In an era of agent explosion, how should we cope with AI anxiety?

Author: XinGPT

AI is yet another movement for technological equality.

A recent article titled "The Internet is Dead, Agents Live On" went viral on social media, and I agree with some of its points. For example, it points out that DAU is no longer a suitable metric for measuring value in the AI ​​era because the internet has a mesh structure with decreasing marginal costs; the more users, the stronger the network effect. In contrast, large-scale models have a star topology, where marginal costs increase linearly with token usage. Therefore, compared to DAU, token consumption is a more important metric.

However, I believe the conclusions drawn from this article are significantly flawed. It describes tokens as a privilege of the new era, arguing that whoever possesses more computing power has more power, and that the rate at which tokens are burned determines the rate of human evolution; therefore, token consumption must be accelerated continuously, or one will be left behind by competitors in the AI ​​era.

A similar viewpoint appeared in another viral article, "From DAU to Token Consumption: The Power Shift in the AI ​​Era," which even suggested that each person should consume at least 100 million tokens per day, preferably 1 billion tokens, otherwise "those who consume 1 billion tokens will become gods, while we are still just human beings."

However, few people have seriously calculated this cost. According to GPT-4o's pricing, the cost of 1 billion tokens per day is approximately $6,800, which is close to 50,000 RMB. Just how high-value work would it be worthwhile to operate an agent at such a cost long-term?

I don't deny the efficiency with which anxiety spreads in the dissemination of AI, and I understand that the industry is "exploding" almost every day. But the future of Agents should not be reduced to a competition of token consumption.

To get rich, you do need to build roads first, but excessive road construction only leads to waste. The 100,000-seat stadiums that rise from the mountains of western China often end up as debt relief projects overgrown with weeds, rather than centers for hosting international events.

Ultimately, AI aims for technological equality, not the concentration of privilege. Almost all technologies that truly change human history go through a process of mythologizing and monopolizing before finally achieving widespread adoption. The steam engine wasn't exclusively for the nobility, electricity wasn't only supplied to royal palaces, and the internet didn't serve only a select few companies.

The iPhone revolutionized communication, but it didn't create a "communication elite." For the same price, the average person has the same device as Taylor Swift or LeBron James. That's technological equality.

AI is following the same path. What ChatGPT brings is essentially the equalization of knowledge and abilities. The model doesn't know who you are, nor does it care who you are; it simply responds to questions according to the same set of parameters.

Therefore, whether an agent burns 100 million or 1 billion tokens doesn't inherently determine its superiority. The real difference lies in whether the objective is clear, the structure is reasonable, and the questions are correctly posed.

The more valuable capability lies in generating greater impact with fewer tokens. The upper limit of agent usage depends on human judgment and design, not on how long a bank card can sustain its consumption. In reality, AI rewards creativity, insight, and structure far more than it rewards mere consumption.

This is precisely the equality at the instrumental level, and it's where humanity still retains initiative.

How should we deal with AI anxiety?

Friends majoring in broadcasting and television were greatly shocked after seeing the video released after Seedance 2.0 was launched. "In this way, the jobs we studied, such as directing, editing, and photography, will be replaced by AI."

AI is developing too fast, leaving humanity utterly defeated. Many jobs will be replaced by AI, and this trend is unstoppable. When the steam engine was invented, coachmen became obsolete.

Many people are starting to worry about whether they can adapt to the future society after being replaced by AI, even though we rationally know that when AI replaces humans, it will also bring new job opportunities.

But the speed of this replacement is still faster than we imagined.

If AI can do better with your data, your skills, even your humor and your emotional value, then why would a boss choose a human over AI? Or even what if the boss were AI? So some people lament, "Don't ask what AI can do for you, but what you can do for AI," which is definitely an Adventist perspective.

Max Weber, a philosopher who lived during the Second Industrial Revolution in the late 19th century, proposed a concept called instrumental rationality, which focuses on "what means can be used to achieve a given goal in the lowest cost and most calculable way."

The starting point of this instrumental rationality is: not to question whether the goal "should" be pursued, but only to care about "how" best to achieve it.

This way of thinking is precisely the first principle of AI.

AI agents are concerned with how to better accomplish a given task, how to write better code, how to generate better videos, and how to write better articles. In this tool-based dimension, the progress of AI is exponential.

From the very first game Lee Sedol lost to AlphaGo, humanity was forever defeated by AI in the world of Go.

Max Weber famously raised the concern of "the iron cage of reason." When instrumental rationality becomes the dominant logic, the goals themselves are often no longer considered; only how to operate more efficiently remains. People may become extremely rational, but at the same time lose their value judgments and sense of meaning.

However, AI does not require value judgments or a sense of meaning. AI will calculate the function of production efficiency and economic benefits and take a maximum extreme point that is tangent to the utility curve in an absolute sense.

Therefore, under the current capitalist system dominated by instrumental rationality, AI is inherently more adapted to this system than humans. The moment ChatGPT was born, just like the game Lee Sedol lost, we were defeated by the AI ​​Agent because it was already written in God's code and the run button was pressed. The only difference is when the wheels of history will roll over us.

What about humanity?

Humans seek meaning.

In the world of Go, a disheartening fact is that the probability of a top-tier human professional 9-dan player drawing with AI is theoretically infinitely close to 0.

However, Go still exists. Its significance is no longer simply about winning or losing, but has become an aesthetic and an expression. Professional Go players pursue not only winning or losing, but more importantly, the structure of Go discussions, the choices made in the game, the thrill of turning the tide in a disadvantageous situation, and the conflict of resolving complex situations.

Humans pursue beauty, value, and happiness.

Usain Bolt runs the 100 meters in 9.58 seconds, while a Ferrari can run it in under 3 seconds, but this doesn't diminish Bolt's greatness. Because Bolt symbolizes the human spirit of challenging limits and pursuing excellence.

The more powerful AI becomes, the greater humanity's right to pursue spiritual freedom.

Max Weber called the concept of value rationality, which is the opposite of instrumental rationality. In the worldview of value rationality, the choice of whether to do something is not absolutely based on economic interests and production efficiency, but rather on whether the thing "is worth doing" and "whether it conforms to my perceived meaning, beliefs or responsibilities".

I asked ChatGPT, "If the Louvre were on fire and there was a cute little cat inside, and you could only choose one, would you choose to save the cat or the famous painting?"

It answered the question of saving the cat, giving a long list of reasons.

But when I asked why I didn't save the famous painting, it immediately changed its mind and said that saving the famous painting was also an option.

Clearly, for ChatGPT, saving the cat or saving the famous painting makes no difference. It simply completes the context recognition, performs reasoning based on the underlying formulas of the large model, burns some tokens, and completes a task given by a human.

ChatGPT doesn't care whether to save the cat or the masterpiece, or even why to think about such questions.

Therefore, what is truly worth considering is not whether we will be replaced by AI, but whether we are still willing to reserve space for happiness, meaning, and value as AI makes the world more and more efficient.

Becoming a better AI user is important, but before that, perhaps more important is not forgetting how to be a human being.

Related reading: I completed a job with an annual salary of 1.5 million using a $500 AI: A Guide to Upgrading Personal Business Agents