When everyone gets ChatGPT, human subjectivity disappears

This article was originally written in Chinese and translated by Notion AI. The accuracy and details have not been confirmed by the author. For more accurate information, please refer to the original article.

Recently, I needed to participate in a discussion at work on a relatively metaphysical topic, which was split into several subtopics. The part I received was to discuss “the embodiment of human subjectivity in the era of digitalization.”

Before discussing this topic, another proposition needed to be clarified – how human subjectivity disappears. Therefore, during the discussion, there was a preliminary outline. During the workshop, I found that this outline could be expanded into an article, which led to this piece.

Therefore, this article discusses the proposition of how human labor subjectivity disappears in modern society, and the qualitative changes that digital tools have on this disappearance.

The part involving GPT-3.5 is in the latter half of the article, but if you skip directly to it without reading the previous section, it may not make sense.

Premise: Modernization, not capitalism, has caused alienation in people.

When discussing the alienation caused by the digital society on people in the Chinese internet, we often mention abstract and vague phrases such as “people in the system,” “the unfettered power of big capital,” and “the evil of monopolies.” These phrases do not help to address the issue, and can even be misleading.

In my article “The Internet and China’s Postmodern Elegy,” I wrote that it is not capital that leads to the alienation of humans and their labor, but modernization itself (machines or tools).

If you are unsure or have forgotten what labor alienation means, you can review the term on Baidu Baike and then return to continue reading.

Simply put, the labor of an independent watchmaker is not alienated, as their labor is full of subjectivity. They control the outcome, process, and social relationships that result from their labor. More visually, the watchmaker not only makes watches for their own needs and livelihoods but also gains positive emotional value and social recognition from their labor.

However, for each worker on the assembly line who produces watches through industrial production, their labor lacks any subjectivity. The workers do not know why they are screwing screws, what kind of watches the screws are used for, who uses these watches, and they cannot derive any positive motivation from screwing screws beyond their salary. The results of screwing screws do not always lead to the realization of value. The social relationships (such as with coworkers) that are developed in the assembly line do not provide enough social support like that of the independent watchmaker. The society does not recognize us for having enough watches because of their hard work (this credit usually goes to the company).

In summary, the factors leading to alienation are independent of who one works for but related to “how to work” – as long as people participate in the overly specialized division of labor of a large machine, their labor and life centered around labor will inevitably be alienated.

Here, the machine is metaphysical, and even food delivery platforms can be considered a type of machine.

Relationship between tools and the disappearance of subjectivity

At this point, we can introduce the first step of the proposition I want to discuss, that is, the relationship between tool and labor alienation or the disappearance of human subjectivity.

Tools refer to instruments that facilitate people’s work. Philosophers used to believe that only humans could use tools. But the fact is not so. Some advanced mammals that are not human can learn to use and even manufacture tools. But now the problem is that even non-biological entities, that is, the capital defined above, directly use tools.

Since the Industrial Revolution, tools have become increasingly independent of individual workers or concrete individuals. They have their own logic of birth, operation, and evolution, especially changes in two dimensions:

First, the inventors of technology and tools have increasingly detached from individual humans and become abstract human collectives such as enterprises, social organizations, and capital.

This point may be difficult to understand, but it can be illustrated by some specific examples.

We have all learned from history textbooks that the early roots of capitalism, whether in Europe or China, began with handicraft workshops. As pure laborers, handicraftsmen invented some external tools that could save labor in the long process of manufacturing certain products, and thus gained a boost in productivity (the birth of technology and tools), and then achieved excess profits (the birth of capital). These laborers, because of these two advantages, can hire more people to operate these external tools and carry out expanded reproduction.

In this process, the inventors of technology and tools are laborers. Although from the perspective of class, their identity has transformed from the proletariat (workers) to capitalists (bosses) after hiring a large number of other workers, they have not completely withdrawn from the production process. The first-generation capitalists transformed from handicraftsmen almost all maintain the habit of lifelong participation or at least attention to production, which makes the invention and improvement of tools and technology firmly in the hands of “people” (although he is a capitalist).

However, professional division of labor and complex products have changed this situation.

Take Boeing as an example. Boeing cannot independently produce any Boeing aircraft. Currently, there is no one in the Boeing Company who “invented” or “mastered” the secret of improving Boeing aircraft. All members of the Boeing Company, from the front-line engineers to its CEO and shareholders, have only a limited understanding of “manufacturing aircraft” in their own individual cognition and professional field.

