Menu
We are living in an era where computers are increasingly capable of performing tasks that, until recently, were thought to require human intelligence. From natural language understanding to decision-making, these advancements have reshaped our perceptions of what machines can achieve. However, framing AI technology as merely ‘human-like’ oversimplifies the reality of intelligence. Intelligence is not only a cognitive phenomenon but is also profoundly social. It is embedded in our science, societies, and cultures, arising through the intricate interactions among individuals, communities, and institutions.
AI holds the potential to enhance these forms of social intelligence, strengthening our collective capacity to tackle complex and emerging challenges while preserving and enriching the fabric of our shared humanity. This vision of AI as a tool for fostering collective intelligence is pivotal, particularly as global challenges—such as climate change, public health crises, and economic inequality—demand collaborative and interdisciplinary solutions. The aim is not to replace human intelligence but to augment it, enabling communities to thrive and societies to innovate in inclusive, equitable, and sustainable ways.
The concept of collective intelligence is deeply rooted in many areas of modern science. In disciplines such as physics, biology, and mathematics, studying phenomena emerging from the interactions of diverse entities has long been a cornerstone of research. Similarly, economics and the social sciences have focused on designing mechanisms, institutions, and systems that are both effective and resilient in complex and dynamic environments. These fields offer invaluable insights into how AI can be developed, deployed, and governed to serve the broader good.
The interdisciplinary conference “AI, Science, and Society: Connections, Collectives, and Collaboration” will provide a platform for exploring how AI can be informed by and contribute to understanding collective intelligence in science, economics, and beyond. By fostering dialogue between experts in AI, natural sciences, and social sciences, the conference seeks to highlight the synergies between these perspectives—not only to deepen our understanding of the current state of AI technology but also to shape its future development in a direction that aligns with societal values and priorities.
Prof. Michael Jordan (Inria/UC Berkeley)
Prof. Eric Moulines (École polytechnique – IP Paris)
Director General and Acting President of École Polytechnique
President of Institut Polytechnique de Paris
French President's special envoy for the AI Action Summit
Professor in the Department of Electrical Engineering and Computer Science and the Department of Statistics at the University of California, Berkeley / Inria
Poincaré Amphitheatre
10:35 am
President of the Mohamed bin Zayed University of Artificial Intelligence, Professor of Computer Science at Carnegie Mellon University, and Co-Founder and Chief Scientist at GenBio AI
11:15 am
Senior Director of Research & Engineering at Google DeepMind
Arago Amphitheatre
10:35 am
The Barnum-Simons Chair in Mathematics and Statistics, Stanford University
11:15 am
Mathworks Professor of Electrical Engineering and Computer Science, Department Head, EECS, Deputy Dean of Academics, Schwarzman College of Computing, MIT
Moderated by Sébastien Meyer, AI Project Manager at the French Ministry of Ecology
From balancing energy efficiency with computational demands to ensuring transparency, reproducibility,
and ethical alignment, this panel will explore the essential innovations and frameworks for developing robust
and responsible foundation models. Special attention will be given to real-world applications in enterprise,
healthcare, and biology, as well as the emerging role of agentic systems capable of interacting with tools and
environments. In this roundtable, leading researchers and industry pioneers will highlight pathways to
ensure AI’s transformative power remains sustainable, inclusive, and aligned with human values.
VP and Managing Director, AI Frontiers, Microsoft Research
President of the Mohamed bin Zayed University of Artificial Intelligence, Professor of Computer Science at Carnegie Mellon University, and Co-Founder and Chief Scientist at GenBio AI
The Barnum-Simons Chair in Mathematics and Statistics, Stanford University
VP for AI models at IBM Research
Senior Director of Research & Engineering at Google DeepMind
French Minister in charge of Artificial Intelligence and Digital Affairs
Chair : Jamal Atif, Professor at Université Paris-Dauphine – PSL
Foundation models have revolutionized AI by enabling versatile representations across diverse disciplines, from natural language processing to images, videos, and structured data in physics, biology, and engineering. They excel in learning from vast datasets and fine-tuning for various applications with exceptional accuracy and efficiency. Key advancements such as self-supervised learning, attention mechanisms, and scaling laws drive their success.
