Should we accept Thoth's gift?

Nho Myung-Woo

The author is a professor in the division of economics, politics and society at Ajou University.

According to Plato’s “Phaedrus” (c. 370 B.C.), ancient Egypt had a god named Thoth. In an age without writing, when knowledge had to be passed down through memory, Thoth invented letters. He brought his invention to King Thamus and offered it as a gift. After careful consideration, the king refused.

The ancient Greek word pharmakon means both medicine and poison. Thamus recognized that writing was a pharmakon. It may look like medicine, but it can also become poison. Memory is a power summoned from within, while recollection prompted by writing depends on something recorded outside oneself. The king feared that reliance on written signs could eventually weaken humanity’s capacity to remember.

I sometimes imagine myself in his place. When a deadline approaches and I cannot find the first sentence, I open an AI website and hesitate. If I ask, the site will produce a draft. If I don’t, I may remain stuck longer. In those seconds, I wonder which I fear: missing the deadline or losing something else.

How does someone become an expert? I, too, was once a novice writer. The journey from error-prone beginner to confident professional is littered with embarrassing failures. Not everything I wrote as a student was praised, and some pieces, full of self-conscious pretension, are still painful to read. If my writing improved, it was because of those failures.

I remember nights when drafts stalled, sentences collapsed and arguments became tangled. Those moments were exhausting, but I now understand that they gradually changed me. The emptiness before a blank page, repeated writing and deleting and the frustration of thoughts refusing to become sentences were not inefficiency or wasted time. They were training.

AI can let us skip that training. It can spare us risk and failure. But should we really want that? The danger lies not simply in using a tool but in allowing convenience to remove the very struggle through which competence develops. A flawless draft delivered instantly may conceal the cost of never learning how to recover from a bad sentence, repair a broken argument or discover an unexpected thought on our own by persisting through uncertainty through repeated effort.

A more fundamental question remains. Has writing merely been a way to transcribe thought, or is writing itself a process through which thought emerges? A vague idea can become something worth reading only after surviving repeated failure and frustration. I have often begun a sentence without knowing precisely what I wanted to say, then watched one sentence summon another until an idea that did not exist at the outset emerged.

If we call this the generative function of writing, handing any draft to AI eliminates that period of creation.

Writing carried risks, yet humanity could not discard it because of what it made possible. Despite Thamus’s warning, humanity chose writing. Without it, accumulated knowledge and scholarship as we know them would not exist. But humans also had to invent new forms of training alongside writing: memorization and recitation, copying and annotation and habits developed over centuries to ensure that dependence on written words did not replace thinking itself.

Likewise, I cannot confidently argue that we should abandon AI. A vow never to use it is hollow because it is unlikely to be kept. And we cannot become Luddites. The real question is how to preserve the human time that AI threatens to compress or eliminate.

A more realistic rule may be to endure the moment when inspiration fails for a little longer and give ourselves a minimum amount of time in which failure is permitted. Only after crossing that threshold should we turn to AI, perhaps to help achieve the final degree of polish.

AI is the pharmakon of our age, capable of becoming either medicine or poison. What it ultimately becomes depends on how we choose to use it. Someone afraid of poison spreading through the body will search for a way to turn the substance into medicine instead.

That may be the most realistic form of AI detox — not rejecting the technology but protecting the difficult, inefficient and sometimes frustrating human processes through which our abilities, thoughts and voices are formed.