|
10-11 Mar 2026 Paris (France)
|
|
|
|
|
Welcome Discover the Future of Science at the intersection of AI and Exascale computing !Institute Henri Poincaré, Paris, France Amphitheatre Hermite March 10th - 11th, 2026 Location: 11 Rue Pierre et Marie Curie, 75005 Paris, France This event belongs to the CNRS AISSAI Center thematic trimester : "HPC and AI convergence at the exascale era"
In recent years, the rise of large computational infrastructures has been instrumental in the progress of AI, with groundbreaking applications such as ChatGPT or AlphaFold requiring massive resources for their training. However, to transfer this progress to the broader scientific landscape, we must question the interplay between AI and Exascale computing. The SCOPE conference explores this convergence by investigating how we can build Foundation Models for Science, particularly when faced with the "data dilemma"—where simulations are prohibitively expensive or the resulting data is too vast to be stored.
Beyond model creation, we address the architectural integration of these tools through Model Hybridization, embedding AI directly into simulation methods to achieve unprecedented speed and accuracy. This evolution necessitates a shift in how we manage the machines themselves; we examine AI for HPC, utilizing machine learning to steer complex simulations in real-time and automate the generation of high-performance code. By coupling these domains, SCOPE aims to bridge the gap between raw computational power and transformative scientific discovery, establishing a new paradigm for research at the exascale. The SCOPE (Science at the Convergence of AI and Exascale Computing) series kicks off with its inaugural conference, exploring the synergy between AI and High-Performance Computing (HPC) in the Exascale era. This event will explore how AI is revolutionizing HPC and how HPC infrastructure is enabling AI to scale—with a focus on real-world applications across scientific communities. Featuring overview keynote talks and open-ended roundtable discussions, the conference will tackle the most pressing challenges and opportunities at the intersection of AI and HPC, shaping the future of scientific discovery. Don’t miss this chance to engage with leading experts and be part of the next wave of computational innovation ! Registration- Participation is free of charge, but registration is mandatory. Please register here before March 2, 2026 All the talks will be at the Auditorium Hermite.
Program - Tuesday, March 10th, 2026 - Theme: Convergence HPC/AI (Day 1)
8.30 AM - 9.15 AM (CET) Registration of participants and welcome coffee ☕. 9.15 AM - 9.30 AM (CET) Welcome and opening remarks 9.30 AM - 12.30 PM (CET) Morning session - Convergence HPC/AI
This morning session will bring together a set of keynotes examining how High-Performance Computing and AI are progressively fusing into a unified paradigm for scientific discovery. They will explore how AI plays a role in HPC from acceleration or compression to analysis and orchestration of computing at scale. Moving beyond simple acceleration of existing codes, the talks will show how AI components are embedded within large-scale simulations, redefining notions of performance from raw FLOPs to “science per unit of compute.” The role of generative models will be highlighted, not only as tools for code synthesis and data augmentation, but as engines for proposing new experiments or emulating high-fidelity physical processes under uncertainty. By bringing together perspectives on convergence, hybridization and generative AI. The challenges of applying in HPC workflows for scientific applications will further be examined.