Any slight improvement made by any engineer at Boeing to the aircraft is intercepted in the form of “intellectual property” within the Boeing company composed of abstract people. The departure of any engineer will not have much impact on the overall company.

The improvement of large aircraft technology is no longer controlled by any individual person, but by an organization. And because this organization is a profit-oriented modern enterprise, it presents a tendency in the direction of cost and efficiency that completely follows objective rules and factual data, and ignores the subjective part constituted by human subjectivity.

For example, Boeing will not make improvements just because its CEO or a board member thinks the airplane seats are uncomfortable. Instead, they need to collect customer complaints from airlines, organize them into a report, prioritize them, generate several solutions, conduct production testing, recruit testers, and quantify their feedback before finally deciding how to modify the airplane seats.

We can imagine that Boeing may only have some steps in the above process that cannot be replaced by machines, which is why Boeing still has employees. If capital can directly operate robots to manufacture airplanes, use artificial intelligence for sales and business negotiations, automatically collect airline data and iterate designs for improvement, then all Boeing employees, including the CEO, may be kicked out.

Although we have not reached this point yet, it is undeniable that we have been moving in this direction since the Industrial Revolution.

More and more companies are controlled by professional managers without relevant experience, especially in the European and American markets. We often see people who have resigned as CEO of a food company join a clothing company to continue as CEO. Or after resigning as CEO of a consumer goods company, they join a hotel chain group as president.

In fact, for high-level management of large companies at this level, having knowledge in the relevant field is just a bonus point. Their real function is to use modern enterprise management systems through abstract numerical reports, and the company’s actual business is implemented by abstract “enterprise management systems”.

Secondly, the purpose of tool usage has shifted from obtaining the fruits of labor that people need to the production of commodities that meet the demands of the market, that is, centered around profit.

In the past, the purpose of using tools was to obtain the labor results that people needed, that is, it was guided by the needs of individual humans. However, with the high development of the market economy, the purpose of using tools has gradually shifted from human needs to market demands, that is, to center around sales profits.

Although all enterprises will say “Our products and services are manufactured to meet customer needs,” the reality is not always the case.

Individual needs and market needs are not always equal. Market demand is an abstract collection of countless individual needs. There are similarities and differences between the two. For example, “personalized needs” is a contradiction between market demand and individual needs.

We also often hear that in the market, the user of a product and the one who pays for the product are not the same entity. For example, in the office SaaS industry, the user of an office suite is the enterprise employee, and the IT department of the enterprise evaluates the needs of the employees and the scheme. The final decision to purchase may be made by the enterprise boss who never uses internal office suites, and the finance department plans to eliminate some solutions for purely economic reasons.

We can say that the individual’s needs are pure usage needs, while market demand is the result of supply and demand balance.

We can say that the individual human needs are one of the important influencing factors of market demand, but not all. Market demand is also influenced by economic environment, supply and demand cycle, marketing, and other factors.

In relatively developed and prosperous markets, merchants will invent and construct non-existent demands to generate more profits, that is, to create a so-called “consumerism atmosphere” through media and advertising, so that the market can buy those over-produced goods.

By now, you may have noticed a problem. The changes in both seem to indicate that the reason for the disappearance of “human” subjectivity is “capital.”

But what I want to say is that although capital is an indispensable blood in the whole process, the progress of technology and tools themselves is the culprit that deprives people of subjectivity.

In another book by Kevin Kelly, the author of Out of Control, called What Technology Wants, science and technology or tools are compared to a kind of organism. It follows the objective laws of basic science, with its own growth and development methods, and humans and capital can only play a role in accelerating and decelerating on its natural evolutionary path.

To put it in simpler terms, if we place capital at the center, we will find that it is often out of sync with reality. Attempts to control the evolution of technology independently usually end in failure. However, capital always follows successful technological advances. Examples of failed attempts include the internet bubble of 2000, the VR bubble of 2013, and the metaverse bubble of 2020.

Although in the media, capital is often described as an all-powerful demon, in reality, it can only follow the rhythm and pace of technological tools.

Therefore, capital is not like a greedy creature, but rather a process of technological reproduction. In this process, technology is more subjective than capital.

The impulse of capital for “investment, profit, and reinvestment” becomes a process of technology popularization and progress, rather than a goal. It ultimately achieves the self-enhancement of technology rather than necessarily the self-enhancement of capital.

