Why universities are banning ChatGPT | Explained

Following a global trend, Bengaluru-based RV University has issued an on-campus ban on ChatGPT, the artificial intelligence (AI) tool that has taken the world by storm. Students at the institute are banned from using the chatbot for academic assignments such as exams, labs and home works.

According to a Hindustan Times report, the RV University ban also applies to other AI tools such as GitHub Co-Pilot and Black Box. Surprise checks will be conducted and students who are found abusing these engines will be made to redo their work on accounts of plagiarism.

Launched by OpenAI in November 2022, ChatGPT is a language processing AI model that is capable of generating human-like text and uses a reinforcement learning method. The chatbot has multiple use cases from serving as customer support to playing chess. The revolutionary piece of technology amassed over a million users within five days of its release.

The popularity of the chatbot and the ethics surrounding it, however, is not lost upon academics and those working within the field of education. The text-based bot has been used, and maybe in some cases abused, by students who have taken advantage of it to turn in assignments and essays.

Already prohibited in some public schools in New York City and Seattle, ChatGPT was also banned at top French university Sciences Po as it was "raising important questions... with regards to fraud in general, and particularly plagiarism," the institute said on Friday. Punishment for using the software may go as far as exclusion from the institution, or even from French higher education as a whole, it said.

Several recently published white papers have found that the chatbot was able to pass a University of Pennsylvania’s Wharton School of Business final exam, four exams from University of Minnesota's Law School and parts of the U.S. Medical Licensing exams. However, it doesn't necessarily ace those exams rather completing them with a low but passing grade.

Professor Christian Terwiesch from Wharton's MBA program found that while ChatGPT did an "amazing job" at basic operations management and process analysis questions, it made "surprising mistakes" in relatively simple calculations in which it was "below the academic performance of a middle school student."

Terwiesch also found that the AI bot was able to correct itself after receiving a hint from a human expert. His findings were echoed by researchers from the University of Minnesota Law School who came to the conclusion that while the AI tool "wasn't a great law student acting alone." They both emphasised the need to have a human in the loop. Similarly, as medical exams are proctored and have an oral component, a human couldn't pass the certification by using the chatbot.

Interestingly, researchers across the three professional fields — law, medical and business — also pointed out the potential of the chatbot to assist students with their education and further help in the actual practice of their learning. Dr. Tseng from California outlined how ChatGPT's generative model could provide doctors with inputs to form their own opinions by comin up with diagnoses by absorbing medical information and tracking historical information.

Recently in an interview with CNBC-TV18 in Davos, Coursera CEO Jeff Maggioncalda even said that ChatGPT could "change education." He, however, warned that users must have good critical thinking to point around the chatbot's dangers despite it being a revolutionary technology.

The chatbot has also been accused of fake generations which scientists call "hallucinations" like when it provided a bogus citation-backed response when asked by a Zurich-based data scientist about a non-existent "cycloidal inverted electromagnon."

"This is where it becomes kind of dangerous," the scientist said. "The moment that you cannot trust the references, it also kind of erodes the trust in citing science whatsoever."

Academics are also worried as students may use ChatGPT-like tools to cheat on tests or turn in unoriginal essays. The advent of AI tools and their normalised usage of them could lead to a decay of critical thinking among students, some believe.

So much so, that a college student from Princeton University even built an app that tells professors whether AI has authored some text, counteracting ChatGPT's extraordinary ability to mimic human language.

Created by 22-year-old Edward Tian, GPTZero is a bot that helps fight AI plagiarism and forgery by distinguishing writing by a human versus AI. It does so by measuring the "perplexity" or complexity of the text and its "burstiness" or the variation of sentences. Text that scores high on both those indicators are more likely to have been written by a human as bots are usually trained on low complexity texts with uniform sentences.

This lapse, however, is not lost upon developer Open AI whose AI safety team is reportedly working on a way to "watermark" GPT-generated text so that its source can be identified through an "unnoticeable secret signal," NPR reports.

While the world gushes over this new, powerful invention, it remains to be seen how we grapple with new technologies and learn to coexist.