The dead Internet theory—the idea that a large portion of social activity on the Internet is composed of interactions between bots—was once a provocative idea. It is now trite, because it is true to some degree and everybody knows it. This has not prevented people from using social media, but it has made them worse for doing so by making them more biased and reducing the quality of online discourse. Rather than indicating that humans have left cyberspace, the “death” of the Internet refers to the demise of its role as a forum for the transmission of knowledge.
Over the past two quarters, I have taken three computer science (CS) courses. In two of them, professors used Claude to write the slides. That is a sensible choice. In my experience, lecture attendance in CS courses decreases by about half by the end of the quarter. The half that stays is only half-listening. Professors are therefore working with very little bandwidth. Those who are actually paying attention are the most interested in the material and are therefore least in need of direct instruction. Many of the rest will self-study using chatbots, do acceptably on exams (having memorized the material), and ace their projects. They will get a good GPA.
If professors are going to teach into the void, they might as well have the bots do it for them; they have better things to do. If they must write assignments that bots will complete in their students’ stead, they might as well have bots do the writing. If courses are to be for the bots, they might as well be by the bots. I am sure the use of AI in teaching is not ubiquitous, because most professors enjoy teaching. But I enjoy talking, and I can only speak so much when nobody is listening.
Perhaps I should congratulate my computer science professors for being at the cutting edge of educational technology, but I am concerned about what this edge will cut. The greatest risk is a decline in the effectiveness of the lectures themselves. I cannot definitively say that AI use has adversely affected lecture quality, but I have found the lectures in which it was used to be noticeably less information-dense. More importantly, however, the information gained from these lectures is identical to the information gained from pasting a lesson plan into a chatbot—it is shorn of the perspective that gives a professor their standing. While this might not matter much for introductory courses—which aren’t even taught by professors in some departments—the more advanced the course, the more relevant an instructor’s own expertise is to the material being taught. Conveying such difficult topics in a way that students understand requires a level of familiarity with the material that large language models (LLMs) do not possess.
Nevertheless, I would prefer an engaging, excited lecturer using LLM-generated slides to a boring one who has painstakingly crafted their slides over years of teaching. It is also true that LLMs are not the only agents capable of producing bad slides—humans, too, count that as one of their many talents. If a professor is to reduce the effort that they commit to teaching, then there are far worse ways for them to do so than to employ the “great averager” to generate their course materials.
Indeed, AI hardly acts alone in degrading the quality of a university education. Online learning has enabled universities to teach more people—without expanding their educational and administrative capacity—at the cost of meaningful interaction between students and educators. Universities employ more and more non-tenure-track faculty to expand their teaching capacity—without breaking the bank—at the cost of higher standards and the wellbeing of these instructors. Research grants have become increasingly competitive, forcing professors to spend greater portions of their time writing grant applications at the expense of both research and instruction. Grade inflation has enabled universities to improve apparent outcomes while lowering educational quality. At the same time, a college diploma has become essential for the vast majority of prestigious and well-paying jobs, with schools seeking to capture a share of the ever-growing pools of tuition, donations, and funding. To universities, education has long ceased to be of greatest concern.
Now, LLMs will enable professors to reduce their teaching workload, whether to ameliorate the issue of overwork or to enable them to focus on their research instead. Students, forced to compete with ever-greater intensity as their job prospects stagnate, must spend as little time actually learning as possible so that they may develop their careers by more efficient means—RSOs, recruiting events, and startups. They are empowered to do so through the gradual decline in standards that grade inflation has precipitated, enabling them to perform sufficiently well while having LLMs do their coursework.
AI, then, does not make professors abdicate their role as teachers, just as it does not make students try less in their courses. It is merely accelerating the existing trends that drive both instructors and students toward maximum efficiency. The purest realization of these incentives could soon become a reality: an expensive diploma that signifies some abstract notion of aptitude and nothing more.
The networks that animate university life have long been in decline. Soon, they will be taken over by the bots.
Aahaan Singh is a fourth-year in the College.
