Du tableau noir au chatbot : arriverons-nous à tenir le rythme ?

Except where otherwise noted, content of these slides is licensed under a Creative Commons Attribution-Share Alike 4.0 international license (CC BY-SA 4.0) Du tableau noir au chatbot : arriverons-nous à tenir le rythme ? Pierre Poulain [email protected] Laboratory of Theoretical Biochemistry teach my students P. Poulain | CC BY-SA 2 Topics not covered today Source: The Guardian, 2024. Source: Joule, 2023. Source: Time, 2023. Source: Reuters, 2025. P. Poulain | CC BY-SA 3 Topics not covered today Source: DALL.E, 2024-07-10, prompt: ”Make a picture of a surgeon” Source: DALL.E, 2024-07-10, prompt: ”Make a picture of a nurse” Source: Business Insider, 2023. P. Poulain | CC BY-SA 4 The ChatGPT (r)evolution From a research result to a mass-market product Source: arXiv, 2017. P. Poulain | CC BY-SA 5 The ChatGPT (r)evolution From a research result to a mass-market product Source: arXiv, 2017. Source: TechCrunch, 2025. P. Poulain | CC BY-SA 6 Students are already using gen. AI / LLMs Survey: “Provide or punish? Students’ views on generative AI in higher education” (UK, 02/2024, 1250 students) More than half of students (53%) have used generative AI to help them with assessments. The most common use is as an ‘AI private tutor’ (36%), helping to explain concepts. Source: HEPI, 2024. P. Poulain | CC BY-SA 7 Students are already using gen. AI / LLMs Enquête : “IA et Enseignement Supérieur : Formation, Structuration et Appropriation par la Société” (FR, 04/2025) 19k étudiant.e.s, 6k enseignant.e.s, 7k personnels support Source: https://mission-ia-sup.forge.apps.education.fr, 2025. P. Poulain | CC BY-SA 8 Students are already using gen. AI / LLMs Source: https://mission-ia-sup.forge.apps.education.fr, 2025. P. Poulain | CC BY-SA 9 Students are already using gen. AI / LLMs Projet LLM@UPCité Samuel Bottani, David Janiszek, Pierre Poulain Janvier → avril 2025 / 470 étudiant·e·s / 270 enseignant·e·s https://hal.science/hal-05347662 Enquête sur la perception et les usages des IA génératives par les étudiant·e·s et les enseignant·e·s d’Université Paris Cité L’IA générative est utilisée par 68 % des étudiant·e·s et 57 % des enseignant·e·s. P. Poulain | CC BY-SA 16 Llama4-Maverick (Groq) 12/06/2025 Deepseek-R1 (Groq) 12/06/2025 From chatterbox to bullshitter…. Correct answer: 61,728,394,938,271,605 P. Poulain | CC BY-SA 17 ChatGPT 3.5 12/06/2025 Correct answer: 61,728,394,938,271,605 ChatGPT 4o 12/06/2025 From chatterbox to bullshitter…. P. Poulain | CC BY-SA 18 ChatGPT 5 08/10/2025 Correct answer: 61,728,394,938,271,605 From chatterbox to bullshitter…. P. Poulain | CC BY-SA 19 ChatGPT 5 08/10/2025 Correct answer: 61,728,394,938,271,605 From chatterbox to bullshitter…. P. Poulain | CC BY-SA 20 ChatGPT 5.1 08/12/2025 Correct answer: 61,728,394,938,271,605 From chatterbox to bullshitter…. P. Poulain | CC BY-SA 21 Read the small prints! P. Poulain | CC BY-SA 22 Source: poster of the movie Lost in Translation. Education P. Poulain | CC BY-SA 23 How could we take advantage of LLMs to improve learning and teaching (social) experiences? Source: College Teaching, 1993. P. Poulain | CC BY-SA 24 The homework apocalypse "To be clear, AI is not the root cause of cheating. Cheating happens because schoolwork is hard and high stakes. And schoolwork is hard and high stakes because learning is not always fun and forms of extrinsic motivation, like grades, are often required to get people to learn." -- Ethan Mollic Source: Post-apocaliptic education, 2024. P. Poulain | CC BY-SA 25 Why students “cheat” Source: Inside Higher Ed x Generation Lab. July 2025. 1,027 students. P. Poulain | CC BY-SA 32 The detection illusion Source: arXiv, 2024. P. Poulain | CC BY-SA 33 Source: arXiv, 2024. Pro tip: remove ‘delve’, ‘crucial’, ‘potential’, ‘significant’, ‘intricate’ from your (possibly AI-generated) papers. Source: medRxiv, 2024. The detection illusion P. Poulain | CC BY-SA 34 The detection illusion ($$$) Source: turnitin, 2025. P. Poulain | CC BY-SA 35 How students use gen. AI? Source: Higher Education Policy Institute. February 2025. 1,041 students. P. Poulain | CC BY-SA 36 How students use gen. AI? Source: Inside Higher Ed x Generation Lab. July 2025. 1,027 students. P. Poulain | CC BY-SA 37 How students (and teachers) use gen. AI? Source: enquête « Enquête sur la perception et les usages des IA génératives par les étudiant·e·s et les enseignant·e·s d’Université Paris Cité », 2025. ●comprendre un cours / devoir (82 % )  ●comprendre un article (59 % )  ●aider à la rédaction, correction, traduction ( + )   P. Poulain | CC BY-SA 38 Teachers adapt Source: https://mission-ia-sup.forge.apps.education.fr, 2025. P. Poulain | CC BY-SA 39 AI workout Targeted mentoring: leverage AI to reword assessment items for improved clarity. Guided exploration: encourage AI to propose a variety of possible solutions to a problem, fostering divergent thinking. Critical evaluation: compare AI-generated outputs to detect possible gaps or inaccuracies. P. Poulain | CC BY-SA 40 Course chatbot Source: Inside Higher Ed x Generation Lab. July 2025. 1,027 students Source: https://koro-ai.com/ P. Poulain | CC BY-SA 41 Course chatbot ●+400 pages ●27 chapters https://python.sdv.u-paris.fr/ P. Poulain | CC BY-SA 48 Conclusion: Train yourself … and get your students trained Source: https://mission-ia-sup.forge.apps.education.fr, 2025. P. Poulain | CC BY-SA 49 Conclusion: Train yourself … and get your students trained https://www.brevetai.fr/ (+ démo pour les enseignant.e.s) Projet pédagogique : IA@UPCité (2026-2027) P. Poulain | CC BY-SA 50 Source: DALL.E, 2024-07-10, prompt: ”Make a picture of crazy male and female scientists making experiments” Conclusion: Experiment... Wisely! 10.5281/zenodo.17859062 ChatGPT-like @UPCité https://pleiade.mi.parisdescartes.fr/