Despite their progress, challenges in interpretability, efficiency, and domain-specific adaptations persist. Addressing these requires deeper exploration of their mathematical foundations and optimization techniques. This workshop focuses on bridging theory and practice, fostering collaboration between fundamental research and applied innovation.
CEO at Kyutai
VP for AI models at IBM Research
Co-founder and CEO of Instadeep
Chief R&D Officer at Owkin and Co-founder at Bioptimus
Entrepreneur - Ex GenAI Meta and Google Deepmind
Chair : Karteek Alahari, Research Director at Inria
Modern generative AI relies on groundbreaking methods like transformers and diffusion processes, rooted in
advanced mathematics such as probability theory, functional analysis, and optimization. Transformers, with
self-attention mechanisms, revolutionized sequence modeling by capturing complex global data
dependencies, driving breakthroughs in natural language processing, computer vision, and multimodal
applications. Diffusion processes use stochastic frameworks to generate high-dimensional data, refined
through Itô calculus, offering advantages in uncertainty and complexity management. Stable diffusion
enhances this by focusing on numerical stability and robust optimization for high-quality content generation.
This workshop explores the mathematical foundations, scalability solutions, and future directions of these
technologies, emphasizing their role in AI-driven scientific discovery, creative applications, and
computational science’s broader landscape.
Director of the Munich Center for Machine Learning & Chair of Computer Vision and AI, TU Munich
Professor at École Polytechnique - IP Paris
Professor at the Collège de France
Research Director at Inria
Professor at the University of Hong Kong
Chair : Jalal Fadili, Professor at ENSICAEN
While machine learning systems are becoming ubiquituous and are evolving at a fast pace, their theoretical understanding necessiates a diverse spectrum of sophisticated mathematical tools and even rethinking or inventing some of these tools.
These frontiers not only open the door to a deeper theoretical understanding but also allow to improve efficiency, address limitations of the current systems and address some of their practical challenges. The workshop will feature a series of four talks by world renowned experts in the mathematics of machine learning.
CNRS Research Director and professor at ENS - PSL
Professor at the University of Liège
Research Director at Inria, ENS - PSL
CNRS Research Director
Professor at the Centre de recherche en économie et statistique (CREST) - ENSAE - IP Paris
Chair : Florence d’Alché-Buc, Professor at Télécom Paris – IP Paris
This session delves into two core pillars of responsible AI: fairness and privacy. As AI increasingly influences
critical areas like healthcare and finance, ensuring unbiased decisions and protecting sensitive data are
paramount.
The discussion will highlight cutting-edge research and practical strategies to address bias in AI systems,
promote fair decision-making, and build trust through transparency and explainability. It will also explore
innovative privacy-preserving techniques, such as federated learning and differential privacy, that balance
data protection with analytical effectiveness.
A central focus will be the interaction between fairness and privacy, addressing the challenges and trade-offs
in achieving both. Experts from academia, industry, and policymaking will share their perspectives, offering
attendees valuable insights into creating ethical AI systems that meet societal needs.
Director at the Max Planck Institute for Intelligent Systems
Research Director at Inria
Research Director at Agence ministérielle de l'intelligence artificielle de défense (AMIAD)
Professorat École Polytechnique - IP Paris
Regius Professor of Computer Science at the University of Southampton.
5:00 pm
Scientific Director at ELLIS Institute and Max Planck Tuebingen, Professor at ETH Zurich
5:40 pm
Chief AI Scientist, Meta
Professor in the Department of Electrical Engineering and Computer Science and the Department of Statistics at the University of California, Berkeley / Inria
Chief AI Scientist, Meta
Scientific Director at ELLIS Institute and Max Planck Tuebingen, Professor at ETH Zurich
Professor at the Collège de France
Mathworks Professor of Electrical Engineering and Computer Science, Department Head, EECS, Deputy Dean of Academics, Schwarzman College of Computing, MIT
Chair : Stéphane Canu, Professor at INSA Rouen Normandie
While technological developments in the field of AI evolve rapidly, the environmental footprint of AI systems – including energy and water consumption, carbon emissions, and materials needs – calls for a critical analysis.
