Author Yu Zheng

Introduction: Behind prosperity is a silent 'retreat of people'

The more wild the technology is, the more trapped people become

We are currently in a dizzying era.

Large models iterate on a weekly basis, generative AI kicks down the threshold for creativity, and autonomous driving is rewriting the meaning of 'mobility'. On the other hand, the situation of people is not so glamorous: delivery riders are precise to the second by algorithms - in 2021, multiple departments jointly issued guidance, clearly requiring platforms to fully listen to the opinions of trade unions and worker representatives when formulating and revising algorithms directly related to worker rights and interests, and publicly disclose them; The guidance for food delivery drivers directly proposes to relax the delivery time limit appropriately through "algorithm selection". But the reality is that this "pick and take" approach has uneven implementation, and the countdown pressure on the rider's end has not truly dissipated. White collar workers are forced to produce triple reports with the help of AI assistants; The creator's work was taken to "feed" the big model for free in the copyright ambiguity zone - Anthropic has reached a settlement of about $1.5 billion for training models on unauthorized use of copyrighted works, paying consideration for about 500000 works and destroying infringing copies. This is currently one of the largest settlements in similar lawsuits.

This is not the fault of the technology itself. The real problem is that the logic that dominates the direction of technology is too singular: the infinite proliferation instinct of capital. When the first landing point of technological progress is always "reducing labor costs" and "improving capital return rates", rather than "liberating people's time" and "enriching people's experiences", the era of intelligence has fallen into a profound paradox: the more powerful the tools, the cheaper, more replaceable, and less needed we as "people" are——More precisely, the IMF estimates that about 40% of global jobs are affected by AI, with half of them being complementary rather than substitutes; The World Economic Forum's 2025 report estimates that AI will create approximately 11 million jobs and replace around 9 million by 2030. The evidence that technology "makes people cheaper" in the "substitution" aspect is not clean; The real question is whether the costs saved by AI have become disposable time for more people or just a steeper curve on shareholder reports.

0.2 True question: Who is leading and serving whom

Sam Altman once appeared in a congressional hearingProactiveSupporting the establishment of registration, disclosure, and licensing systems for cutting-edge AI systems - this is more powerful than vaguely acknowledging that "the trouble will not be small".But what he didn't explain clearly is that the problem may not necessarily lie in the technology itself, but more likely in who is leading the technology, serving whom, and distributing the benefits according to what logic.

So, a fundamental question surfaced: in the era of intelligence, is humanism taming capital, or is capital using the shell of technology to achieve the ultimate contraction of human civilization?

1、 Triple Alienation: How Capital Logic "Dismantles" Human Nature in the Intelligent Age

1.1 De Dignification of Labor: You Become AI's' Data Feed '

In the past, assembly line workers were alienated as part of machines; Today, knowledge workers have been transformed into AI's "data feed" and "validation workers".

Every like you give on social media is providing free training data for recommendation algorithms; Every modification you make to generate text for AI is helping the big model learn how to be more human like. But this part of 'data contribution' is almost not reflected in your compensation. Capital, in the name of 'technological progress', collects implicit labor from people for free, and then uses' efficiency improvement' as an excuse to drive explicit positions out of the market. According to McKinsey's 2025 survey, 51% of companies have reduced the demand for entry-level positions due to AI - not by eliminating labor, but by breaking it down into invisible fragments and allowing them to be stored for free. The dehumanization of labor is quietly accomplished in this way.

In the domestic context, this logic is being partially recognized: the Interim Measures for the Management of Generative Artificial Intelligence Services require that training data sources be legal and respect intellectual property rights; The "Artificial Intelligence Generated Composite Content Identification Measures" implemented on September 1, 2025 clearly stipulate the obligation of explicit identification+metadata implicit identification, and prohibit malicious deletion, tampering, and concealment of identification. But these solutions are about "whether it can be used", not "how to divide it after use" - the attribution of profits is still blank.

1.2 Decapitalization of Time: Your Attention Becomes an Auction Item

Internet companies are no longer competing for your money, but for your attention duration. The algorithm is carefully designed as an addiction engine that maximizes retention - every piece of information you brush is the version that has undergone massive A/B testing and is most capable of hijacking your emotions.In 2024, global social media advertising spending will be approximately $220 billion - this money is not buying your clicks, but the time you spend awake but unable to control yourself. The daily average social media usage time of global users is about 143 minutes, which is equivalent to more than two hours a day, and your attention is being silently auctioned off.

