David Egan Philosophy

You’ve probably heard of this new technology that’s been transforming the way people think, communicate, and learn. It threatens to disrupt our lives at the most fundamental level. It presents a pastiche of human thought so plausible that many mistake it for thinking—indeed, are in danger of losing the ability to think without it. The risk that it will sow confusion and undermine the social order is very real. Resisting its influence will require constant vigilance.

The technology I’m talking about is writing, and the voice is Plato’s. The Greek alphabet was several centuries old when Plato set stylus to wax but writing as a tool for public broadcasting was very new. Plato’s dialogues attest to this change: they record, in writing, spoken exchanges between Plato’s teacher Socrates and his interlocutors.

A lot of what Plato says about writing has echoes in contemporary commentary on large language models (LLMs) and other forms of machine learning. In hindsight, Plato’s worries can seem overblown—literacy is now widespread and the sky remains unfallen. Does that mean that we shouldn’t worry so much about how LLMs will transform society either?

Not necessarily, and not just because history never exactly repeats itself even when it rhymes. It’s also that Plato wasn’t entirely wrong. Literacy has radically altered every society it’s touched. From the far side of that transformation, it’s hard to see it as a catastrophe. But the way that people live, think, and conceive of themselves has changed, often in ways we scarcely register. Plato’s warnings are still relevant, and their distance from the present can give us some valuable perspective.

Consider what Plato says about writing toward the end of his dialogue, Phaedrus. Writing will make people lazy and forgetful, he says. People will lose the habit of internalizing what they learn and just trust that, if it’s written down, they can call it up when needed. People who rely on writing will believe they understand much more than they actually do.

What’s more, writing risks supplanting thinking with its false semblance. Written words look like they express thoughts but they’re inert, Plato says, unable to think for themselves or answer questions one might want to put to them. You might feel like the author is speaking to you when you read a text but no true dialogue with the author is possible.

And writing is careless with the truth, he says. A written text “roams about everywhere, reaching indiscriminately those with understanding no less than those who have no business with it, and it doesn’t know to whom it should speak and to whom it should not.” A piece of writing gives the appearance that you’re having a conversation with someone but the appearance is deceiving. The text has no mind of its own. It doesn’t even know you exist.

In effect, LLMs are Plato’s worst nightmare. All the dangers he perceives in the written word take on monstrous new form in algorithms that churn out oceans of text, arranging fine-seeming phrases that not only lack any genuine commitment to the truth but that don’t even have a conception of reality to which they might try to be faithful. With writing, Plato says that the words come untethered from the mind of their author. With LLMs, there was no mind to begin with.

With two and a half thousand years of hindsight, we can see that writing brings certain advantages. Writing—and subsequent technologies like printing and digital technology—vastly expands the range of people we can learn from and exchange ideas with. Many of you who are reading this have never spoken with me in person (although some of you have—hi mom!) and we all come several thousand years too late to talk to Plato. Yet here I am telling you about this beguiling genius from ancient Athens.

Writing also facilitates sustained reflection. Take philosophy, for instance. Written philosophy can develop arguments and build systems of thought that are far too complex to hold in one’s mind all at once. Writing doesn’t just preserve our thoughts. It also transforms what we’re capable of thinking.

Unlike his teacher Socrates, Plato didn’t shun the written word—his complete works run nearly two thousand pages. He clearly saw that writing has its uses, but he also saw dangers. Foremost among them is that writing risks giving us a mistaken idea of what we do and don’t understand. “I read it somewhere” is a common shorthand for something you claim to know but can’t really explain.

In effect, Plato noticed that writing was the first technology to significantly externalize our knowledge. He remarks that writing “is external and depends on signs that belong to others,” in contrast to what people know “from the inside, completely on their own.”

Subsequent developments in information technology have furthered this externalization. An unprecedented amount of knowledge is just a Google or Wikipedia search away. Machine learning algorithms raise curious conundrums for what we can even claim to know. If a computer crunches a proof of a mathematical theorem of such complexity that no human mathematician can ever survey it, has the theorem been proved?

Plato’s response to the externalization of thought is pertinent today. He says that the best way to secure your understanding is through dialogue. If I can justify what I say, in conversation and in my own words, to a responsive listener, then I’ll find out whether I do indeed know what I’m talking about and not just mouthing things I read somewhere. And if I can’t explain myself clearly, I’ve learned something about the extent of my ignorance.

Plato’s writings model this dialogue form. The method of dialectic has had immense influence on how the humanities have been taught ever since. A seminar room is essentially a forum for conversation.

I’ve been thinking about Plato and LLMs as I prepare to return to the classroom this autumn. University instructors, especially in the humanities, have been scrambling to respond to the existential threat that LLMs pose to the college essay. I don’t have any great answers myself but it does give me renewed appreciation for how much valuable thinking happens through conversation.

And I think—I hope—the coming years will only enhance the value we place on the art of conversation. For as long as written words have required human writers, the cost of producing texts—in terms of time, effort, and expertise—has exceeded the cost of reading them. That’s starting to change. LLMs can churn out plausible copy faster than people can read it. As supply exceeds demand, the prestige of the written word will wither.

Plato would welcome this shift. Writing has its uses—and those uses will only expand with the spread of LLMs. But the injunction to “think for yourself” will increasingly send us back to thought’s original home in words spoken and heard. 

1. According to a recent estimate, wild terrestrial and freshwater animals expend 76,000 gigajoules a year in altering the land around them. That’s the equivalent landscape-altering force of several thousand extreme floods a year. (source)
2. In the mid-eighteenth century, 1% of London’s population died from tuberculosis every year. That’s nearly twice as many deaths, proportional to population, as occurs in present-day London from all causes put together. (source)
3. Approximately 10% of the Chinese infants given up for international adoption between 1991 and 2024 were forcibly taken from their birth parents. (source)
4. It’s possible to reconstruct what the Earth was like—and what life forms lived on it—hundreds of millions of years into the past due to its stratified geology. None of this would be possible on the moon, Mars, Mercury, or Venus. We can thank a combination of volcanos, plate tectonics, sedimentation, and erosion that happens at just the right pace. Mercury and the moon have become volcanically inactive, Mars no longer has the water or atmosphere to lay down sediment, and Venus has been so volcanically active as to erase its own geological past. (source)
5. Milwaukee is built on an ancient fossilized coral reef. (source)
6. Sweden nearly became the fifth country (after the USA, USSR, UK, and France) to have nuclear weapons before abandoning their nuclear weapons program in the mid-1960s. (source)
7. The word “symbol” derives from an Ancient Greek practice of breaking a piece of ceramic in two and later putting it back together again. Two people who would be at some distance from one another would break the “symbolon” in two and divide the two pieces between them. When one wanted to send a message to the other, he would send his half of the symbolon (the word originally means “to put together” or “combine”) with the messenger so that, in fitting the two halves together, the recipient of the message could be confident that the message had originated from the sender. (source)