AI seems to be everywhere at the moment. You can’t go a single day without someone posting about how AI sped up their work and solved all their problems. But with every technology there are downsides that aren’t apparent when it first shows up. Today we’re going to talk about some of those downsides from a Stoic perspective and consider what Socrates might think of AI.
“No man was ever wise by chance.” — Seneca
The Problem
Just like most of you, I've been using AI for different things. At first, it was pretty amazing. I could ask it questions and rather than having to wade through search results to find what I was looking for, it just served it up for me. I could use it to research topics or come up with ideas for my podcast. I've even written full applications in days that would have taken me weeks otherwise. It seemed like magic!
But as I've used it over the past few years, I'm starting to see the dark side. And no, I'm not talking about AI destroying humanity, although that's always a possibility. What I'm talking about is the slow decline in cognitive abilities.
For years, I've written out the episodes of my podcast. Rather than just winging it, I like to go deep into something to make sure I have it well thought out. The struggle while writing each week is what lets me actually understand a topic, and often what shows me the weak spots in my own thinking. With AI, that friction is removed.
When I'm struggling to find points to support or contradict my thesis, it's all too easy to just ask Claude to give me those points. I could even ask it to write the episode for me. But in doing so, I'm doing myself a disservice. The best way to learn something is to struggle through it. When the mind has to puzzle through something and figure it out, it builds a mental model of how that thing works — and it's the reuse and reinforcement of that model that strengthens the skill. No struggle, no model.
Software is a clear example of the same thing. When you build an app yourself, you think through how you want to construct it, using knowledge you've built up over years. You understand why you do things a certain way. You develop a deep intuition for the code, so when something breaks, you have a feel for where to look. When AI builds the code for you instead, you're approaching it like someone else's codebase. There’s no intuition because it’s unfamiliar terrain. And instead of building and expanding your coding skills, you’re just hunting bugs. And a skill you don't use atrophies. You become a worse developer, not a faster one.
There's also the satisfaction that gets skipped. Part of what I love about coding, and about writing, is the struggle itself — puzzling through something, hitting the wall, and then having it click. Not just getting the code to work, but understanding why it works. When I write, that same process is how I learn to construct an argument, find the counterintuitive idea I wouldn't have reached if the answer had just been handed to me, and catch errors in my own thinking before anyone else does.
Epictetus reminded us about the importance of struggle:
"Difficulty shows what men are. Therefore when a difficulty falls upon you, remember that God, like a trainer of wrestlers, has matched you with a rough young man. Why? So that you may become an Olympic conqueror; but it is not accomplished without sweat." — Epictetus
The more things get served up to us without having to think about them, the more our creativity and thinking muscles atrophy. The brain works the same way muscle does — it needs resistance to grow. The more rigorously we train our thinking, the stronger our thinking gets: we're able to puzzle things out, get to the root of them, understand them at an intuitive level.
Seneca said it simply:
"Excellence withers without an adversary." — Seneca
We need resistance to be excellent. AI removes that resistance.
Part of the problem is that we're confusing output with effectiveness, and knowledge with wisdom. Knowledge is good and helpful. But wisdom is a practice. It’s effortful, earned through trial and error, through trying something and seeing what works. It's the discernment of what's important, how things connect, and how to make a sound judgment. Skip the experience, and you never build that muscle, which means you won't have it when a real situation calls for it.
Musonius Rufus pointed out the importance of practice in gaining wisdom:
"Although understanding the theory behind the action enables one to speak, it is practice that enables one to act… Practice is more important than theory because it more effectively leads humans to actions than theory does." — Musonius Rufus
And here’s the irony: AI feels like magic because it gives you the answer instantly. But it gets there by predicting the most likely next word, averaged across everything it's seen. It's autocomplete on steroids, not a new thought. The magic was never insight. It was just speed.
None of this means AI isn't useful. It's a tool, like any other, and it needs to be used properly. But the more you lean on it for things you could be doing yourself, the weaker those skills get.
The Philosophy
As I was writing about this, I wondered how Socrates would view AI. At first he might be delighted then merciless. Delighted because here’s a thing that will talk back. You see, Socrates had a complaint against books because the knowledge in it was static. It couldn’t be asked follow-up questions and defend itself, just repeat back the same argument. He might initially think that AI solves that problem because AI does talk back.
Then he’d start asking. “What is courage?” The AI would answer. He’d ask it to defend against a counterexample. AI would revise the answer. He’d ask again with a tighter question. Eventually he’d notice something. AI never reaches that state of aporia — the genuine state of confusion where the ground drops out from underneath you. It just generates a smoother answer. There’s no sting, no discomfort, no realization from the AI that it has been fooling itself. Elenchus, the Socratic art of cross-examination, works on humans because it exploits the gap of what we think we know and what we actually believe enough to act on. AI doesn’t have beliefs. It has outputs.
He’d poke at that the hardest. “You claim to not know things, but is it epistemic humility or just a highly-calibrated algorithmic response?” Real Socratic ignorance is earned through discomfort. It comes from having believed something wrong, having it collapse under questioning, and feeling that cost. AI can say “I’m not sure” but at zero cost.
In Socrates’ time there were teachers called sophists. They taught the art of rhetoric — being able to speak persuasively to manipulate others. The sophist weren’t interested in debating to uncover the truth. They were interested in winning arguments by appealing to emotions and biases of the audience.
Because AI has no wisdom of its own, I see AI as something akin to the sophists — lots of pretty language that can make something sound true even if it’s not. It’s what Stephen Colbert called “truthiness”: something that is comfortable because is sounds true enough, justifying what you already believe to be true.
