The ChatGPT AI developed by the OpenAI startup opened a new cheating hole for many students who kept trying to prove its effectiveness. Indeed ChatGPT has shown itself capable of creating dissertations from scratch, which has temporarily turned teaching and learning methods upside down in higher education.
At the University of Strasbourg, in France, around twenty students, who had cheated using this tool during a distance exam, then had to take it again face-to-face. ChatGPT and other artificial intelligences have therefore been banned from schools and universities all over the world, like Sciences-po Paris which has asked all its students not to use it. The Minister of National Education, Pap Ndiaye, even mentioned more global measures last Thursday on France Inter. “We will have to intervene on this, we are thinking about the right way to intervene” he declared.
Fraud detectors that teachers can use
Several initiatives have been developed to enable teachers to detect fraud when in doubt. Starting with OpenAI, which is very well placed to recognize the AI work of its own robots but also of any other AI. The American company provides its detection tool called “AI Text Classifier” which gives teachers the level of probability that a text has been generated by an AI. The tool is rather limited at the moment: limited to 1,000 characters, and its reliability rate is only 26%.
Also in the United States, the prestigious Stanford University has announced that it is developing a counter-attack to ChatGPT itself: “DetectGPT”. Five researchers from this university located south of San Francisco have developed this service. Although it is not the first tool to emerge since the ChatGPT controversy, DetectGPT is however the one that seems to be the most reliable for the moment. It would recognize texts written using an AI with nearly 95% success.
Another tool “GPTZero” searches the perplexity score of texts. Developed by Edward Tian, a student from the American University of Princeton (New Jersey), the tool is also free and it is clearly intended for teachers by allowing them to test a set of text files on the fly. Finally, the “Draft & Goal” tool is also widely used and promises a reliability rate of 95%: you have been warned!
Nightmare or super-tools for teachers?
The chase against ChatGPT is seen as a “short-sighted” reaction for its defenders. Just like the calculator or automatic translation tools, spell checker, AI can be a great teaching aid tool. Technology will always be a help but students will ultimately have to know how to analyze, decipher and understand; national education and higher education must take this tool into account in order to ask students for more and in a different way.
The best weapon against ChatGPT will be to reinvent students’ homework and tests. As Elon Musk, one of the founders of OpenAI, said in January, “It’s a new world. Farewell, homework! “.