt, regulate, self-discipline and persevere. Problems created by people can in the end only be solved by them. Intelligent robots are not a kind of ’new man’: they remain projections and reflections of human rationality, and must develop within its moral and institutional limits. Provided it is oriented to people, rather than to profit and power, AI’s development can embark on a bright and correct path.
(Originally published in Hong Kong AM730, February 13, 2023)
2022年底，美国人工智能研究实验室Open AI推出了一款新型自然语言处理工具ChatGPT，引发商业、技术、社会、伦理、法律诸多层面的激烈反响。这是人工智能技术发展的一个里程碑，也是对人类存在与人类伦理的重要考验和挑战。人工智能本质上属于一种机器文明，是人类理性应用的结果。但是当这种定向和拟人化的机器文明充分积累和发展到一定程度时，就有可能对人的劳动及其存在构成替代，对人类社会的基础结构、交往方式及伦理/法律体系造成颠覆性影响。
人工智能的技术发展与商业应用，既是资本逐利的必然趋势，也成为国家间实力竞争的新战场。中美都是人工智能技术大国，也是商业应用大国，具有新冷战性质的高科技范畴的对抗竞争必然波及这一领域。美国制定有关法律和政策极力推动人工智能发展并抢占制高点。ChatGPT的出现，是资本与国际竞争共同作用的结果。面对人工智能的发展及其带来的广泛挑战，中国应当做好技术储备、伦理引导和法律规制，确保自身在人工智能领域的技术优势和道德合法性。
机器文明的两面性
人猿相揖别，人与自然动物的关键区别在于制造和使用工具，即理性劳动。人的理性应用于劳动，根本目的在于人的生存和幸福，实践后果是催生了工具文明乃至于机器文明。在古代社会，由于科学技术与人的实践能力的局限性，人类仅仅能够制造和使用简单的石器/铁器工具，人类文明尚难以达到机器文明的程度。进入资本主义社会以来，人类科学技术取得突破性进展，发明创造层出不穷，工厂普遍采用先进生产机器，资本和国家均对机器研发和创新应用保持浓厚兴趣和大规模投入。资本的关切是利润，机器能够实现更高效率和更多利润。国家的关切是实力，机器能够带来更强的国家实力和竞争力。
在工业革命时代，人与机器的关系已经出现两面性和矛盾张力：一方面，机器的进步带来更好生产力，对人的劳动有解放和提升的功能；另一方面，机器的普遍运用又降低了劳动者的地位和议价能力，造成劳动者对机器的仇视。机器文明存在固有的两面性：既是人类劳动的辅助者，也是人类劳动的替代者。在机器功能较为单一且主要体现辅助功能时，人类对机器是喜爱和接纳的。当机器功能复杂化和拟人化达到对人类劳动的部分替代时，机器本身成了一种“新劳动者”，成了人类劳动的竞争者，人类对机器就会产生厌恶和排斥。ChatGPT当然是资本的新宠，也是国家竞争的优势杠杆，但是对普通劳动者及其社会关系而言却引发了严重猜疑和恐慌，内含着被替代的生存性恐惧，由此激发出某种幽怨和排斥情绪。
按照人文主义逻辑，人是“万物的尺度”，是自然立法者，人对劳动工具包括机器享有发明权和主权，人与机器之间是一种拟制的“主奴关系”，机器是服务人类利益的“不说话的奴隶”。在此意义上，人的主导性伦理得以确立，人必须在伦理和法律上确保这种主导性不会被颠覆或架空。对机器文明的两面性，人需要发扬其辅助效用，抑制其颠覆反制风险。用好人工智能是人类之福，用不好就会成为人类存在的颠覆性力量。
人工智能对人的威胁
一切机器文明的背后都是对人的理性的模仿，但模仿程度不同就会产生不同强度和完善度的人工智能。按照模仿程度，人工智能可以分为弱人工智能、强人工智能和超人工智能。这是一个机器对人的理性模仿逐步升级甚至超越的发展谱系，也是对人的存在威胁性逐步提升的谱系。弱人工智能只是机器对人的理性的简单、片段化的模仿和运用，对人类社会的威胁是完全可控的。强人工智能涉及对人的理性的多功能整合，是人工智能目前阶段发展的主流，ChatGPT就大致属于这一阶段。强人工智能已经具有人类理性的集成性和一定程度的自主学习与实践能力，只是尚未完备到对人的完整取代。
