We're asking the wrong questions about AI and college.

I wonder if you share this experience with me:

Someone you work with — a colleague, a prospective contractor or consultant, a student — sends you a proposal. And when you open it, you immediately recognize the stylistic marks of Claude or Chat GPT. Font sizes, headers, that greenish navy title color, recognizable vocabulary.

And before you even consider the content of the piece, you’re inclined to dismiss it. Your “AI radar” has detected a rat. Hypocritical? Maybe. Like everyone else, I use AI instruments all the time. But I find it difficult to take the time to read a 4-page proposal (or anything else for that matter) if I can tell it is AI-generated.

It’s no mystery why this is the case. AI-generated content doesn’t come from someone. On a certain level, we don’t care if the the quality of thought is good because it isn’t that person’s thought. We don’t even know to what extent they believe the content themselves. At best, we can treat the proposal as a heuristic exercise, a conversation piece that maybe helps us discover what we ourselves believe, where we agree and disagree. It doesn’t matter if the content is better researched and more cogent than it would be if a real person made it. If you’re like me, then you still skim through it with some disinterest.

I wonder if this is a kind of luddite discrimination. Is it something I need to evolve out of? There’s a problem, though, even with the way that question is posed. Since when did technology demand that we “need” to do anything? It doesn’t feel quite the same as adapting to the widespread adoption of the internet because the internet offered information and a platform; it didn’t simply replace our own responsibility to create ideas.

The topic of college always seems to come up when people talk about AI risks.

Whenever I read editorials on AI and college, two additional topics are always thrown in the mix: jobs and the category of “the human.” Three pieces on the topic from this last week are worth mentioning.

One is Jay Caspian Kang’s New Yorker piece, "Will AI Make College obsolete?” The tension Kang detects is that even if colleges fly the flag of liberal education and the importance of learning how to think carefully, learn from the past, and communicate persuasively, they’ll be making this case to young people who lack the intellectual experience required to recognize the good of those capacities in the first place.

Kang writes:

The future of college as we know it may rely on the ability of people who have a stake in the credentialling economy to convince the youth that there is still value in classroom instruction, in writing papers without A.I. assistance, in talking to imperfect humans about misshapen ideas. But they will be making this case to a generation of students who learned many things—skateboarding, the piano, cooking—from YouTube, and who have been able to ask Claude to assist them in every academic endeavor they’ve undertaken.

I’m haunted by this reference to “imperfect humans” because the imperfection of our minds and consciences is part of the human condition and will continue to be so regardless of future advances in technology. …More on this later.

The second piece I read was Adam Harris’s interview with Ian Bogost in the Atlantic, “Colleges are at a Breaking Point.” It’s subtitle reads, “The AI job market has made tuition look like a dubious investment. But it only exposes the deeper identity crisis in American higher education.”

The interview rehearses the usual observations:

• College is too expensive.

• The return on investment isn’t there anymore.

• Colleges are scrambling to figure out what to do about AI.

But it recognizes that people still “buy” college anyway. Most people sort of just cross their fingers.

So there’s a number of different pressures that certainly didn’t exist when I was going to school. It was like, Well, I’m gonna go to college, and everything will be okay. I don’t know how, and then it was. And that’s just not the case anymore.

I’d argue that it is very much still the case that most college-seeking students and families still cross their fingers and bite the financial bullet. What is changing, though, is a growing interest in alternative routes, parallel to the traditional scripted degree pathway.

I appreciate Bogost’s conclusion, but let’s be honest. It lacks persuasiveness:

We don’t know what the future holds. It is changing rapidly. The best thing you can do is build a great deal of knowledge in a large number of domains at a basic level, and prepare yourself for change, rather than preparing yourself for a sort of singular professional life that is sure not to come about in the way that you expected and is sure not to persist in that way for your whole career.

The reason why I ultimately don’t find this convincing — and why the college-going market place also finds it unconvincing — is the ambiguity of the mandate to “prepare yourself for change” rather than follow a typical educational route to law, or a medical profession, or finance. It’s just a hard claim to find a real grip on. That does’t mean it isn’t true. It’s certainly true. Going to college for a singular pre-profession is a bad idea. Polls and research show the growing need for interdisciplinary T-Shaped Skills over narrow professional training, while at the same time employers care less about formal degrees with the growing prevalence of AI in all industries. And as a result, the entry level jobs historically available to college grads are disappearing.

I agree that college students need an interdisciplinary, creative, and intellectual formative education — I built a college around it — but my observation remains: to most, it is an unconvincing claim, at least not persuasiveness enough for them to invest real time and money in for their children.

The third piece I read this week was Pope Leo’s Encyclical, Magnifica Humanitas, “ON SAFEGUARDING THE HUMAN PERSON IN THE TIME OF ARTIFICIAL INTELLIGENCE.” Pope Leo’s sharpest argument is about the dangers of the privatization of AI capacities. Knowledge is for the common good, he says, and the privatization not only of capital but also of learning and knowledge threatens our collective pursuit of the common good.

