AI & The University

TL;DR: Carson Gross argues that in the age of AI, university CS education still needs to teach students to write code and read code; AI has created both an assessment crisis and a new opportunity to teach with real-world projects. ## Who is the speaker I'm Carson Gross, and I teach computer science at Montana State University. I don't have a PhD, so I'm not one of those "insiders." But I really enjoy teaching. I spent a long time in the private sector, then my startup fell apart, so I went into teaching. I help maintain HTMX, and another library called Fixie Hyperscript, which is a project I hold very dear to my heart. I also wrote that "Grug Brain" article. Today I want to discuss two big questions: Can universities effectively teach students to build software that actually works? And, for those of you who aren't on the front lines of teaching, what does AI actually look like on those front lines? Before I start, I want to quote Upton Sinclair, which helped me understand a lot of things—why, when I try to explain something to people, they just don't get it: > "It is difficult to get a man to understand something, when his salary depends on his not understanding it." To summarize with a meme: Why don't they get it? Please, think of the professor. My view on AI is somewhere in the middle. I'm trying to balance the "AI psychosis" camp against the "if I find AI anywhere I will kill you" camp. ## Universities and AI's "Weird Moment" Universities right now feel like they're at a strange moment in everything. Is a CS degree still worth it? That's a perennial question, because you have examples like Bill Gates, and all those people without formal degrees, or who dropped out, who have been hugely successful in tech. Anyone in this room with enough experience has worked with someone truly brilliant—often one of the best programmers you've ever met—and then you find out they studied poetry, or something. And now things have gotten worse, especially in undergraduate education, where there's a fairly serious AI crisis. One example is UC Berkeley. I did my undergrad there, and actually dropped out of the Berkeley CS program mid-way. They have a foundational intro course, I think it's the Scheme course. Recently the failure rate in this course spiked to 35%, up from around 10%, and they believe AI is a major cause. Another more recent example is an economics professor at Brown University. He gave a midterm exam, done in the very typical university way—online, take-home, do it whenever you want. Then he received answers that were almost impossibly perfect. He was very suspicious: even at Brown, come on, not every kid could have been paying attention in class. So he said: "Alright, for the final exam, everyone comes to the classroom, we'll do it in person." When the students showed up, you could see their scores. In the end, only 23% of the class actually passed the final exam, while on the midterm they'd nearly all received A's. This is really bad. Maybe it's obvious in hindsight, but it caught university professors off guard. But I want to single out one person: Student 22. Student 22 failed both times, but scored better on the in-person exam than on the take-home one. I'm not particularly political myself, but I think we should find this person and maybe organize a political party around them, because I'd rather accept an honest failure than many other types of people. ## Declining enrollment and industry shifts You may have heard—our department is certainly worried about it: CS enrollment is declining. Part of it is due to a bunch of very high-profile big tech layoffs, and a lot of AI fear. I can tell you right now that the environment for my graduating students is very tough, and it pains me deeply, because they're all good kids. If any of you have hiring power, I'd encourage you to recruit these kids—they really need the opportunities right now. There's also a broader industry crisis: software engineers at many companies are now expected to act more like managers—we're managing agents, but are we good managers? I'm not a good manager. I like writing code. I want to write code. Unfortunately, for many people this is being forced, not even optional. I think this will change, because costs will exert their usual pressure, but that's just my guess. The reality is that many companies are discouraging people from writing code right now. I do know some senior developers, very good senior developers, who don't even open their editor anymore. They just don't write code. This seems crazy to me, but you know, I'm an out-of-touch academic, what do I know? That sounds pretty pessimistic, and it is indeed somewhat pessimistic. There's certainly a lot of upheaval happening, and I don't think all of it is good. But I do think there's also something positive. That's why I find myself in the middle of the "doomsday" camp and the "unhinged" camp and all the others—I see the good side, and I see the bad side. ## The contract between universities and students When a student comes to a university CS department, what's the contract you've made? The contract is really this: we're going to teach you computer science, and then we're going to evaluate and certify that you're actually capable. Then you can take this certificate of capability to a company, or do whatever you want with it. They'll accept it, and that's your ticket in—your ticket to a well-paying career. I want to break this into two big parts: the teaching of computer science, and the evaluation of computer science. AI affects both of these in strange ways. ## Should we teach coding or not? Let me take a show of hands: how many people here think CS departments should teach kids to write code? I'm guessing at a conference like this, everyone thinks: "What do you mean? Of course, isn't that what CS is?" But what you might not expect is that there's a strong current of thought in CS education circles that we should NOT teach kids to write code. Kids should learn to code on their own; that's not the department's job, that's "bootcamp" activity. I entered my teaching career through a very non-traditional path. I wrote code for 20-plus years. Inside CS, there's a group of people more enthusiastic about discussing theory and formal methods, who see writing code as low-level work, the kind of thing done by people with jobs. I've talked with these people, not just in my department but in many others. I genuinely think this is a very misguided view. Most CS students won't go into esoteric theoretical work. That doesn't mean they can't appreciate esoteric theoretical work, but they're going to enter the real world, and they're going to work with computers. And to work with computers, at least until recently, you needed to be reasonably proficient at writing code. One point I try to make to some of my peers in CS education: 90% of CS undergraduates will never go to graduate school. And of those who do go to grad school, maybe 1% will become PhDs and continue into academic work. In my view, we've gotten things a bit backwards.