Home > TERATEC FORUM > Workshop 8

Teratec 2022 Forum
Wednesday, June 15 - Application Workshop

Workshop 08 - 2:00 to 5:30 pm

High performance AI in the industry
Chaired by Cristel Saudemont, France Director, Supercomputing & AI , Higher Education and Research and Frédéric Parienté, senior manager in the solutions architecture and engineering group, Nvidia

© TERATEC 2022

After revolutionizing our smartphones, artificial intelligence (AI) is transforming all industries and science. This workshop presented various application areas of AI for industry, whether they are existing uses that are directly transferable, such as automatic natural language processing (NLP), or new techniques specific to the industry, e.g. numerical simulation and optimization problems.

First, the speakers from Hugging Face and then LightOn presented large language models (LLMs) which have demonstrated great versatility, both in the NLP domain with applications for text summarization, translation, recording transcription, etc., and beyond with applications for computer code generation, machine vision, reasoning (Socratic models), etc. While the entry ticket to train such LLM models is high ($5-10M), the French community has equipped itself with the Jean Zay computer hosted by IDRIS and grouped in the BigScience project to democratize these LLMs and teach them the French language.

In the second part, presentations from Altair, CERFACS and NVIDIA give many examples of applications of new AI techniques for the engineer's tasks, be it design, manufacturing or decision making. Today's engineers use accurate numerical tools, but they are complex, expensive, and largely monolithic. Deep Neural Networks (DNN) trained on either experimental data, direct numerical simulation data, or physical equations (PINN), offer great possibilities: better parameterization of sub-machine phenomena, exploration of the design space in real time, collaborative work, multi-physics optimization, etc. A key condition for the democratization and effective adoption of these AI techniques will be their seamless integration into current engineering tools.

Finally, the last speaker from Artelys reminds us that more elementary techniques, such as classical optimization or statistics, can fulfill the mission, and that they should not be sacrificed at the altar of deep learning (DL).

With the participation of :

Cristel Saudemont, France Director, Supercomputing & AI , Higher Education and Research and Frédéric Parienté, senior manager in the solutions architecture and engineering group, Nvidia


BigScience: Collaboratively training a large multilingual language model
Lucile Saulnier, Machine Learning Engineer and Thomas Wang, Machine Learning Engineer, Hugging Face

Download the presentation
Abstract & Biography

Practical lessons from training extreme-scale language models for the multilingual Muse API
Julien Launay, PhD Student at ENS/Machine Learning Research Scientist, LightOn

Download the presentation
Abstract & Biographies

Simulation, HPC et AI pour Accélérer l'Electrification des Chaînes de Traction
Vincent Leconte, Senior Director of Business Development, Electromechanical Solutions, Altair

Download the presentation
Abstract & Biography

What can AI do for High-Performance Computation?
Corentin Lapeyre, Research Scientist, Coordinator of the Helios workgroup, CERFACS

Download the presentation
Abstract & Biography

Physics-Informed Neural Networks with NVIDIA Modulus: Application to external flow problems
Niki Loppi, AI/HPC solutions architect, Nvidia

Download the presentation
Abstract & Biography

Robust optimization in power networks
Jean-Luc Bouchot, Engineer, Artelys France

Abstract & Biography

For any other information regarding the workshops, please contact :

Jean-Pascal JEGU
Tel : +33 (0)9 70 65 02 10
2, rue de la Piquetterie


© Ter@tec - All rights reserved - Lawful mention