What is frugal AI?

🎥 Replay of May 15, 2025 webinar: "What is frugal AI?"

On May 15, an exciting webinar was held on the theme of "What is frugal AI?", organized in collaboration by Telecom Paris Alumni's Numérique et Environnement and big data & AI groups , X-Environnement and Ponts Alumni'sTransitions group.

This event brought together academic experts, industrial players and members of the research community to explore a question that has become central: how can we make artificial intelligence more sober and sustainable?

As AI technologies, and in particular generative AI, become more widespread, their ecological footprint is exploding. Massive consumption of electricity, intensive use of computing resources, increasing demands on networks: today's models pose real environmental challenges. This webinar provided the keys to a better understanding of these impacts, as well as concrete avenues for action.

- Francis Bach, Director of Research at INRIA and member of the French Academy of Sciences, opened the session by defining the contours of frugal AI. He explained the issues involved in the energy consumption of current models, and the limits to their scalability without technological or conceptual breakthroughs.

- Anne-Laure Ligozat, researcher at the Laboratoire Interdisciplinaire des Sciences du Numérique (LISN) and member of the CNRS EcoInfo group, presented a detailed analysis of the life cycle of AI solutions, highlighting the importance of assessing environmental impacts right from the design phase.

- Benoît Courty, co-founder of Codecarbon.io, highlighted tools for measuring the carbon footprint of software, a key step towards making technological development more accountable and guiding technical choices towards greater sobriety.

- Bertrand Charpentier, co-founder of Pruna AI, shared his experience of reducing the CPU consumption of AI models, proposing concrete optimization methods for technology companies.

🎧 By listening to this webinar again, you'll discover how to measure, understand and reduce the environmental impact of artificial intelligence. An essential resource for engineers, researchers, decision-makers or students concerned with combining technological innovation and ecological responsibility.