It’s a position really worth considering. This isn’t just an advance in the “information age.” The information age has become the “learning age,” or the “intelligence age.” Meaning, AI has the capacity to deprive people of their intelligence. If we let it, it can prevent our young people from learning to learn.

The observation that AI threatens the intellectual growth of young adults is well rehearsed at this point.

But we still haven’t found an answer.

Once upon a time, college was for knowledge, and knowledge was for freedom, virtue, and prosperity (college —> knowledge —> happiness).

Then things changed, and college was for a degree and a degree for a job (college —> degree —> job).

Throw AI into the mix. AI is perceived as replacing the need for knowledge and also compromising colleges’ promises to get you a job. Universities are going one of two directions — or they’re combining the two. You can incorporate AI literacy into college education, and/or you can double down on liberal education and the formation of the mind.

It’s telling to me that some of the best AI-centered work I’ve seen at big universities is extra-curricular. One example was an “AI Lab” where students seemed to be doing cool things with AI instruments, but it wasn’t for course credit. Another example was an “AI Prototyping Group” a friend of mine ran at Harvey Mudd. Students made amazing things over the course of a term, but they didn’t it through non-credit-bearing once-weekly lunch meetings. Of course, there are countless AI literacy requirements integrated into degrees at this point, but the best work I’ve seen is outside of the traditional classroom. I think the reason for this is not that professors hold back on AI education out of confusion or fear but because they know, even if instinctively, that there’s a limit to AI education.

And that limit is the shelf life of any current AI training. “We don’t know what the future holds,” as Bogost says.

But chew on this.

AI advances on the information age. But it doesn’t make us smarter. In fact, it’s likely to make us less intelligent — to undermine our ability to analyze arguments and viewpoints, to construct our own arguments and communicate them cogently, to read well. In short, if we’re not careful, AI will take away our capabilities, our power and freedom to do things.

That’s an old challenge. GPS has taken away our ability to navigate cities on our own. Industrialization in agriculture almost took away thousands of years of accumulated knowledge about sustainable farming.

But AI’s threat is different because it isn’t threatening to rob us of knowledge only but also of intelligence. It may be more similar to how computers made our bodies less healthy because more people removed themselves from manual labor.

…Examples abound, but the point is that we’re at risk of losing our power of UNDERSTANDING. Claude makes it less necessary to understand things in order to act on them. And what’s scary about that is that UNDERSTANDING IS A GOOD OF THE MIND. It isn’t an instrumental good, like money or schools. Understanding is a good in and of itself. It is an intellectual virtue, an act of human excellence.

This isn’t another “we still need humanity” argument against AI. It isn’t even an argument against AI at all. That’s because there’s no contest between technology and understanding to begin with. They aren’t in opposition to one another. Technology is an instrumental good, and understanding is a virtue. We can use technology to help us pursue happiness. But we need understanding in order to be happy.

To be clear, when I say “understanding,” I mean the comprehension of why something is what it is. Where it came from. What it’s made of. What defines its form and identity. What makes it intelligible to us as itself. And what it’s for. Unlike many, I believe that AI does indeed have the capacity to perform moral reasoning. Why not? It can be trained according to a system of ethics — be it utilitarian, or virtue ethics, or rationalism. AI can “understand” morality metaphorically. But humans must possess understanding to reason morally, and they must reason morally if they’re to be happy. All of the best paradigms for happiness that know hinge upon this very capacity.

So what, then, should college teach about AI? …”the book test”

Indeed, who knows what the future holds. I’m in line with those who champion an education that trains students in adaptability, resilience, and the sorts of practical skills that prepare grads for an uncertain future, including technical knowledge. But more than that, I’m also in line with those who champion the liberal arts, an education in the arts of reading, communicating, analyzing, computing, … in short, the arts of understanding.

Consider this example. Let’s call it “the book test.” If someone were to write a book about how to use AI technologies now, it would be obsolete by the time it was released. But, by contrast, if someone wrote a book about technology, about intelligence, or about artificiality (the made-ness of things), it could potentially stand the test of time.

College education — if “college” is even the best way to get it anymore — should prioritize those forms of understanding that could be written into a book whose value can last for years or decades. And what better way to acquire that kind of understanding than by reading the texts that already have withstood the test of years, decades, centuries, even millennia?

“It’s not rocket surgery,” as the saying goes. We don’t know what the future holds. We don’t know exactly how AI will change the economy and job prospects for future grads. What we do know is that we ought to invest in things that last. Plato survived every technological revolution since the 4th century BC. So did Aristotle, the Apostle Paul, Boethius, Augustine, Dante, Euclid, Aquinas, Shakespeare, and Newton survive the technological advances that followed them.

Shown in this light, phrases like “the credentialling economy” and "degree platformization” sound silly.

The goods of college are the same as the goods of the mind.