Rue Saint-Guillaume, Paris. “I feel like I’m living a revolution. Célia, a student at Sciences Po Paris, is upset. We are however neither in May 68, nor at the Sorbonne. But at Sciences Po, in March 2023. Nor does she comment on the demonstration in progress, not far from rue Saint-Guillaume, place de la Concorde, which occurred suddenly after the government resorted to 49-3.
No, she’s talking about the ChatGPT phenomenon. Its use by students has been banned for a few weeks. The students received an e-mail prohibiting the use of OpenAI software without the prior agreement of their professors.
Ironically, it was this email from the administration that introduced Célia to ChatGPT. “I didn’t know,” she said. “So I went to see it out of curiosity. This disturbed me a bit. Because it’s very well done. »
“For the moment, as the answers are not detailed enough, we are not going to write ten pages with ChatGPT. It’s more of a stepping stone,” she says. Enough to give “first avenues of research, points of tension which are still quite fair”, details Célia. But ChatGPT should not lead to “a total abandonment of our own thinking”, she specifies.
“He invents a lot of books and quotes”
Mathieu and Daphné, first-year students, are seated side by side in the large amphitheater. They attend a conference organized by the establishment around “higher education and artificial intelligence”.
Mathieu has already experimented a lot with ChatGPT. “Not bad on some points. For example, to ask for an opinion, “what does this author think of this?” “.
“And at the same time, there are a lot of limits. He invents a lot of books and quotes. But it’s much more convenient than search engines. To find good formulas, formulations of sentences, it is rather practical”.
Would he use ChatGPT if his teachers agreed? ” Yes. Because it’s easier. And often if we use ChatGPT, it’s to do simple things. Because complicated things, he does not know how to do them. For example, for the history of Germany very summarized, it is good, because it is simple. »
Daphne has a very different opinion. “I like to produce my own stuff. With ChatGPT, I feel like it wasn’t really me doing the work, I was just asking a question. »
A dichotomy that the lapidary Daphne summarizes as follows. “There is an idea of morality behind the return of work. »
And this is precisely the subject of the debate at this conference.
“A lesson is going to be put on ChatGPT, and we are going to make a Powerpoint of it! It’s called laziness! »
Mathias Vicherat, the director of Sciences Po, clears the bomb from the start. It is not stricto sensu to forbid ChatGPT to Scien Po, which is a “improveable” tool. The challenge is first of all to “see what the implications of this tool are both in terms of pedagogy and in terms of research”.
“In reality, the revolution has already taken place,” says Asma Mhalla, a specialist in the political and geopolitical issues of technology, and teacher at Sciences Po. “The question is ‘who?’ Who creates knowledge? What vision of the world? (…) And implicitly, there is obviously the question of European technological sovereignty…”
“The data is shared without any thought of what is behind it,” warns Cécile Badoual, Vice-President Training at Paris Cité University. “A course will be put on ChatGPT for example, and then we will make a PowerPoint of it! It’s called laziness! It’s fun and it’s a test. However, the course was swung and it was picked up. The data we share can be sensitive, things that can be very important are shared and will no longer belong to the people who gave them. So there is an extremely urgent issue of ethics and raising awareness among students and teachers. »
“ChatGPT is the worst solution”
“This is a very good opportunity to discuss artificial intelligence. But starting from ChatGPT is almost the worst solution,” says Professor Dominique Boullier. “It’s a living room success, certainly, but OpenAI is not open. »
“What’s interesting is that Google and Meta were holding back from publishing their solutions because they were aware of the problems it posed. They had a sense of responsibility. But what OpenAI is doing by publishing ChatGPT is forcing the debate, pushing everyone else to enter the market faster than they wanted. »
“It’s an obvious business strategy, it’s the cultural model of disruptive companies, both anti-state and furiously anti-regulation. And in the educational field, it is still a real problem. “And the researcher to call for validation tests, “as we do in all industries elsewhere”. But this point of view is debated within the school itself.
“Teaching the art of the prompt is also teaching that there are hallucinations”
“Teaching the art of the prompt also means teaching the fact that there are hallucinations from these artificial intelligence algorithms”, defends Alain Goudey, Deputy CEO in charge of digital at the business school and management Neoma BS.
“It’s precisely learning to recognize them, learning to manage them. It is precisely to learn to use this tool for what it should be, that is to say a kind of mirror of thought. It is not a provider of solutions, it must be a tool, which allows humans to go perhaps faster, but above all further in their thinking, in their ability to try to prototype ideas, to make them concrete, to render them into images. »
“It is at this level that we must position the tool. The problem, today, in the field of education, is “I have to give the right answer”. But that’s not the real issue. I prefer to assess students on their ability to think. I prefer that the answer is wrong but that there has been a real investment and an explanation of this process of reflection. »
“And my role as a teacher is to get them to think, and to equip them not to repeat mistakes. The culture of the note, the culture of the right answer is in my opinion something that must be overcome. »
In the end, ChatGPT could quite simply upset teaching, but in a subtle way?
We will have to “re-discuss evaluation methods, training methods, pedagogy itself and student support,” says Cécile Badoual. And the human must not be forgotten. There are a lot of things that go through the oral, through the interaction with the student, that will never be in ChatGPT”.
So maybe the bots won’t replace the teachers.