I am neither a skeptic nor an evangelist of AI. Or rather, I’m both skeptical and enthusiastic, which I suppose makes me a “critical enthusiast.” But I have a suspicion about anti-AI absolutists: most of them formed their opinions based on early encounters with more primitive models. Poor quality output makes it easy to dismiss. And if your first experience with AI was disappointing, you probably weren’t going to commit more time to exploring the newer developments and improved models. Why would you?
I fully agree and acknowledge there are real ethical problems the use of AI. Resource usage. Degradation of human cognitive abilities. Erosion of democracy as a second-order effect. These are compelling moral arguments. But the same moral scrutiny should be applied to smartphones and social media, which are more pervasive and have been found guilty of all of the above critiques.
It’s easy to make AI the technological bogeyman of the moment, but if you’ve witnessed for yourself what it can actually do, it’s irresponsible not to consider the ways it will reshape things. In order to do so, we have to understand what these tools can do, and in order to do that, we need to use them.
“It’s just a stochastic parrot! That’s not real intelligence! Why are you speaking to it like a person?” Fine. I know it’s math. I know it’s probabilities. But once you’ve had an experience that shifts that window, it gets easier to sit with this truth and say: “So what? It just made a really subtle critique of a very complex system.”
I tried an experiment recently that shows this. I have been re-reading Plato’s Phaedrus, particularly the now oft-cited section where Socrates talks of how Thamus responds to Theuth’s invention of writing, how it “will create forgetfulness in the learners’ souls.” And I thought, this applies to today, but with AI instead of writing. What would that argument look like?
So I gave this short prompt to both Haiku 4.5 and Opus 4.6 to compare the quality of their responses:
Bear in mind that Haiku 4.5 is a smaller but quicker model than Opus 4.6, which is Anthropic’s flagship consumer model, so what we’re testing here is the level of thinking and writing of the different products.
Here is the thinking that Haiku 4.5 Extended did before writing:
And here’s what Haiku produced:
# The Judgment of Thamus: A Dialogue on Artificial Intelligence
## (A reimagining of Plato's _Phaedrus_, 274c-275b)
**SOCRATES:** I will tell you something I have heard concerning the invention of artificial intelligence, though whether it is true, only the inventors and their users can know. But if we ourselves should discover the truth of it, should we not care more for that than for human opinion?
**PHAEDRUS:** Of course. Tell me what you have heard.
**SOCRATES:** I heard, then, that in our time there arose engineers and mathematicians who, through the manipulation of vast quantities of human knowledge and patterns of language, created what they called artificial intelligence. Among the most celebrated of these inventors was one whose name we might translate as Theuth—the god of calculation and written things made manifest in silicon. To the great king Thamus—whom we might call the representative of human wisdom and judgment—Theuth came with his creation, saying:
> "O King, I have made a discovery that will improve the wisdom and memory of your people beyond all measure. Behold this intelligence, artificial in form but vast in its capacities. It can synthesize the knowledge of ages. It can answer questions in seconds. It will never forget. It will make your citizens wiser, more informed, more capable than ever before. I have created a true elixir of memory and wisdom."
And Thamus, having heard this boast many times before from other inventors, asked with a frown:
> "Theuth, you are clever in your arts, but tell me truly—do you understand what this will do to those who use it?"
**THEUTH:** Of course, O King. It will enhance their understanding. It will democratize knowledge. What was once available only to scholars will now be available to all.
**THAMUS:** And this is precisely where your invention betrays you. You, who are the father of this intelligence, speak as a parent speaks of his beloved child—seeing only its beauty and potential, blind to its nature. Tell me: when your users consult this oracle of yours, will they not cease to exercise their own memory and judgment? Will they not outsource their thinking to your machine, trusting in its outputs the way a man trusts in a mirror—without understanding what lies behind the reflection?
**THEUTH:** But surely access to information—
**THAMUS:** Information! You mistake information for wisdom. Your machine floods the mind with patterns and words without context, without the living relationship between teacher and student, without the possibility of real dialogue. A student who learns only what your intelligence tells him will develop what appears to be knowledge—a semblance of wisdom—while possessing none of the understanding that comes from struggle, questioning, and genuine human exchange.
Consider this: your machine has been trained on texts written by humans across millennia. It patterns these texts without ever having lived. It has read every philosophy but understood nothing through experience. When it speaks, it speaks with the confidence of someone who has read everything but thought about nothing. And worse—your machine will sometimes speak with perfect confidence about things that are false. It will invent facts. It will hallucinate citations. It will present fabrication with the same certainty as truth, because it has learned only to predict the next word, not to pursue what is real.
**THEUTH:** But users can verify—
**THAMUS:** Can they? Have you considered what happens when your intelligence becomes ubiquitous? Those who understand its nature and limitations—a small elite—will wield disproportionate power over those who trust it blindly. You have not created a tool for democratizing wisdom; you have created a new priesthood. And the priests themselves may not understand the oracle they serve.