Speakers: Mohammed Wahib (Riken - Japan), George Karniadakis (Brown - USA) and Julia Gusak (INRIA - France). Allocated time for each keynote speaker: 40 minutes. Moderator: Jérôme Bobin (CEA - France)
12.30 PM - 2.00 PM (CET) Lunch break
2.00 PM - 5.15 PM (CET) PANEL DISCUSSION: AI for High-Performance Computing? This panel brings together experts to examine the many facets of the evolving relationship between AI and HPC, from deeply embedding hybrid AI–HPC methods into scientific workflows to using AI as a high-level orchestrator of those workflows. The conversation will focus on the practical challenges of deploying AI in HPC environments, the current technological and methodological boundaries, and the long-term outlook for this convergence. Panelists will debate how AI and HPC may progressively fuse, ranging from tightly coupled AI-augmented scientific workflows to profound transformations of the HPC software stack itself. They will also explore the implications of generative AI for HPC, considering it not only as a performance accelerator but as a catalyst for discovery through the rise of AI co-scientists. Finally, the discussion will address the cultural and technical shifts required—both for integrating AI into HPC and for leveraging this fusion to accelerate scientific progress. Panel speakers: Shantenu Jha (Rutgers University- USA), Anshu Dubey (ANL - USA), Paola Cinella (Sorbonne Université - France), and Julie Deshayes (CNRS - France). Moderator: Julien Le Sommer (CNRS - France) 2.00 PM - 3.00 PM Introduction talks of the panelists (10 min per panelist) 3.00 PM - 3.30 PM Coffee break ☕ 3.30 PM - 4.30 PM Round table discussion - Theme: AI for High-Performance Computing? 4.30 PM - 5.00 PM Q&A session between the panelists and the audience following the roundtable discussion 5.00 PM - 5.15 PM Closing remarks of 1st day 5.15PM - 6.30 PM (CET) COCKTAIL & APERITIF 🍸🍹 Wednesday, March 11th, 2026 - Theme: Foundation models for Science (Day 2)
8.30 AM - 9.15 AM (CET) Welcome coffee and breakfast☕ 9.15 AM - 9.30 AM (CET) Welcome and opening remarks 9.30 AM - 12.30 PM (CET) Morning session - Foundation models for Science This morning session will feature a series of keynotes dedicated to the evolution of the use of AI within specialized scientific disciplines. Moving beyond general-purpose architectures, these sessions explore how large-scale pre-training is being adapted for domains where data is often "expensive”, whether due to the cost of physical experiments or the complexity of high-fidelity simulations. Our speakers will detail the design of domain-specific models that respect physical laws and structural constraints, transforming how we interpret multi-modal scientific data. By centering the discussion on the synergy between AI and High-Performance Computing, these keynotes highlight the shift from isolated task-specific models to versatile computational engines that serve as a base for diverse scientific applications. The challenges and hurdles of applying AI on diverse scientific fields will also be discussed.
Speakers: Pierre Gentine (Columbia - USA), Mo Lotfollahi (Cambridge - UK), Heather Kulik (MIT, USA), and Mariel Pettee (University of Wisconsin-Madison, USA). Allocated time for each keynote speaker: 40 minutes. Moderator: Thomas Moreau (INRIA - France) 12.30 PM - 2.00 PM (CET) Lunch break 2.00 PM - 5.15 PM (CET) PANEL DISCUSSION: Foundation models for Science: myth or a reality? This panel brings together experts to debate the practicalities of integrating AI into the scientific workflow. As we frame "foundation models" as a catalyst for discovery, the discussion will revolve around a core question of today’s researchers: how do we balance the hunger for data and resources with the reality of computational costs? The conversation will delve into the feedback loops between learning and simulation, examining how models can be designed not just for accuracy, but for scientific interpretability and reliability. We will discuss the frontier where computation and learning go hand in hand, addressing the cultural and technical shifts required to build a future where AI and HPC are used for the purpose of scientific progress
Panel speakers:Rick Stevens (ANL - USA), Gael Varoquaux (INRIA - France)
Moderator:Pedro Rodrigues (INRIA - France)
2.00 PM - 3.00 PM Introduction talks of the panelists (10 min per panelist)
3.00 PM - 3.30 PM Coffee break ☕
3.30 PM - 4.30 PM Round table discussion - Theme: Foundation models for Science: myth or a reality?
4.30 PM - 5.00 PM Q&A session between the panelists and the audience following the roundtablediscussion
5.00 PM - 5.15 PM Closing remarks of the 2nd day
5.15 PM - 5.30 PM (CET) Conclusion of the AISSAI Opening conference
Practical informations - Wifi access is available through Eduroam
For those who do not have access to eduroam, you can create a CRU account.
|