This workshop explores the complex relationship between artificial intelligence development and sustainability, with a primary focus on environmental considerations. It aims first to present the latest state of-the-art research and ongoing initiatives that leverage AI to achieve sustainable development goals across various sectors. Concurrently, it seeks to address current and prospective strategies to minimize the environmental impact of AI systems themselves, promoting energy-efficient algorithms and sustainable infrastructures.
Associate Professor at Télécom Paris – IP Paris
Professor at Free University Brussels – VUB
Artificial Intelligence Researcher & Climate Lead, Hugging Face
Professor at KTH Royal Institute of Technology
Professor at the Hertie School
Chair : Cédric Auliac, Head of the Artificial Intelligence Program at CEA
This workshop explores how AI is transforming numerous fields across the sciences and humanities. In scientific research, AI accelerates discoveries by enabling the analysis of vast datasets, optimizing experimental design, and uncovering complex patterns beyond human capability. In education, AI enhances personalized learning experiences, supports adaptive teaching methods, and provides tools for greater accessibility and inclusion. Additionally, in the preservation of cultural heritage, AI aids in the restoration of artifacts, digital archiving, and the creation of immersive experiences. By bridging disciplines, AI fosters innovation, reshaping how knowledge is created, shared, and preserved.
Professor at the Technical University of Darmstadt
Professor at ENS - PSL and New York University
Research director at Inria
Scientific Director of Max Plank Institute for Security and Privacy and Professor at the Korea Advanced Institute of Science and Technology (KAIST)
Research Director at Inria and Professor at Carnegie Mellon University
Chair: Isabelle Bloch, Professor at Sorbonne Université, LIP6; Invited Professor at Télécom Paris – IP Paris
Artificial intelligence is transforming medicine and healthcare by enabling breakthroughs in diagnosis, treatment, and personalized care. This workshop will explore cutting-edge AI applications in areas like medical imaging, genomics, electronic health records, and wearable devices. Key advancements include disease detection, patient outcome prediction, and drug development. AI also supports public health by analyzing trends, optimizing resources, and devising prevention strategies.
However, challenges such as ensuring fairness, transparency, interpretability, data privacy, and regulatory compliance remain barriers to widespread adoption. Experts from AI, medicine, and computer science will discuss recent innovations and chart a path forward for AI-driven healthcare solutions.
Technikos Professor of Biomedical Engineering, University of Oxford
Research Director at Inria
Research Director at Inria
Professor at University of Würzburg and Institut Pasteur
President of T-Life
Geneticist, Professor at the Collège de France, Managing Director of the European Molecular Biology Laboratory (EMBL)
1:30 pm
Artificial Intelligence Researcher & Climate Lead, Hugging Face
2:05 pm
Professor at University of Montreal
Plenary followed by a roundtable (see the details below in the special Session dedicated to "Science of AI Safety and Security")
3:05 pm
James Bryant Conant University Professor at Harvard University; Director of GETTING Plurality Research Network
4:00 pm
Professor, Stanford Institute for Human-Center AI and Director, Stanford Digital Economy Lab
4:40 pm
Professor at the Collège de France and INSEAD, Visiting Professor at the London School of Economics and Fellow of the Econometric Society and the American Academy of Arts and Sciences
Moderated by Alice Albizzati, Founding Partner at Revaia
This roundtable will explore the transformative impact of AI on society, technology, and sustainability. Bringing together leading experts in economics, technology, ethics, and innovation, the discussion will delve into how AI is reshaping industries, redefining social interactions, and challenging traditional educational paradigms.
From the regulatory frameworks needed to balance innovation and ethics to the role of AI in addressing global challenges like climate change and inequality, this panel will provide a comprehensive overview of the imperatives for shaping a future built on AI. The discussion will explore the opportunities and responsibilities that lie ahead for governments, companies, and individuals in the AI-driven world.
Professor, Stanford Institute for Human-Center AI and Director, Stanford Digital Economy Lab
James Bryant Conant University Professor at Harvard University; Director of GETTING Plurality Research Network
Professor at University of Montreal
Professor at the Collège de France and INSEAD, Visiting Professor at the London School of Economics and Fellow of the Econometric Society and the American Academy of Arts and Sciences
French Minister for Higher Education and Research
Digital technologies are transforming the world, offering immense benefits for societal and planetary well-being while raising questions about risks, responsibilities, and the future of societies. The Global AI scientific community emphasizes the need for coordination to address these challenges, particularly high-risk AI.