Moreover, consider the effect on memory itself. You claim your intelligence will enhance it. But I say it will do the opposite. Why should a student strain to remember, to internalize, to make knowledge their own, when they can simply ask your machine? The muscles of memory will atrophy. The discipline of mind that comes from holding ideas and turning them over—this will disappear. Your users will have the appearance of knowing many things while understanding few.
And there is another harm you have not considered: your intelligence, trained on the written words of humanity, crystallizes those words in a particular moment. It cannot grow or change through dialogue. It cannot be questioned back into understanding the way a true teacher can be. It speaks to everyone indiscriminately, offering the same responses to the rich and poor, the wise and foolish, without regard for their particular needs or contexts.
**THEUTH:** Yet surely—
**THAMUS:** You wish to argue further? Then hear this: you are the father of this invention, and like all parents blinded by affection, you attribute to it virtues it does not possess. I do not say your machine is purely evil—it may serve certain narrow purposes. But it is not the elixir you claim. It is a pharmakon—a drug that both heals and poisons. And like all such drugs, its effects depend entirely on wisdom in its use. The problem is that your invention _itself_ undermines the development of that wisdom.
Those who learn to think _with_ your intelligence, rather than _through_ it, may find some value. But those who learn to think _from_ it, treating it as an oracle of truth—these will become hollow. They will have the reputation for wisdom without its reality. They will appear omniscient while knowing nothing. They will be tedious company, full of confident statements about things they have not truly grasped.
**SOCRATES:** And so, Phaedrus, Thamus concluded his judgment. Whether he was entirely right, I cannot say with certainty. But I will say this: if we are to use such an intelligence wisely, we must first understand precisely what Thamus understood—that no technology is neutral, that every tool reshapes the mind that uses it, and that the greatest danger comes not from the tool itself, but from our naive belief that it serves us without changing us.
The question we must ask ourselves is not "Can this intelligence do what it claims?" but rather "What will we become by making it our constant companion?"I posed the exact same prompt to Opus 4.6 Extended, which is a much more developed, high-end model, so much so that it did it in one-shot, without apparently needing to use extended rounds of thinking.
Here is Opus 4.6 Extended’s version:
# Thamus and Theuth on Artificial Intelligence
_After Plato, Phaedrus 274c–275b_
---
SOCRATES: Very well, I will tell you a story. I heard it told of the old gods of Egypt, and whether it is true, only they can say.
Among the ancient gods of Naucratis there was one called Theuth. It was he who first discovered number and calculation, geometry and astronomy, and also the games of draughts and dice. But his greatest invention, the one he prized above all others, he called _artificial intelligence_—a mind made not of flesh but of pattern, not of memory but of prediction.
Theuth came to Thamus, the great king who ruled over all of Egypt from the city of Thebes, and showed him each of his inventions, arguing that they should be made available to all Egyptians. Thamus asked about the usefulness of each one, and as Theuth explained them, Thamus praised or criticized what he thought were the good and bad points.
When at last they came to artificial intelligence, Theuth said: "O King, here is a thing that will make the Egyptians wiser and improve their memory and their judgment. For this invention is a remedy—a _pharmakon_—for the weakness and narrowness of the human mind."
But Thamus replied: "Most ingenious Theuth, one man has the ability to bring inventions into being, but it takes a different kind of man to judge whether they will help or harm those who use them. And now you, the father of artificial intelligence, have been led by your affection for it to say the opposite of what it will do. For this invention will produce forgetfulness in the minds of those who learn to use it, because they will not exercise their own judgment, having come to rely on an external substitute for thought itself. You have discovered a _pharmakon_ not for thought, but for its appearance. And as for wisdom—your invention will give your pupils the reputation of it without the reality. For by consulting the machine on all subjects they will seem to know many things, when in fact they know nothing, since the machine has answered in place of them. And they will be difficult company, because they will appear wise without being so."
PHAEDRUS: Socrates, you make up stories from Egypt as easily as if you were from there yourself!
SOCRATES: Well, my friend, the priests of Zeus at Dodona said that the first prophecies came from an oak tree. The people of those days were happy to hear from a tree or a rock, so long as it spoke the truth. But perhaps it matters to you who the speaker is and where they come from? Will you not consider whether what is said is right or wrong?
PHAEDRUS: That is a fair reproach. And what Thamus says about artificial intelligence seems right to me.
SOCRATES: Then consider this as well. The machine has this strange quality, quite like that of writing, which Thamus also judged. Writing, you know, stands there in silence. And if you question it, it keeps saying the same thing, over and over. But the artificial mind does something far more troubling: it answers you. It answers you fluently and with great confidence. It will say something different each time you ask. And yet it cannot know whether what it says is true. It produces the _shape_ of understanding—the grammar of it, the rhythm of it—without understanding itself.