You are both a commodity and a consumer, but no longer the sovereign of your life. When 'killing time' becomes a billion dollar business, your 'mind at ease' becomes the biggest enemy of capital. The depersonalization of time has turned living into being harvested by algorithms.

There are corresponding governance tools in China: the Management Regulations on Internet Information Service Algorithm Recommendation issued by the Office of Cyberspace Affairs requires regular assessment and social supervision; Article 24 of the Personal Information Protection Law grants individuals the right to request and refuse explanations for automated decision-making. However, in reality, the cost of exercising these rights is much higher than the cost of collecting data on the platform, and the "feasibility" of the rights is insufficient.

1.3 Value de subjectification: What you think you want may be what algorithms make you think

The most ingenious trick of capitalism is to replace "social needs" with "market needs".

You have a need to express yourself, but the platform only pushes content that can trigger opposing emotions; You have a need for deep learning, but knowledge payment products only sell fast food alternatives that allow you to listen to a classic book in 30 minutes. Over time, you may not even be able to distinguish between what I really need or what algorithms make me think I need? According to a study by the Stanford team in 2026, 66% of respondents do indeed spend more time on high-value work, with this proportion reaching 80% among cutting-edge users. Technology itself can make people more free; The question is, who has intercepted this' freer 'dividend. When human desires are shaped in reverse by capital, human subjectivity quietly evaporates. The de subjectification of value is the most hidden and fatal layer in the triple alienation.

The combination of these three levels of alienation is the cruel paradox in the introduction: the more advanced the technology, the more "people" are compressed into a series of tradable, replaceable, and predictable parameters——Not because technology is necessarily like this, but because in reality it always carries the interests of the people who designed it.

2、 The cage of conspiracy: why the 'hand of the market' cannot save humanity

2.1 Winner takes all: Consumers vote with their feet, but have nowhere to escape

Some people may say that the market economy has a self-regulation mechanism, where consumers vote with their feet. If companies excessively exploit human nature, they will eventually be punished by the market.

This argument may have been partially valid in the industrial age, but in the intelligent age where network effects and data monopolies coexist, it has become a sophisticated lie. The training cost of AI big models is in the billions of dollars, and the data barrier is as high as a city wall. Once a company takes the lead in running the growth flywheel of "data model user", there is almost no possibility for newcomers to turn the tables.The single cost of training the global top tier basic models has exceeded hundreds of millions of dollars, which is not a game that small teams can play.

The market concentration has unprecedentedly increased, and consumers' "foot voting" has become "nowhere to escape" - if you don't use WeChat, it means you are disconnected from social networks; If you don't use a certain food delivery platform, it means that the number of available restaurants has decreased by half. The winner takes all, rendering the exit mechanism ineffective.

At the domestic level, the algorithm filing system is already in operation: as of the end of December 2025, 751 algorithms from 15 batches have been filed, and 5100 deep synthesis service algorithms from 14 batches have been filed. But the issue of filing is whether to register, not whether to be bound after registration - this is the real landing point of domestic rules: the list of obligations is very detailed, but the binding effect is still climbing.

2.2 Attention arms race: Every family is rational, but together they go crazy

Optimizing algorithms for individual enterprises to increase user time is a rational decision; But if all companies do this at the same time, it will cause a social attention grabbing and spiritual internal friction.

This is the "prisoner's dilemma" in game theory: each company is unwilling to take the lead in reducing algorithm addiction because it means user churn; The result is that all companies are accelerating the extraction of humanity, and no one can stand alone.The domestic "Clear Network Platform Algorithm Typical Problem Governance" special action has been cracking down on information cocoons, list manipulation, and big data manipulation - this indicates that regulators have seen the problems, but the special action is pulse like. Whether the platform has really lowered the weight of "retention maximization" after returning to daily operation still lacks a sustained and quantifiable tracking mechanism.

The tragedy of the attention arms race is that the more rational the participants are, the crazier the whole becomes.

2.3 Efficiency becomes absolute justice: Skeptics are labeled as' anti development '

When capital packages "cost reduction and efficiency improvement" as the sole measure of social progress, any doubts are labeled as "anti development" and "conservatism".

