French university Sciences Po is one the first higher education institutions to ban the use of the artificial intelligence text generator ChatGPT to complete assignments, in order to prevent fraud and plagiarism. The company that created ChatGPT said it has released software that educators can use to detect AI-generated text.
Sciences Po sent a message to all students and faculty announcing a ban on all AI-based tools by students and faculty in the network, based in Paris, with campuses around France, including Lille, Toulouse and Grenoble.
“The use of ChatGPT, or any other tool using AI, without the transparent referencing is strictly forbidden in the academic space,” wrote provost Sergei Guriev.
The decision is a first for a French university, and Sciences Po said the penalty for using the tools would be “expulsion from the institution" or even a ban from French higher education as a whole, though it did not specify how the use of the tools would be detected.
Universities and schools around the world have been debating the use of ChatGPT, which was released in November and can be used to generate text in response to simple prompts that can be taken for human writing.
In an experiment, professors at the University of Minnesota Law School had the chatbot answer questions from final exams in several subjects, including constitutional law and taxation, and when graded alongside student answers, it consistently achieved low, but passing grades.
AI cannot be ignored
Many French students have been using ChatGPT to write homework assignments, but Guillaume Leboucher, who created the AI for school foundation, at the Institut de France, is not concerned.
Instead of banning the use of ChatGPT, he would like educators to see it as a tool – a new generation of search engine that allows users to interact in simple, human text.
“The interface has the potential to radically change our relationship with AI in our daily lives," he told RFI.
While there are ethical issues raised by its use by students, “we encourage teachers to have students use ChatGPT, to play with it, while warning them that it will not help them get better grades on their homework if they just cut and paste text generated by a chatbot.”
“The challenges related to the AI-based language generation tools will necessarily and rapidly change teaching and evaluation practices,” Sciences Po provost Guriev wrote.
“We know very well that it will be part of the digital ecosystem,” Sciences Po's Directof of Studies, Myriam Dubois-Monkachi, told FranceInfo radio. "How do we live with it?”
In order to block students' use of the tools, some schools have announced plans to assign fewer take-home writing projects and have more oral and hand-written exams.
Unreliable detection tools
Many are looking for software tools to detect AI-generated text. Some third-party tools, including GPTZeroZ have already been created.
OpenAI, the creator of ChatGTP, which has received billions of dollars of investment from Microsoft, released its own detection tool on Wednesday.
In a blog post, the company said that its AI classifier is a language model trained with pairs of human and AI written text on the same topic.
“We recognise that identifying AI-written text has been an important point of discussion among educators, and equally important is recognising the limits and impacts of AI generated text classifiers in the classroom."
The company said the classifier is unreliable, particularly on short texts of fewer than 1,000 characters.
It also says the classifier is not as effective on non-English languages, making it of limited use for schools like Sciences Po.
"We're making this classifier publicly available to get feedback on whether imperfect tools like this one are useful," OpenAI said.