In high school, I took Latin. I actually did somewhat well at it, getting a Manga Cum Laude on the National Latin Exam. I was far from the best in my class, however. I owe this to the fact that I was taught Latin by one of the best teachers I've ever had: Jeff Souther. He is an old man with a big gray beard, who had a story about the time (as students embellished it) he beat Chuck Norris in a fight at a martial arts tournament. It is undoubtable to anyone who knew him that he smoked weed on occasion. He taught some classes at the high school and some classes at the local college, although I took the last class he taught at the high school. He is still alive and well in 2026, as far as I can tell.
I have many fond memories of Mr. Souther's class. It was full of the biggest nerds in the school, and especially the older students had a lot of camaraderie. One of the students translated a Roman recipe for a sort of plum cake and cooked it for us on Saturnalia. For a few months, we started every day with about 10-20 minutes of debate about whether 9/11 was an inside job. I could never tell how sincerely he held his belief on that, or whether he was simply goading us into employing rhetoric against his position. I learned more about making a convincing argument in that class than I probably did in four years of other high school classes combined.
When you study something like Latin, or indeed math, you constantly get asked questions like "what are you going to use it for?" from strangers, relatives, even my friends. It's a question I have had to design canned responses to over the years. For Latin my response was something like studying Latin makes it easier to pick up on the structure of romance languages, or that it helps sharpen your knowledge of English. Both of which are true. Now that I am so heavily invested in mathematics, I often have to point out to people that mathematicians can make a lot of money in the stock market, or rattle off some engineering applications of the specific types of mathematics I study. This is usually enough to satisfy people.
Once, a student asked Mr. Souther a variation of that question. The wording was something like "what's the point of learning this?". Mr. Souther, exasperated, replied "The point is that you will become educated!"
Again I have to stress how different this viewpoint was from what I saw in my other classes, or even what I was told from my parents. In every other class, the point was to display that you are capable of studying so you can get in to a good school, with the eventual goal of getting trained to do some high-paying job. A secondary goal is to show prestige and lord your knowledge over your peers, to be known as one of the smart kids. Mr. Souther's tautology said something different: education is its own reward. Having the knowledge and discipline necessary to learn things for yourself, and being able to communicate your understanding effectively to other people, are valuable in themselves; the fact that they are also useful for gaining resources or social standing is just icing on the cake.
Mr. Souther, of course, had similar canned responses to mine when he needed to justify his life's work to some school higher-up or a concerned parent. I believe that, like me, his responses didn't involve the real reasons he thought that what he did was valuable. Like a space engineer listing tech spin-offs of the Apollo program, he was forced to show off the by-products of his work and not the purpose of it.
It's twelve or so years later now, and in that time chatbots have gone from a funny novelty to a powerful technology that is shaking the foundations of our society. Some ridiculous percentage of the GDP of the United States is caught up in the "AI" market, people make comparisons to the Manhattan Project (in that allegory, where is Hiroshima?), and it is disturbingly not uncommon for people to have "AI girlfriends" and "AI boyfriends". Discussions about "AI" are inescapable. Pictures generated with stable diffusion are on billboards, and someday soon they'll be on TV too. We are constantly reminded that large language models can apparently pass difficult exams with ease. Many people think that anyone not currently using a chatbot in their work is falling behind, and that someday soon chatbots will automate all white-collar work.
Underpinning our collective fascination with "AI" lies a fundamental question: if you can prompt a chatbot to analyze any data, produce any program, write any text, or draw any image, why should we bother knowing how to do these things? What is the point of knowledge when a machine can do it better?