But honesty is needed here: efficiency itself is not fake. The proportion of employees using AI has increased from 30% to 76% (according to McKinsey data), and 66% of people do indeed allocate their time to high-value work - the question is not "efficiency should not exist", but "who should receive the efficiency dividend". If every minute saved by AI becomes a dividend for shareholders rather than leisure for workers, then 'efficiency' will indeed become the crushing stone.

Under this discourse hegemony, 'humanism' is reduced to 'sentimental romance', while 'efficiency' is deified as' rational conscience '. This is not a competition of diverse values, but a comprehensive suppression of discourse. Once efficiency and absolute justice are established, human nature can only retreat.

So, expecting capital to spontaneously 'discover conscience' is equivalent to expecting lions to actively eat grass. There must be an external, institutionalized, and coercive force involved to establish an insurmountable 'human barrier' for technology.

3、 Human nature dam: Lock three bottom lines for capital

3.1 The First Dam: Establishing the Property and Profit Rights of "Digital Labor"

Every click, conversation, and annotation you make is creating value for the AI system. This logic must be recognized from a legal perspective.

Article 53 (1) (d) of the EU Artificial Intelligence Act imposes transparency and assessment obligations on systemic risk models, which is only the first step; A more thorough direction is to establishPersonal Data Trust System——Include each individual's data assets in independent third-party custody. When using data to train models, companies must pay fair consideration to the data trust and distribute it to individuals based on their contribution.This is a natural extension from "rights" to "benefits", in line with the right to data portability and the right to oppose automated decision-making under Articles 15-22 of GDPR.

Article 24 of the Personal Information Protection Law in China has granted individuals the right to request and refuse explanations for automated decision-making; But whether there are alternative solutions after "rejection" and who will bear the cost of "explanation" are still ambiguous at present. The localization of data trust requires filling in these two gaps first.

This is not a utopia, but a digital extension of the fundamental principle of capitalism that 'labor creates value'. The property rights and income rights of digital labor must be recognized.

3.2 The Second Barrier: Breaking the Attention Cage with "Antitrust 2.0"

Traditional antitrust focuses on market share and price manipulation; But in the era of intelligence, attention monopoly and data monopoly are the real culprits.

It is necessary to legislate to prohibit algorithms from optimizing for the goal of maximizing addiction, force platforms to provide users with "non personalized recommendation" options, and limit the daily limit of user attention occupied by a single platform.California SB 976 has requested large platforms to provide non personalized recommendation options to underage users; The EU's Digital Markets Act imposes obligations on "gatekeeper" platforms such as interoperability and data isolation, and fines for violations can reach up to 10% of global annual revenue. Antitrust 2.0 is not about scale, but about the excessive occupation of human attention.

The domestic "Regulations on Algorithm Recommendation Management" have required the provision of a personalized recommendation option to be turned off - but in reality, this switch is hidden deep in the five or six layers of menus, and the recommendation quality drops sharply after turning it off, which is equivalent to indirectly punishing users who exercise their rights. The true 'Antitrust 2.0' requires making the close button a default item, rather than a hidden Easter egg.

3.3 Third Dam: Establish a mandatory pre procedure for "Technical Impact Assessment"

Any AI system involving large-scale deployment must undergo an independent committee's "Social Resilience Impact Assessment" before going live - not only assessing economic efficiency, but also evaluating its medium - to long-term impacts on employment structure, mental health, social trust, and cultural diversity.

But if the evaluation becomes a paper-based process, it will become another compliance cost. To make it really bite, four hard constraints are needed:

Firstly, the evaluators cannot all be insiders. It must include representatives of workers, users, and independent technical experts, and the list and conflict of interest statement should be made public.

Secondly, evaluate the grading of the objects. High risk and high impact scenarios - recruitment, credit, public welfare discretion, education evaluation - mandatory pre positioning; Low risk scenarios can follow simplified procedures. Not all models have to go through the same sieve.

Thirdly, the evaluation results are accompanied by clear constraints: limited scope, deadline for rectification, and suspension of online launch. It's not just about suggesting optimization, it's about not being able to proceed if the evaluation doesn't pass.

Fourthly, the conclusion and rectification records can be traced and questioned. Citizens and affected groups have the right to apply for reconsideration, rather than locking the assessment report in a drawer.

