(Paris) The artificial intelligence ChatGPT, capable of writing texts in response to simple questions, has spread like wildfire in the educational world, prompting teachers to question the advisability of banning it or take advantage of it.
In mid-December, just a few weeks after the tool was made available by the young Californian company OpenAI, eight Australian universities announced that they were modifying their exams and considering that the use of artificial intelligence (AI) by students sounded like cheating.
In 2023, their tests will now be “monitored” with “increased use of paper and pen”, said the leader of the “group of eight” Vicki Thomson, quoted on a daily blog The Australian.
More recently, after several media outlets reported on the growing use of the tool by students around the world, particularly encouraged by TikTok videos, New York public schools restricted access to ChatGPT on their networks. and terminals.
The tool “does not help develop critical thinking and problem-solving skills, which are essential for academic success and lifelong success”, says Jenna Lyle, spokeswoman for the education department. of the American city, in a statement to AFP.
ChatGPT is a conversational robot that has been “trained” with huge amounts of data gleaned from the net and can “predict” the likely outcome of a text. But, failing to reason, he produces an astonishing mixture of correct answers and factual or logical errors, more or less difficult to detect.
He happens, for example, to cite the whale shark (a fish) among marine mammals, to be mistaken in the size of the countries of Central America, to “forget” certain historical events such as the Battle of Amiens in 1870 or to invent bibliographic references from scratch.
Within the world of education, however, some voices are being raised to integrate this innovation into teaching methods.
“ChatGPT is an important innovation, but no more than that of calculators or text editors”, which ended up finding their place at school, explains to AFP Antonio Casili, professor at the Institut Polytechnique de Paris and author of Waiting for robots (Threshold).
According to him, “ChatGPT can help to make a first draft when you find yourself facing a blank sheet, but, after, you still have to write, give a style”.
“Counterproductive” ban
The expert also notes that ChatGPT partly reverses the philosophy of teaching, based on the teacher who asks questions.
This time, the student must himself interrogate the machine, “it’s an opportunity for us to see how the students carry out the tasks entrusted to them, to make them work on checking the facts, and to check if the generated bibliographical references are correct”, analyzes Mr. Casili.
For Olivier Ertzscheid, researcher at the University of Nantes (western France) in information sciences, the ban on the tool is in any case “counter-productive”, because it reinforces the desire of students to ‘utilize.
As after the arrival of Wikipedia or search engines, the challenge for teachers is, according to him, “to experiment with the limits” of these tools.
Finally, the response is organized to detect texts generated by an AI. The online service GPTZero, for example, is preparing an offer dedicated to education professionals and OpenAI is working on a “statistical watermark” applied during text generation. Cheaters are warned.