④ Educating Puppies vs. Higher Education

A major Korean university recently introduced "mega-lectures"; classes of 400 to 1,000 students, delivered in a hybrid format. They held up Harvard's CS50 and Michael Sandel's Justice as models. The pitch sounded reasonable: solve enrollment bottlenecks, expand student choice, raise the intellectual bar for everyone.

There was just one problem. Harvard's large lectures work because of what happens after the lecture. Small-group recitations. Teaching assistants who know each student's name. Weekly sections where twenty students sit in a circle and argue. The lecture is the front door. The education happens in the rooms behind it.

This university built the front door. The rooms behind it don't exist.

I wish I could say this was an isolated case. My own university's innovation office has been pushing 500-person lectures. They call it "educational innovation." When AI can deliver knowledge more efficiently than any human lecturer, the single most important thing a university can offer is close, high-density contact between professor and student. These institutions are investing in the exact opposite.

Anyone who reads the earlier essays in this series will ask: if your lab works so well, why doesn't every classroom operate that way? The answer is not that universities don't want to. It's that they can't. And the reasons interlock in a way that makes any single fix useless.

Start with the ratio. In my department, one professor teaches sixty to seventy students. Meaningful mentoring requires ten at most. We're off by a factor of six or seven, and the gap is widening; administrations are cutting faculty and increasing enrollment, not the reverse.

The obvious response is to hire more professors. But that requires money, and here the numbers become absurd. University tuition in Korea was effectively frozen for sixteen years (Not SIX, but SIXTEEN years! and the reason is obviously politicians). During that time, inflation rose, salaries rose, technology transformed every aspect of education, and university revenue stayed flat. To put this in perspective: a premium puppy daycare in Seoul charges roughly two million won per month. University tuition, when divided across the year, comes to less than 600,000 won, about a third of what people pay to have their dogs looked after. And yet the prevailing public sentiment remains that tuition is too expensive. The result is a vicious cycle: underfunding degrades educational quality, which makes the public even less willing to pay.

So you try to work within existing resources. Redesign curricula. Reallocate faculty time. But the Ministry of Education controls much of the curriculum structure, leaving universities with remarkably little room to innovate on their own. Programs like LABA are possible only because they fit inside an existing institutional loophole. Scaling this to an entire department, let alone a university, is structurally impossible.

And the regulatory instinct is heading in the wrong direction. Last year, the Ministry released draft ethics guidelines for AI use in universities: students must disclose AI usage when submitting assignments, syllabi must specify permitted AI tools, and violations can be treated as academic dishonesty. While my students use AI in virtually every stage of learning, and build systems that would be impossible without it, the bureaucracy's concern is not how to harness this tool for education, but how to police it within the old broken format.

And even if you solved every structural problem(the ratio, the funding, the regulation) you'd still face the final wall: people. Where are the professors who can do this work? Doctoral programs train researchers. They do not train mentors. No one teaches PhD students how to give feedback, how to redirect a student's thinking, how to rebuild someone's confidence after a failed week. Professors who can do these things learned them on their own, through temperament and effort. The system doesn't produce them.

What the system doesn't produce can't scale.

Ratio requires money. Money requires political will. Political will requires regulatory freedom. Freedom requires people who know how to use it. Each wall holds up the others.

Meanwhile, education experts repeat the same advice: make lectures more participatory, redesign assessment, focus on critical thinking. They're not wrong. But this advice always lands on the individual professor. Change your teaching. Innovate your evaluation. As though a single person standing in front of seventy students, receiving no institutional support, operating inside a rigid curriculum, can simply decide to do things differently. That is not a solution.

I know this firsthand. I run LABA without any teaching credit from my university. No funding. No course load recognition. Boot camps during break, weekly lab meetings, constant feedback: they are all on my own time and money. The only reason I continue is a personal conviction that this is the right thing to do. But education that depends on individual conviction rather than institutional design is not sustainable. One professor's stubbornness is not a system.

Here is what I find most troubling. Think about what a large lecture actually is. One person delivering the same content at the same speed to a large number of people. That is precisely what AI does best. Expanding large lectures means competing with AI on AI's home ground. It is a fight universities cannot win. If universities are to survive, they must provide what AI cannot: small-group mentoring, real-time correction of thinking, the kind of human dialogue that reshapes how a student sees a problem.

Instead, they are doubling down on exactly what AI will replace, while cutting exactly what AI cannot.

This is why what works in LABA cannot scale to every classroom. But the part that truly frightens me is something deeper; a consequence that this broken system is about to bring. That's what the next essay is about.