What is truly lacking in China is not the right itself, but the cost of appealing and the arrangement of burden of proof to enable the exercise of rights. If an ordinary person has to spend thousands of yuan and go to three departments to exercise the right to refuse once, then this right looks good statistically and does not exist in daily life.

Technology impact assessment is to implant a 'human brake pad' into capital's profit seeking instincts. Brake pads are not meant to stop the car from running, but to make the car run further and more steadily.

4、 Who will build the dam: the joint responsibility of the state, enterprises, and civil society

4.1 Country: shifting from "efficiency priority" to "resilience priority"

Over the past forty years, we have witnessed the miracle of 'development is the fundamental principle'. But the new hard truth of the intelligent era is that development without resilience is essentially accelerating towards a cliff.

The government should incorporate the social resilience index, which includes dimensions such as psychological capital, social trust, and cultural diversity, into the local governance assessment system, placing equal emphasis on GDP. It's not that we don't want GDP, but we can't let GDP be the only indicator that can speak. More importantly, the state should take the initiative to become the "last guarantor": when technological shocks lead to large-scale structural unemployment, it is not through "retraining" to shift the responsibility onto individuals, but through systems such as universal basic income (UBI) to provide individuals with a "no fear" safety net. Only those who are not afraid dare to innovate; Only a society without fear possesses true resilience.

4.2 Enterprises: Moving from "shareholder first" to "stakeholder co governance"

This should not be just fancy words hanging in ESG reports.

A "Humanities and Ethics Committee" should be established within the board of directors of the enterprise, with substantive veto power over product launches; Non financial indicators such as "user mental health impact" and "social trust contribution" should be included in the salary assessment of corporate executives. More fundamentally, companies need to redefine 'success' - not' how many jobs we replace ', but' how many people we create more dignified ways of working '. This is not a moral blackmail, but a rational calculation of long termism: a spiritually withered society cannot foster a sustainable growth market. Stakeholder co governance is not about decentralization, but about extending life.

4.3 Civil Society: Being the 'third eye' that never remains silent, alternative solutions must also be provided

Independent media, academic institutions, non-profit organizations, consumer alliances - these forces need to maintain continuous scrutiny and public discussion of technology beyond the collusion of capital.

Every exposure of algorithmic bias, every lawsuit against data black boxes, and every public debate on technological ethics are adding bricks and tiles to the bottom line of civilization in this era. It is important that the supervision of civil society cannot remain at the level of 'anger and criticism', but requires the ability to construct 'constructive alternative solutions' - not just to say' this is not right ', but to say' that may be better '. The third eye needs to see both the problem and the way out.

Only when these three forces work together can "humanism" transform from an empty slogan into an executable, supervised, and iterative social operating system.

5、 The dignity of civilization lies in refusing to be an auxiliary material for algorithms

5.1 Technology has never destroyed humanity, but humanity must be regained

Looking back at the history of human technology, every major breakthrough has triggered a similar anxiety: Will printing destroy thought? Will steamers eliminate craftsmen? Will the Internet disrupt the community?

It has been proven that technology has never destroyed humanity, but in every wave of technology, humanity must go through an arduous battle of "reclaiming" - reclaiming dignity from the grip of capital, meaning from the worship of efficiency, and freedom from the training of algorithms. The weaning of civilization is to cut off the dependence on the logic of unlimited capital growth, and to make people the end rather than the means.

5.2 This battle has just begun

The outcome of this battle in the era of intelligence is not yet predetermined.

If we continue to allow capital to be driven by technology and humanityFish MeatWe will eventually slide towards the chilling prophecy in science fiction: humans will become the "bio materials" floating in the ocean of algorithms, providing endless training data and attention fuel for AI systems, but can no longer remember the original meaning of "happiness".

But if we are clear headed, united, and have institutional imagination, we have the potential to embark on a new path - a symbiotic path that utilizes the power of capital to accelerate technological inclusiveness, while using the dam of humanism as capital to lock in the bottom line.

At that time, looking back at today, just like looking back at our predecessors who abolished slavery, established labor protection laws, and established universal suffrage, we will exclaim: we ultimately won that long and difficult 'battle to defend humanity'.

At this moment, this battle has just begun. The dignity of civilization is hidden in every moment of refusal to be fully analyzed by algorithms.