超人工智能则属于人工智能发展的完善且极端化的形态，最早由英国哲学家兼人工智能伦理专家尼克·博斯特罗姆（Nick·Bostrom）提出并给出了如下定义：
“一种几乎在每一个领域都胜过人类大脑的智慧。”
这一定义有着非常敏感和关键的知识信息：
其一，超人工智能是对人的理性的概括性超越，跨过了人的理性优越性的总体极限，是一种“超人”智慧存在；
其二，“几乎”提示了超人工智能可能在某些人类专属的特定领域尚未取得超人优势；
其三，超人工智能阶段的机器对人具有从劳动到社会管理各方面的优势，人类的立法者地位和主导性可能遭到颠覆和动摇。
从人工智能的当前技术形态来看，弱人工智能的时代已经过去，强人工智能的时代正在展开，超人工智能的时代构成未来图景。ChatGPT本身是一种强人工智能的技术系统，在智能对话、文字创作与社会管理方面有着显著的技术优势和应用前景，且经由这一方向的定向积累、迭代和优化，可以朝着超人工智能方向逐步发展。人们所恐惧的就是它朝向超人工智能发展的未来趋势。由于资本和国家的共谋，这一趋势必然加速发展，人工智能对人的威胁必然更具现实性和挑战性。
仅从ChatGPT的应用来看，其威胁和负面性也是可预测和可见证的：
其一，对较为广泛的工作岗位的替代，包括智能客服、代码开发、图像制作、文字脚本、文稿起草、新闻写作、法律咨询、基础教学等；
其二，学术领域的诱导抄袭，如利用这一工具进行论文写作和剽窃等；
其三，教学领域的机器依赖和实践能力的退化，对学生学习能力、动手能力和道德实践能力都将造成严重损害和弱化；
其四，社会层面的文化安全与道德安全，包括算法专制及激进思想渗透、政治正确文化泛滥等因素。这些威胁性是现实的，必然伴随这一程序的广泛应用而更加凸显。
人工智能的发展必须坚持人的立法者地位和人的主导性，坚持以人的存在和人的价值为前提和基础。这一底线具有根本法则的规范地位，应当拘束一切投资者、技术开发和社会管理者。
智能机器人是一种“新人”吗？
在神权时代，人的理性曾经以信仰名义交付给神的理性。在民主时代，人的理性为自我和自然立法，由此开创了现代文明。但作为人的理性解放的产物和助手，机器文明的发展却在逐步走上一条人的异化的道路，即在人的存在之外构造出一种对人的理性的极致模仿乃至超越的新理性形态。
人工智能由弱到强再到超强的发展演变，就是人的理性自我解放又自我异化的显著轨迹。人类作为理性立法者，能够一直控制人工智能吗？智能机器人的算法会“智能到”突破人的理性防御吗？算法背后的“少数人”对“多数人”的技术专制如何破解呢？算法歧视和平台霸权是否构成人类不平等的新的来源和基础呢？智能机器人会对人类劳动、存在及其社会关系产生颠覆性影响吗？智能机器人会最终成为人类的统治者从而彻底颠覆人与机器的“主奴关系”吗？机器人能够作为一种“新人”而主导世界吗？这些终极提问都具有极强的存在论意义和哲学挑战性。
对超人工智能的伦理与法律挑战，人类社会必须有充分的危机感和辩证的技术哲学观，以可控的发展主义引导人工智能始终为人类服务，并对人工智能的威胁性加以综合规制：
其一，国际法层次，联合国及有关国际组织应主导制定人工智能全球发展伦理准则或公约，提供规范性约束和指导；
其二，国家法层次，各国应当制定本国人工智能发展的伦理准则、政策框架和法律机制，并注意进行国际协调；
其三，行业自律层次，具有行业影响力的公司应当加强伦理准则制定与合规体系建构，推动形成人工智能国际产业标准；
其四，家庭与个人自律层次，每个家庭和每个人都应当理解人工智能的利弊，在工作与生活应用上注意平衡与自律，维持人类的底线道德及其社会实践的基础能力。
总之，人工智能是人的理性产物，必须服务于人类利益并由人类进行立法和主导，才能确保人工智能合乎人类伦理并促成人类更好的生存与生活。面对ChatGPT带来的广泛而深刻的挑战，人类社会、各国、各群体及每个人都有责任思考如何适应、规制、自律和坚守。人产生的问题，最终只能由人来解决。智能机器人不是一种“新人”，仍然还是人类理性的投射和反映，也必须在人类理性的道德和制度限度内发展。只要坚持人工智能以人为本，而不是以利润和权力欲为本，人工智能发展就能走出一条光明而正确的道路。
（原载香港AM730, 2023年2月13日）