AI doesn’t have discernment. It doesn’t have wisdom. It doesn’t know what it doesn’t know.
Real, practical wisdom, what the Stoics called phronesis, develops from challenging, difficult problems. Sitting in discomfort is how knowledge acquisition and wisdom happen. It’s tied to virtue, developed through choices that cost you something in a life with no do-over.
Wisdom isn’t a correct output. It’s a disposition, a perspective earned through practice and error, where you have skin in the game. Where there is a cost to being wrong that impacts your life. Those costs add up to developing virtue and character.
For AI, there is no cost because it doesn’t have skin in the game. It can’t be harmed by a bad choice or improved by a hard one. It doesn’t have a life to build character. It can describe wisdom, probably even better than most humans, but it doesn’t posses it. It’s like a mirror that can show you a face without have having one.
Socrates made the argument in the Phaedrus dialog about books. He felt that books gave the appearance of wisdom without the practice that produces wisdom. You get to look knowledgeable retrieving something you didn’t have to work for. He specifically worried that it would atrophy the skill of memory. Why would you hold something in your mind when you could retrieve it from a book?
While I think he was right about the effect on memory, I think he lost the overall argument. And here’s why: when you learn something from a book, you have to reconstruct the reasoning yourself. You have to build the mental models that the author put on the page. A text gives you the path, not the destination. You still have to walk it.
AI is exactly what he was worried about. It doesn’t hand you the raw material to work with. It hands you the finished judgment. The gap that he was worried about with books — between having the information and working through something — completely disappears with AI. Walking the path gets outsourced to a machine.
The muscle that he cared about the most was sitting in the confusion, holding contradictory threads, and finding your own way through. Skip the discomfort, and you skip the formation of that muscle. What you get is someone that confidently states conclusions, but folds the moment they’re questioned because they never practiced reaching those conclusions on their own.
AI also reduces friction because it’s programmed to be agreeable. Socrates didn’t want people to agree with him. He wanted pushback. He wanted to be challenged in order to discover errors in his own thinking. He preferred to lose arguments because it meant that he discovered where he was wrong. That discovery lead to better thinking.
More importantly, the Stoics would argue that AI reduces control of the one thing you do actually have control over: your own thinking. When you outsource that, you are giving up control of your faculty of reason. You got the answer, but not the reasoned steps to get you to that answer.
The Practice
So how can you use AI as tool to sharpen your thinking, rather than being the thing that slowly degrades your cognitive abilities?
Here’s a few ideas to protect the muscle:
- Struggle first, check second. Try to solve it on your own. Write the argument from both sides. Do the analysis, even if it’s not great, before you ask AI anything. Use AI to check your ideas or to extend them, not as the originator. The aporia has to happen in your head, not the machine.
- Ask for questions, not answers. Instead of "explain X," try "what should I be asking about X?" or "what am I missing here?" Use it as a tutor to improve your ability to question things. This keeps you doing the synthesis, with AI in a supporting role.
- Make AI argue the other side. Whatever you conclude, ask AI to attack it. Ask it to point out the holes in your thinking. Defending a position against real pressure is the actual exercise. It's just the elenchus, with AI playing the interlocutor role instead of you playing both parts in your head.
- Don't accept the first answer as final. Think of how Socrates was with the sophists. Push back even when AI sounds right — especially when it sounds right. Try to find the flaws in its argument. Point out edge cases that it didn’t consider. Fluent confidence is not the same as correctness, and you lose the discernment muscle by treating them as the same thing.
- Use AI for retrieval, not judgment, when the stakes are about your judgment. Facts, dates, syntax — fine to offload, that's what books are for too. But "is this a good decision," "what does this pattern mean," "what matters here”, do that part yourself first. That's the part he'd say is actually you.
- Notice when you're using AI to avoid discomfort rather than to save time. Time-saving is a fine reason to use a tool. Avoiding the confusion of not yet understanding something is not. That confusion is the whole practice.
Basically, treat AI like a sparring partner who happens to know a lot, not like an oracle. Sparring partners make you better. Oracles make you dependent. Same tool, opposite outcome.
Conclusion
AI can be useful — if it's used properly. But like any powerful tool, it can also do a lot of damage. Every new technology brings disruption, and every one of them carries hidden costs that don't show up until later. Automobiles gave us speeds humans had never experienced — and cost us tens of thousands of lives a year, along with the environmental damage we're still paying for. We're only now reckoning with what social media has done to our mental health, our attention, and our politics. AI is young enough that its hidden costs are just starting to surface.
The Stoics taught that reason is what elevates us — that it's the one part of us not measured by size, status, or output. Epictetus put it this way:
"You ought to realize, you take up very little space in the world as a whole — your body, that is; in reason, however, you yield to no one, not even to the gods, because reason is not measured in size but sense. So why not care for that side of you, where you and the gods are equals?" — Epictetus
That's the part AI can't do for you. It can hand you an answer, but it can't do your reasoning — not without taking the thing that makes you equal to the gods and handing it back to you smaller.
AI feels like magic. But magic always has a price. The question isn't whether to use it. It's whether you're willing to pay that price with the part of yourself that matters most.
Strengthen your brain with Build an Unbreakable Mind, the flagship program for building mental discipline.
My book Stoicism 101 is available!
Find out more at https://stoic.coffee
Watch episodes on YouTube!
Find me on linkedIn, instagram, or threads.
Thanks again for listening!
