How Dassault Systèmes Is Building AI That Understands Physics - Ep. 296

In short

In this episode of the NVIDIA AI Podcast, Noah Kravitz interviews Nicolas Souryier from Dassault Systèmes about the next generation of agentic AI systems. They discuss industry world models, virtual companions, and the underlying technologies, highlighting the partnership between Dassault Systèmes and NVIDIA. The conversation covers the principles behind these systems, how they differ from traditional AI, and the future vision of autonomous, proactive virtual companions.

Key takeaways

• Agentic AI systems are transforming industries by enabling virtual testing and optimization before physical creation.
• Industry world models are grounded in science, fueled by industry knowledge, and sovereign by design, providing a more reliable AI than observation-based generative AI.
• Virtual companions augment human workers, freeing them to innovate and solve complex problems.
• Trust in agentic systems is built through scientific foundation, human oversight, and transparent data management.
• Dassault Systèmes uses a hybrid approach to AI models, combining proprietary and open-source solutions to meet performance and regulatory needs.

Chapters

1. 00:00

   Introduction to Agentic AI Systems

   Noah Kravitz introduces Nicolas Souryier, VP of 3D Experience Platform R&D at Dassault Systèmes, to discuss agentic AI systems, industry world models, virtual companions, and the systems driving them.

2. 01:22

   Dassault Systèmes and the 3D Experience Platform

   Nicolas explains his role and the 3D Experience Platform, which serves as the foundation for Dassault Systèmes' 12 brands, including CATIA and SolidWorks. The platform enables customers to imagine, design, simulate, and build almost anything.

3. 02:58

   Virtual Twins Explained

   Dassault Systèmes provides factories for customers to create virtual twins, which are scientific, multidisciplinary, multiscale representations of products. This allows products to be tested virtually under real conditions before physical creation.

4. 03:58

   Transforming the Platform to Agentic AI

   Nicolas discusses transforming Dassault Systèmes' platform architecture into an agentic platform, shifting from a SaaS model to an agent-as-a-service model to bring AI to all customers.

5. 04:39

   Virtual Companions and Industry World Models

   The discussion shifts to defining virtual companions and industry world models, explaining their use within Dassault Systèmes and how they differ from traditional generative AI.

6. 05:09

   Industry World Model Principles

   Nicolas details the three core principles of AI for industry: grounded in science, fueled by industry knowledge, and sovereign by design. This contrasts with classic generative AI, which learns from observation without scientific foundation.

7. 08:33

   Virtual Companions in Action

   Virtual companions turn the intelligence of industry world models into action, understanding user intent and reasoning with industry world models to orchestrate actions within the context of the user's business. The goal is to augment, not replace, human workers.

8. 10:14

   Building Trust in Agentic Systems

   The foundation for trust includes scientific background, human-in-the-loop accountability, and IP lifecycle management (IPLM) to enforce lineage and traceability of AI interactions.

9. 11:46

   NVIDIA Technologies and Agentic Workflows

   NVIDIA technologies infuse every layer of Dassault Systèmes' architecture, from AI factories and GPUs to CUDA-X libraries and Omniverse technologies, accelerating AI training, inference, and simulation. The partnership focuses on understanding, reasoning, and execution.

10. 15:11

    Open vs. Proprietary Models and Hybrid Approach

    Dassault Systèmes uses a hybrid approach, building its own models while relying on best-in-class models from NVIDIA and other providers. Model selection is based on performance, sovereignty, and regulatory constraints.

11. 18:54

    Use Case: Recreating Aircraft with Virtual Companions

    Nicolas shares a use case where a customer, Nayar, is recreating virtual twins of existing aircraft without access to original designs, using Leo, the mechanical designer virtual companion, to automatically generate 3D parts from multiple sources.

12. 20:58

    The Future of Agentic Systems

    Dassault Systèmes' strategy aligns with NVIDIA's announcements on Nemotron, AIQ, and agentic technologies. The vision is for virtual companions to continuously monitor and proactively optimize factories and supply chains in real-time, creating a closed-loop autonomy.

13. 22:48

    Learning More About Dassault Systèmes

    Listeners can learn more on the Dassault Systèmes website (3ds.com) and LinkedIn page, where they communicate more about AI and their collaboration with NVIDIA.

Get the next The AI Podcast recap

A short recap of every new episode, by email. Free for up to 5 shows.