A recent column published in The Crimson White argued the merits of AI in college education, heralding the technology as an oncoming “revolution” that educators need to adjust to. After all, generative AI is going to be a part of most people’s daily lives for the foreseeable future. It’s going to replace all sorts of jobs, people like writers, composers, coders…copy editors.
While the CEOs of consumer AI companies like Anthropic’s Dario Amodei might love this narrative to be true, the actual results of the technology have been far more muted. For the average student, this ChatGPT revolution is only revolutionary for academic dishonesty, and some universities, including The University of Alabama, are still embracing the technology with open arms, to mixed effects.
Large language models like Open AI’s ChatGPT or Microsoft’s Copilot, commonly referred to as AI, are probability models. They intake data and, through a complicated process involving trillions of individual calculations, spit out words in an order that resembles the most probable human output.
While it would be reductive to say that AI as a whole isn’t a groundbreaking technology with massive research implications, the use-cases for LLMs, especially as an academic tool, are far less clear.
They always have a chance of “hallucinating” inaccurate information, are vulnerable to bias from its training data and have a bad habit of disseminating major disinformation. These traits limit AI’s viability in a professional setting, and it’s starting to show.
From the statistical side, the Census Bureau recently reported that the AI adoption rate for large companies is beginning to decline, and the Dallas Fed reported that fears of AI job replacement are likely overblown, at least for now.
In terms of individual applications, a study from Model Evaluation and Threat Research, an AI research group, found that programmers who used AI were actually slower than their AI-foregoing peers. Separately, researchers have found that programmers who use AI tend to create more security flaws. Other studies have shown employees tend to negatively judge co-workers that use AI.
If LLMs aren’t good at business or coding, what are they good at? Considering their limitations and strengths, the best real use-case for AI is cheating.
The potential of the AI-cheating market hasn’t gone unnoticed. Tech startup Cluely, founded by a Columbia student who was expelled for — unsurprisingly — cheating, has raised millions of dollars on the promise that its technology will help consumers “cheat on everything.” Other AI companies are certainly aware of this demographic too: OpenAI made ChatGPT Plus free for college students during the spring 2025 finals season.
Some schools are resorting to testing by blue book, cutting out technology entirely to ensure no cheating takes place.
Microsoft and Google have both promoted their respective premium AI services on campus, with the former offering Copilot access to all students through the Microsoft Office Suite. The University is also investing heavily in proprietary AI models, including a new “BamaGPT,” with the construction of an AI data center designed to locally host that technology.
While the potential of AI cheating looms over higher education, many institutions are simultaneously investing heavily into the technology, including the University, which is currently building out AI infrastructure and allowing multiple companies to sell AI products to students.
Even those that do wish to do their own work can’t escape AI, reporting anxiety over false accusations of AI abuse as faculty try to scan their work with often inaccurate AI-checkers like Turnitin and GPTZero.
It seems that, at least at the University, parts of academia have already embraced AI and its students’ usage of the technology, but those very same tools have provably eroded skills that will be essential for students’ futures.
A pre-review study headed by MIT researcher Nataliya Kosmyna found that students who used AI to write essays showed worse neurological performance, struggling to recall information from their “writing” and showing weaker brain activity. Scientists have also published papers warning about using LLMs for academic research, citing their tendency to generate misleading information.
AI is important, and the ramifications of its abilities will be widespread for years to come. Students need to learn the limitations of the technology and how to use it correctly, but let’s not be naive: Academia is already adjusting to AI, and it’s not going very well.