Before I get in to this, I want to disabuse you of one notion right now. These applications of machine learning are not "just tools like any other tool". The way they are marketed, and often the way they are used, is fundamentally different from what we usually think of as a tool. For example, consider a self-driving car. Many people like to rephrase the famous Trolley Problem in terms of self-driving cars: is it better for the car to do nothing and let five people die, or turn and kill one person? However, this fundamentally changes the nature of the problem: no longer do we think of what decision a person with ethical standing should make, but instead what outcome is preferable for a machine with no ethical standing to do. This is a qualitative, fundamental shift from a car that, perhaps, has some kind of fly-by-wire system, or a driver assist functionality. The self-driving car is a machine that is designed by a human, and (perhaps) guided by a human, but not controlled by a human. This is why, for instance, Tesla is trying so hard to market self-driving car systems while also claiming that legally it is not really fully self-driving, so that they can continue to hold the human accountable for any issues and avoid any accountability for themselves.
This reasoning extends to all use of chatbots and other forms of generative "AI". In all cases, the human becomes merely a guide, a prompter, taking the role of a superior commanding their subordinate rather than a person using a tool. However, their subordinate now has no ethical standing; there is no person who can be held accountable for either making a mistake or carrying out illegal orders.
So, let's talk about why knowing things is good. Reason number one: you can say things you know, and you can be held accountable for them. This might sound like a bad thing. After all, who wants more responsibility? But the fact that people stake their name on the things they say and do is the bedrock that society is built on. We would not have such good bridges if we didn't require civil engineers to personally put their reputation and indeed their ability to make an income on the line by signing off on bridge designs. This feature of humanity cannot be replicated by machines, at least not until they become people with selves. No matter how accurate your model is in assessing the quality of a bridge, someone needs to be responsible for it, and only a moron or someone who fears no consequences would sign off on something their chatbot generated without understanding it themselves, at the same level as if they had done it themselves.
Bridges are important, but I think this principle applies to much less important things as well. It is critically necessary to hold people accountable for what they say and do, and if you hold people accountable for what they say and do, then only a moron would sign off on code they don't understand, or send out a document they haven't thoroughly read and checked for errors. This is why if you want real productivity gain from "AI", you need to know exactly as much as if you were doing everything yourself. Being a good prompter does not cut it, and never will. If you are seeing productivity gains despite knowing nothing, I think it is probable that you are simply not being held accountable. Sometimes, you shouldn't. Nobody should be judged for vibe coding a side project or doing casual research. I will leave it up to you to decide whether there is enough accountability in other spheres of life, though.
So what if everyone's a moron that blindly trusts their chatbot? If the chatbot is good enough, the results will still be better than the alternative, surely.
Setting aside the dystopian conclusions of that mindset, and setting aside that a society of morons wouldn't be able to honestly evaluate whether their society is preferable to another, my answer to this question is the same as Mr. Souther's. Knowledge, skill, and education are intrinsically valuable, at the individual level. My basic axiom is that a person is better off the more capable they are, regardless of whether they get anything out of it.
Agree or disagree, the logical conclusion from this axiom is that a society of people who blindly trust their chatbot to do things for them is intrinsically worse off, because (as any educator will tell you), watching someone or something else do something does not impart the same knowledge, ability, or wisdom as doing it yourself.
Skill and knowledge require constant practice. I barely know Latin any more. If you took me right now and sat me down in front of the qualifying exams I passed, I probably couldn't pass them, and not just because it's late at night and I'm tired. At best, you could say that I know what I don't know when it comes to subjects I am out of practice in. I think that as people come to rely on chatbots, they start to practice less, and they start to lose some of their skill and knowledge. This is a bad thing, and if I'm being honest it is the main reason I remain hesitant to adopt chatbots in my work. I simply value the knowledge and skill I get, or maintain, by doing things myself, even if the material outputs are worse or slower or both.
Maybe knowing what you don't know is enough to take you out of moronhood, when it comes to using chatbots. I hold myself to a slightly higher standard, but it's not something I would judge another person for. One way or another, though, you can't substitute a chatbot for knowledge. And likewise, as we go in to the century of what has to far contained economic malaise, mass extinction, and political instability, we can't allow chatbots and their marketing teams to decrease whatever regard for knowledge we still collectively possess. We're going to need it.