A themed breakfast event was held in the salons of Lyon’s City Hall on July 4th a part of the COLT 2025 conference held in partnership with AILyS, a center of excellence in artificial intelligence at the Lyon-Saint-Étienne university site, which includes ENS de Lyon. Here’s a look back at this gathering, which brought together specialists from around the world and highlighted the region's and Lyon-Saint-Étienne’s vibrancy in innovation and research in artificial intelligence.
From June 30 to July 4, 2025, the COLT 2025 conference was held in Lyon—an international flagship event in the field of Learning Theory. The conference was organized in association with AILyS, the dedicated AI center of excellence at the Lyon-Saint-Étienne academic site.
Within this context, on July 4th, a special “AI Breakfast” was held in the salons of Lyon’s City Hall on the theme: “The Role of Fundamental Research in the Age of LLMs (Large Language Models).” Large Language Models are AI programs capable of understanding and generating natural language text, much like a human, by analyzing massive volumes of data.
A Meeting between local stakeholders and international AI experts
This gathering, which included a coffee break and a roundtable discussion, provided an opportunity for AI specialists from around the world to meet local academic and economic stakeholders in an informal setting. The event also gave organizers a chance to introduce these knowledgeable visitors—over a hundred attendees—to the newly launched AILyS center of excellence, before inviting expert speakers from various AI-related fields to speak on key issues in the sector.
Aurélien Garivier, coordinator of the AILyS project and professor at ENS de Lyon, opened the discussion by highlighting the launch of AILyS in April 2025 and the projects that will follow within the academic community:
“Thanks to COLT 2025, Lyon is currently the world capital of AI theory for a week. Beyond this exceptional event, we are working to build a visible, effective, and leading AI cluster—not only for research but also for training and innovation,” he said.
“Our project is called AILyS; it is supported by all academic institutions, universities, grandes écoles, and hospitals that have joined forces in this joint initiative. It is starting today with collaborative funding from these institutions. We’ve just hired a project manager, Alexandra Marquet-Basset, who will join us in August with all her experience and energy.
Beyond the academic sphere, AILyS is also committed to working with the strong industrial ecosystem of our region. In parallel, we are currently identifying a few concrete projects, stemming from our efforts in 2023–2024, that must begin immediately with specific regional support. To launch these projects and this new momentum, a workshop will be held on November 12 at the Digital Campus in Charbonnières—mark your calendars!”
From Machine Learning to Generative AI
Aurélien Garivier then passed the floor to Julien Jacques, professor at the University of Lyon 2, who moderated the discussion between three researchers involved in innovation and experts from distinct disciplines:
- Laurent Daudet: Professor at Université Paris Cité, specialist in signal processing and machine learning, co-CEO and co-founder of LightOn—a French AI company focused on developing generative AI technologies.
- Véronique Maume Deschamps: Professor at Université Claude Bernard Lyon 1, specialist in statistics, director of the Camille Jordan Institute, and former director (2018–2022) of the Agency for Mathematics in Interaction with Business and Society (AMIES).
- Éric Moulines: Professor at École Polytechnique, specialist in statistical learning and signal processing. Member of the French Academy of Sciences.
“Artificial intelligence experienced a revival in the 2010s with the rise of deep learning and results that far exceeded the state of the art in areas such as image recognition,” Julien Jacques began. “But aside from those performance gains, the applications remained within the realm of traditional machine learning. In the 2020s, with the emergence of generative AI and LLMs, a wide range of new applications has emerged.”
After asking the panelists to explain what a Large Language Model (LLM) is, he invited them to weigh in on topics such as:
- “Is AI influencing fundamental mathematics, either as a research tool or as a subject of study?”
- “Are AI and LLMs important topics for the Academy of Sciences?”
- “What are the main challenges that LLMs pose to research in computer science and mathematics?”
- “Has the emergence of LLMs changed the relationship between academia and industry?”