Formation IA

• The "Journée du Numérique 2025" focuses on the educational challenges and pedagogical approaches related to AI.
• The French Senate released a report on AI.
• AI as a science is 70 years old and contains numerous legacies, especially in language.
• Mastering correct language leads to better AI results.

The Question of Cheating

• General prompts are bad prompts and waste time.
• AI can assist popular classes at home.
• AI provides assistance 24/7.
• Evaluating AI generatively becomes necessary if continues to assess it as it is currently done, leading to a risk of being condemned to evaluate generative AI.
• Human potential always needs to be the choice that is being evaluated.

CREIA

• CREIA should aim to build with AI instead of confronting it.
• CREIA does not employ AI experts, but seeks to promote AI acculturation in education.
• CREIA is a space for formation and magistracy.

Conference

• A conference from 9:45 to 10:45 will focus on "Intelligence artificielle : entraîner des modèles et leur faire confiance" (Artificial Intelligence: Training Models and Making Them Trustworthy).

• Michaël P., an AI researcher at INRIA (the national research institute for research in science and digital technology) Lille, and Florène C., Head of Research and Development at INSPE de Lille HdF, will be speaking.

• ELIZA was created in 1966 for chatbot therapy, but it was less effective.

• Deep Blue was created in 1996 for chess games.

• AlphaGo was created in 2015 for more complex games.

• GPT Chat was created in 2022.

• The value and definition of intelligence needs to be understood.

• Generative AI acts as an assistant to the teacher.

• Generative AI produces content.

• Simple AI is embedded for adaptive learning.

• Weak AI performs a specific task in a narrow domain.

• Strong AI, which can understand and have feelings, does not exist yet.

• In education, two types of AI exist: simple embedded AI and generative AI.

• AI adapts to the content of students and assists the teacher, but this has been going on for years.

Points of Vigilance

• Training data is statistical.

• Biases appear in illustrations.

• AI hallucinations occur.

• ecological impact is a concern.

• The distribution of the carbon footprint of digital technology in France is significant.

• Terminals make up 79%.

• Data centers and servers make up 16%.

• Infrastructure such as cables and antennas make up 5%.

• It is important to organize regular, methodical, and deliberate questioning to develop students' critical thinking about how AI works and how it is used.

• All important decisions are validated by the teacher.

• Teachers must establish a framework in order to avoid exposing students' personal data to an AI system.