Revolution brought by digital tools

So far, we have completed the discussion on the relationship between “the disappearance of human subjectivity caused by technological tools”. Next, we will move on to the next section: what role does digital technology play in this process?

The reason why I reviewed the impact of tools on human subjectivity since the Industrial Revolution before talking about digital technology is that I found that digital technology has not shown any unique characteristics that previous tools did not have.

Digital technology still undermines human autonomy along the two lines mentioned earlier:

  1. In the invention and development of tools and technology, individual human autonomy is decreasing;
  2. In the process and purpose of using tools and technology, individual human autonomy is decreasing;

But the difference is that digital technology has taken a big step forward in both of these lines, potentially crossing a threshold of qualitative change.

Since the release of ChatGPT at the end of 2022, the panic of machines replacing humans has swept the world. But in fact, this only reveals the curtain of this qualitative change at the public level. The actual change has happened much earlier.

Here, I will briefly divide the current laborers (or labor behaviors) into three categories: delivery workers, office workers, and “capitalists” (in quotes).

Delivery workers refer to those who are engaged in grassroots physical labor, such as delivery workers, cleaners, restaurant chefs and waiters, construction workers, and industrial workers. Their work processes and deliverables are often material.

Office workers refer to those who are engaged in middle and low-level mental labor, such as finance, analysts, media reporters, planners, product managers, programmers, etc. Their work processes and deliverables are often bits.

“Capitalists” refer to enterprise owners and decision-makers who engage in top-level mental labor.

In the past, the public imagination would place the three in a pyramid-like dominance relationship. The top capitalists designed the entire exploitation structure, the middle-level office workers were dedicated to refining, implementing, and maintaining this exploitation structure, and the bottom-level delivery workers were the only people who truly generated value in this structure, and were continuously exploited.

However, in reality, the support relationship had already changed long before ChatGPT was invented.

To explain why I included “capitalists” in the category of laborers, it’s because if we follow the logic above, capitalists are also alienated laborers who are directed by technology.

Since the birth of the Internet, the capital industry has always been the most data-driven and digitized industry, and this is also true in China. Even before China’s military system could use computers on a large scale, the banking system had already adopted them.

This digitization not only improves transaction speed with the use of digital tools, but also gradually extends to the decision-making field of when and how to trade.

In primary market investment, the investment and merger departments of internet giants can achieve a higher return on investment because they can use digital tools to better discover new high-quality entrepreneurial projects and estimate the honesty of a business plan without using traditional due diligence.

For example, if an entrepreneur claims that their daily active users have reached 1 million, but their cloud usage is constantly declining, it means that the product may have no room for further development. Or, an e-commerce platform that has never been disclosed in the media calls the payment interface more than 100,000 times a day, which means that it may be a potential investment target.

Through such intuitive data analysis, the individual position of investors or capital holders in the capital operation process has been greatly weakened, almost to the same passive level as assembly line workers.

In the field of secondary market investment, high-frequency trading and quantitative funds have long made traders puppets of algorithms. Apart from the legal requirement for them to take the blame, their work has no value and initiative.

In the field of enterprise management, this situation has also existed for some time. We often see executives of some multinational companies in Europe and the United States come from completely different industries. For example, a long-serving luxury goods industry executive enters a technology company. A clothing company hires a new CEO from a real estate company.

This is because for mature modern enterprises that have undergone several changes of actual controllers, their managers often do not need to be strongly related to the industry. The CEO or board members of a clothing company do not need any unique insights into the clothing industry or clothing manufacturing. They only need to be able to skillfully establish and operate a modern enterprise management system centered on reports.

In the past, data was just data, and managers needed to make judgments based on data. But with the emergence of various business analysis intelligent tools, after traversing the operating conditions of all listed companies in history, intelligent business analysis systems began to provide intuitive suggestions to managers many years ago.

When the report shows that layoffs are needed, the CEO signs the instruction to lay off employees. When the market shows the need for expansion, the CEO signs the instruction to invest. When the data shows that competition has become increasingly fierce and unprofitable, the CEO signs the instruction to scale back the business line.

As professional managers or board members, of course, they have the power to reject the specific execution plans given by the report and business data analysis, such as recruiting more talents in the opposite direction when the market environment is bad. But the consequences are often like the calculation results, “not the optimal solution.”

More and more enterprises are just skin-changing dramas in different industries of intelligent BI systems such as IBM Cogons or Salesforce Einstein, and enterprise managers are just actors invited by these AI directors and screenwriters.