The adoption of the Global Digital Compact (GDC) marks a historic step in global digital governance, aiming to bridge digital divides and establish frameworks for emerging technologies. Key initiatives include forming a multidisciplinary Independent International Scientific Panel on AI and launching a Global Dialogue on AI Governance, with Spain and Costa Rica as co-facilitators. The panel will guide evidencebased policymaking, promote scientific understanding, and support less advanced countries in AI, while the dialogue will foster knowledgesharing, capacity-building, and voluntary standards.
This roundtable will explore how initiatives like the Global Partnership on AI (OECD), G-20, and AI Safety Institute Network can contribute to shaping the UN Independent Scientific Panel on AI, establishing it as a central advisory body for global AI policy and standards.
Algorithmic management tools can offer productivity gains for firms and benefits for workers, such as improved managerial decisionmaking. Yet the realities of increasingly automated and digitalised management can also raise some concerns, and it is crucial that such tools be used in ways that respect the rights and safety of workers.
To delve into these issues, the OECD is launching a new report on the use of algorithmic management tools, assessing their prevalence, their impacts, and current measures used by firms to ensure their trustworthy use. The report draws on a novel survey of over 6 000 firms in six OECD countries. It documents what algorithmic management tools are already widespread. Senior policy makers, experts and social partners will discuss the findings and policy measures to promote the benefits and address the risks of algorithmic management tools in the workplace.
Speakers:
There is an ongoing international conversation regarding the role of thresholds, understood as predefined tolerance levels which trigger action if met, as a tool in frontier AI governance. This session will explore the role of thresholds in frontier AI governance, with reference to thresholds used in safety-critical industries such as aviation and existing thresholds for frontier AI defined by AI companies and global governments.
Director, Oxford Martin AI Governance Initiative
Head of department - Evaluation of Artificial Intelligence and Cybersecurity, LNE
Head, OECD AI and Emerging Digital Technologies Division
Panel 1: Why (or why not) thresholds?
Executive Director, Frontier Model Forum
Seconded Expert, EU AI Office
Head of Research, Safer AI
Head, NEA Division of Nuclear Safety Technology and Regulation
Panel 2: Setting thresholds in practice.
Chief Technical Officer, UK AI Safety Institute
Research Scientists, Google DeepMind
Senior Director for Global Public Policy, Microsoft
US science envoy AI, CEO at Human Intelligence
Associate Professor, George Washington University
This session will highlight the findings of the International Safety Report on AI Safety. It will comprise a keynote address by Yoshua Bengio, as Chair of the report, and will be followed by a panel discussion and audience Q&A.
Full professor at Université de Montréal, Founder and Scientific Director at Mila
Professor in Computer Science at UC Berkeley
Professor of Brain-inspired AI, and AI Governance, Chinese Academy of Sciences; Member of UN Advisory Body on AI
Director and Cofounder of ELLIS Alicante
National Coordinator for AI, France
This round table explores the impact of international standards and how standards can be used to achieve compliance, accelerate innovation and reach the global market, while guaranteeing trusted solutions for end-users.
Directeur normalisation à l’AFNOR
Chief standardization officer à Naa.ia
Professor at EPFL and convenor of ISO/IEC JTC 1/SC29/WG 1 on JPEG normalization
Technology specialist at the AI Office de la Commission européenne
These sessions explore the challenge of rigorously assessing artificial intelligence systems as their capabilities rapidly expand. Featuring five keynotes from evaluators, it will delve into methods, challenges, and best practices for assessing advanced AI systems.
Keynote 1: PRISM Eval
Co-Founder & Director of Governance and Standardization, PRISM Eval
Co-Founder & Chief Technical Officer, PRISM Eval
Co-Founder & Chief Executive Officer, PRISM Eval
CO-Founder & Chief Science Officer, PRISM Eval
Keynote 2: Apollo Research
CO-Founder & Chief Science Officer, apollo Research
Keynote 3: Metr
AI Policy at METR
Member of Technical Staff, METR
Keynote 4: European AI Office
Founding member of AI Safety Unit at European AI Office
Technology Specialist (AI Office)
Keynote 5: ANSSI
Directeur général de l’ANSSI