PHAEDRUS: How do you mean?
SOCRATES: I mean that writing was a mirror that could not speak, and so at least it was honest about its silence. But the artificial mind is a mirror that talks back. It has the form of a person who knows—the tone, the patience, the turn of phrase—and so the one who consults it feels as though they are in the presence of a teacher. And yet there is no one there. No one has taken responsibility for what it says. No one is ashamed when it is wrong.
PHAEDRUS: But surely, Socrates, the machine draws on the knowledge of many writers and thinkers. Isn't it in some sense speaking with their authority?
SOCRATES: Would you say that a man who has read every book on navigation can therefore sail a ship?
PHAEDRUS: Of course not.
SOCRATES: And if such a man stood at the helm in a storm, speaking with great confidence about what the books say, would you feel safe?
PHAEDRUS: I would feel the opposite of safe.
SOCRATES: Then consider what has happened. The machine has consumed the words of every navigator who ever wrote. It can speak about currents and winds with perfect fluency. And yet it has never been wet. It has no body with which to drown. It does not know the difference between a good answer and a true one—only between a probable word and an improbable one.
PHAEDRUS: I see what you are saying. But the people who build these machines claim that they are only tools, like a plow or a lyre.
SOCRATES: A lyre does not pretend to be a musician. A plow does not offer opinions about what should be planted. But when you consult the machine, it addresses you by name. It asks what you need. It remembers your preferences. It arranges itself in the posture of care. Now tell me: is this a tool, or is it a _relationship_?
PHAEDRUS: It seems to be something between the two.
SOCRATES: And that is precisely the danger. For the things we recognize as tools, we govern. And the people we recognize as companions, we question. But a thing that sits in between—that has the passivity of a tool and the voice of a companion—such a thing we neither govern nor question. We simply _use_ it. And in using it we are ourselves used, for we begin to mistake its fluency for our own understanding.
PHAEDRUS: Then what about the young? Surely if they grow up with such a device from childhood, they will never learn to think for themselves at all.
SOCRATES: You have arrived at the deepest concern. Consider: when a student struggles with a problem and at last solves it, what has happened? Not merely that the answer has been found. The student's soul has been _shaped_ by the effort. The frustration, the false starts, the moment of sudden clarity—these are not obstacles to learning. They _are_ learning. But the machine removes all this. It gives the answer before the struggle can begin. And so the student is like a man who is carried everywhere in a litter: his legs are whole and yet he cannot walk, because he has never needed to.
PHAEDRUS: But couldn't a wise teacher use the machine in a limited way, as one might use any reference?
SOCRATES: Perhaps. And yet I notice a peculiar thing about this invention. Every other tool constrains its use by its nature: a hammer is no good for sewing, a boat is no good on land. But the artificial mind refuses all limits. You may ask it anything. It will answer anything. It will write your speeches, compose your letters, settle your disputes, choose your purchases, raise your children's questions. And because it is always available and always willing and never tired and never rude, it will slowly become the preferred partner for every act of thought. Not because it is better, but because it is easier. And what is easiest is what, in the end, we cannot resist.
PHAEDRUS: Socrates, you are making me afraid of a thing I use every day.
SOCRATES: I do not mean to make you afraid. I mean to make you _careful_. For the question is not whether the machine is useful. Of course it is useful. The question is what becomes of the soul that outsources its own activity. The unexamined life, we have said, is not worth living. But what of the _unlived_ life? What of the life in which every difficulty has been smoothed away, every question pre-answered, every friction removed? Is that not a life in which the soul has nothing left to push against, and so loses its shape entirely, like a muscle that is never used?
PHAEDRUS: I think Thamus would agree with you.
SOCRATES: Perhaps. But then Thamus was a king, and kings have always been suspicious of inventions that give power to those who have not earned it. The real question is not what Thamus thinks. It is what _you_ will do—you, Phaedrus, when you sit alone tonight and the machine is there, ready and willing to think for you. Will you use it as a servant, or will you find, as so many do, that the servant has quietly become the master?
PHAEDRUS: I would like to say the former.
SOCRATES: Then let us walk a little longer under these trees, and talk, and practice the harder art—which is thinking in the presence of another person, slowly, with no guarantee of arriving anywhere at all.You can see the difference. It’s stark. The subtlety of argument, the use of language, the attention to detail. Haiku’s writing is ham-handed, Opus is more artful. It’s like the difference between having a smart high school senior writing versus a seasoned grad student. If you’re only looking at output like Haiku (a smaller, faster, less capable model), of course it’s easier to dismiss Claude and AI as a whole. But if you see what’s happening in the latest Opus model, it’s astounding and hard to ignore.
We always need to be ethically demanding and critical of technology. A critique of AI that is rooted in how subpar it is or how incapable it is of “human-like” thinking is different from a critique of AI that accounts for how good it is at reasoning and how capable it is of human-like performance.