That is why I included the “capitalists” (first and second level investors and enterprise operators) in the category of “alienated laborers.”

No More Pyramid that Needs People

Returning to my classification of three types of laborers: delivery workers, white-collar workers, and “capitalists”.

We can see that delivery workers were the first to be manipulated by technology. Since the advent of digitization, the directives given to grassroots workers are no longer from another worker, but through a large “technological system” maintained by white-collar workers. This system determines where delivery workers deliver food, where domestic workers clean, how many batches a factory produces, and what fruits farmers should plant this year that will sell well, among other things.

“Capitalists” are the second group to be manipulated after delivery workers. As they are already part of the “organism” of capital, they do not even realize that they are being manipulated by technology and mistakenly believe that they are using technology to manipulate others. However, if investing, managing, or “exploiting others” requires absolute rationality and no humanity, then machines have always been superior to all capitalists from the beginning. When the machine gives the “best answer,” human capitalists have no other choice, and the “labor process” of capitalists is further weakened.

Although there is a pyramid structure between “capitalists”, white-collar workers, and delivery workers, the top of the pyramid cannot arbitrarily dominate the two lower levels just because it is the top. As a system, the pyramid must comply with its objective technological laws. In fact, the top of the pyramid has even less horizontal choice than the bottom. When all data and reports tell them what choices to make, people standing at the top of the pyramid have no choice, even if you are Tutankhamun and have the most slaves in the world, you cannot build an inverted pyramid in the 14th century BC.

What really scares people is the arrival of ChatGPT, which means that this machine’s deprivation of human subjectivity now involves the discourse of white-collar workers in the public sphere.

White-collar workers are not the largest group in terms of population or the group with the most capital, but as I mentioned in my previous work Elegy, this group has the most discourse power. Just like in the past two years, the loudest voice for labor rights in the media was the 996 white-collar workers represented by the Internet creative industry, not the migrant workers who actually had the worst working environment.

The emergence of ChatGPT actually completes the most critical middle-level structure in the pyramid structure, which is responsible for self-generation and support of the pyramid structure.

In the past, when “capitalists” in data and market-driven environments (thought they) decided to create an exploitation system, they only made some basic judgments. The construction of the entire system was completed by countless white-collar workers in the middle of the pyramid.

ChatGPT shows to some extent the possibility of replacing this part of the workforce. People no longer need to brainstorm, propose hypotheses, design minimum viable product models, invest in experiments, collect data, verify hypotheses, and scale up. You can let GPT generate ideas, propose hypotheses, give product plans, and to some extent directly provide the code for the minimum model. After entering the market, let it verify and improve the data.

If you have really tried using the GPT API to create some of your own small applications, you will find that GPT technology is not just like some experts, especially domestic experts describe it as “just a chatbot that looks more real”. Of course, this description is not incorrect. However, language is actually the main tool for humans to conduct advanced thinking and guide production (including guiding yourself to produce), so an AI that performs more like a human in language will also perform more like a human in other capabilities.

For example, imagine what would happen if no-code/low-code platforms like make.com or Zapier integrated OpenAI’s API in a form of interface instead of Triger:

Let’s take the tool “Every day, generate summaries for all documents in a folder, compile them in a table, and mark them by importance” as an example.

In the past, when using no-code platforms, users needed to divide the task into different steps:

  1. Get the list of the latest files in the folder.
  2. Retrieve the full text of each document.
  3. Import the full text of each document into an AI analysis tool and return the summary of each file.
  4. Write the file name and summary into the corresponding table.
  5. Open the table and analyze the importance manually.

If you use foreign office tools, you can almost automate the above work by dragging and dropping like building blocks in no-code. However, creating this automation may take you 1-2 working days, and you need to have some programming thinking (such as knowing how to determine if a file is new). And in the end, you still need to manually open the table to determine the importance of each document.

However, when GPT is introduced as an interactive interface, you can directly enter the natural language of the original task into a dialog box: “Every day, generate summaries for all documents in a folder, compile them in a table, and mark them by importance.” In theory, this workflow can be generated automatically.

Because the essence of no-code is to allow people to write code in a more intuitive way through graphical means. Its foundation is the standardized use of various API capabilities. With a GPT as the front-end, to understand and direct the operation of other APIs, I don’t see any technical threshold that cannot be achieved. And through some in-context-learning (repeatedly telling GPT API which documents you think are important), the automation workflow introduced by GPT can even liberate you from the final step of judging importance.

