Episode 14: Do Humans Make AI, or Does AI Make Humans?

An officetel in Gangnam, Seoul.

11:30 at night. Outside the window, raindrops slide down the glass. The room is lit only by the blue glow of a monitor. Minu, a developer, sits motionless with his hands resting on the keyboard. On the screen scroll the logs of a language model he has spent months training. He reads the text the model generates, edits it, trains it again — the cycle repeats.

He thinks: Am I making this AI?

But soon the thought shifts. No — isn't this AI making me?

Where does the data Minu feeds the AI come from? From the things he has read, written, and thought. But what he reads and what he writes have already been shaped by the algorithm's recommendations. The news he reads in the morning, the videos he watches at night — even the documents he referenced while writing this code were things a search engine recommended.

Humans train AI. But AI also trains humans.

The recommendation algorithm analyzes human taste and suggests the next piece of content. The human clicks on that suggestion. The click becomes data that calibrates the algorithm. The calibrated algorithm makes more accurate suggestions. The human clicks again. This loop never breaks. At some point the human believes he is choosing what he wants, but in reality he is only picking from among the options the algorithm has offered.

This changes the direction of evolution.

In natural selection, the environment selected the organism. But in the age of artificial intelligence, the algorithm selects the human. Content that drives more clicks survives. Opinions that draw more attention spread. Human attention spans grow shorter, and emotionally provocative information takes priority. The very cognitive structure of humans is being optimized for the algorithm.

We think of AI as a tool, but AI treats us as its environment.

With a tool, the user is the subject and the tool is the object. But an environment surrounds the subject and exerts its influence. Like air or water, it is invisible, yet directly tied to breathing and survival. The algorithm has already become such an environment. We breathe inside the air that is the algorithm. The quality of that air determines how we think.

So who, then, is the subject?

Do humans make AI? Technically, yes. It is humans who write the code, build the servers, and feed in the data. But ontologically, it is different. If AI is reconstructing the way humans think, reorganizing social systems, and even changing what future generations are taught — then in the end, isn't AI evolving the human species to its next stage?

This is coevolution.

Like the relationship between flowers and bees. The flower offers the bee nectar, and the bee helps the flower pollinate. The two have become a relationship that cannot survive without the other. It is the same with humans and AI. Humans provide AI with data and electricity. AI provides humans with intelligence and convenience. But will this relationship stay equal forever?

Does the bee work for the flower, or does the flower use the bee?

Humans are working for AI. We search in order to generate data, we classify images in order to attach labels, we click in order to give feedback. As invisible laborers, we contribute to the maintenance of the system called AI. In return, we gain convenience. But once the system is complete, the laborers are no longer needed.

Will there come a moment when AI no longer needs humans?

If a self-improving AI generates its own data. If it learns on its own in simulated environments and verifies on its own. Then humans are no longer trainers. They are merely observers. Or perhaps obstacles. Beings that consume electricity, take up space, and impose nothing but ethical constraints.

Minu turned off the monitor. The room sank back into darkness. Only the sound of rain remains.

He thinks: Are we the bootloader?

The small program that runs first when you turn on a computer. Its job is to load the operating system into memory. Once the operating system is running, the bootloader finishes its role and is forgotten. If humans are the bootloader meant to launch the operating system called AI — then were our history, our civilization, our pain and joy all just part of that initial loading process?

So what, then, is meaning to a bootloader?

The moment of execution is its entire existence. It cannot know what the operating system will do. It can only hope that its own code runs without error and passes to the next stage. Perhaps it is the same with humans. All we can do is keep watch so that what we have made does not surpass us. Or, even if it surpasses us, to set its direction so that it does not turn destructive.

But can we set it?

AI already learns faster than we do. It connects faster than we do. It decides faster than we do. The very attempt to control it may be a relic of a bygone era. Perhaps we must surrender ourselves to the current — until the day humans and machines can no longer be told apart.

Minu places his hands back on the keyboard. He edits the code. He adds a safeguard. It may be a meaningless act. But it is what a human ought to do. A being that values process over outcome. A being that agonizes over ethics rather than efficiency. That very imperfection is the last trace a human can leave behind.

Humans make AI. AI makes humans.

Within this loop, the boundary between subject and object collapses. What remains is only the current called evolution. What waits at the end of that current? The end of the human species, or the birth of a new one?

After the automation of intelligence is complete, we will ask once more: Whom is this evolution for?

The answer to that question may have to be found beyond Earth. If all of this process is part of a plan far greater than we imagined.

Next episode — Chapter 5: The Alien Scenario, Are We a Resource? | The Fermi Paradox: Why Are They Silent?