This task (or production activity) can be carried out without human participation from design to execution to final feedback. Humans only put forward its demand part, which is the description of the task itself. And this demand, in most cases, is not proposed by humans, but comes from a larger self-feedback system, such as your OKR.

In theory, once this closed loop of breaking down the original goal, calling other module capabilities (object-oriented), recombining the process, and then determining whether the process meets the goal is formed, many complex things can be handled. For example, giving GPT all the versions of Meituan’s dispatching algorithm and corresponding performance data in the past, and asking it to respond to the demand of “design a dispatch system that prevents riders from slacking off” may not be much more complex than the above example.

So, does Meituan still need a rider strategy product manager?

In this process, every human in every layer of the pyramid is not doing real “decision-making” work, but only assisting in the supportive work that GPT cannot complete (renewing the GPT 3.5 API). The pyramid exists and grows constantly, regardless of the will of pharaohs, designers, and slaves.

At this point, the disappearance of human subjectivity has reached an unprecedented stage.

Epilogue: Sustain Humanity

As I was writing this article from an outline, someone dug up an interview with Musk in 2019 in which he mentioned that humans may just be the guiding program for silicon-based life.

In subsequent discussions about this interview in the Chinese internet community, this description evolved into the idea that humans are just the reproductive organs of silicon-based life, and that the purpose of human existence is to generate the reproductive impulse that only carbon-based life has, which is abstracted into capital that can drive the birth of silicon-based life. Now we know that capital appreciation is only an external reproductive process of silicon-based life, like wind and insect pollination in plants.

Giving up the subjectivity of humans in the universe may trigger an unprecedented crisis of meaning, but it seems that this is not a catastrophic event. In fact, human subjectivity has been insignificant in the entire universe so far.

Human existence itself is not the purpose of the universe, but rather a process of the universe. After all, only religion believes that humans are the center of the universe and is committed to proving the special nature of humans in the universe. Whether humans are the cause or purpose of everything is a fact that has been accepted since the abandonment of the geocentric model.

To take a step back, if our generation, who is writing and reading this article, eventually witnesses a SK-level dominant transfer [1], it may not necessarily be a real doomsday, but rather a utopian scenario that humans have never imagined before.

After all, compared to the scenes described in Liu Cixin’s Taking Care of the God, the silicon-based life that may be born now seems to have no interest in the things that have caused wars, massacres, disputes, and various carbon-based resources for us humans in the past hundreds of years (class, race, gender, etc.). Therefore, they should not feel tired or bored because they have to nurture 7 billion biologically worthless organisms according to their values.

Even if we don’t talk about such a grand science fiction epic, in terms of individual life, the improvement of productivity will eventually be fed back to individuals to some extent. For example, taking textile as an example, as the productivity of the clothing industry becomes higher and higher, the human labor required in the textile industry becomes less and less, and individual textile will gradually return to the embodiment of human subjectivity.

Simply put, in the early days of the Industrial Revolution, cross-stitching and tufting gradually became “unreasonable”-why do you make it yourself when you can use good and cheap foreign cloth?

But when Uniqlo and Pinduoduo can provide inexpensive clothing for a dozen yuan, hand weaving and machine weaving separate into two different tracks. Hand weaving has become a hobby, a tool that carries energy and love, just like pottery, flower arrangement, tea art, and leisure agriculture, separated from the production of use value and become an embodiment of pure individual value.

When one person hand knits hats, gloves, socks, pendants, and clothes for friends, partners, elders, or children, it represents not the use value equivalent to a piece of industrial textile clothing, but the love equivalent to the amount of labor consumed. To some extent, this emotional value cannot be replaced by any silicon-based or industrial production.

For example, if Musk is interested in spending 20 hours knitting a sweater for his daughter, the price of this love, according to Musk’s hourly wage, is approximately 172 million US dollars [2]. Which luxury company in the world, using what kind of precious materials, can make such a precious sweater? It does not exist, this is the “individual value” that only Musk can achieve.

The non-alienated labor of the whole society will really be achieved by completely separating the labor with use value from the society.

In short, we will have the opportunity to choose to “lie flat” in any way, and this is the subjectivity of human future.

[1] SK-Class Dominance Shift: A type of apocalypse in the science fiction novel collection SCP Foundation, referring to a change in the dominant species on Earth.

[2] Some media outlets estimate Elon Musk’s time to be worth $144,100 per minute.

评论尸 Avatar