3DEXPERIENCE WORLD 2024 (Dallas) – “Dassault Systèmes is developing generative artificial intelligence (AI) models to help 3D designers says Gian Paolo Bassi, SVP of 3D Experience Works. Designers will be able to draw, design, assemble and simulate products in our software, including Solidworks. This will contribute to the growth of the 3D design industry and significantly increase the productivity of professionals.”
If generative AI is starting to be presented on the company’s products, it is because according to the manager of the French publisher, Dassault Systèmes has been working on the issue for a very long time. However, the company has not provided a precise timetable for the release of tools with these new AI tools.
“Dassault Systèmes’ AI model for 3D design is different from other general-purpose image models, such as Stable Diffusion’s StabilityAI, OpenAI’s DallE, Google’s Imagine2, and others,” he says .
Suchit Jain, the VP business development of Dassault Systèmes shows how SolidMacro works, a personalized GPT that allows you to create Solidworks macros for product description.
“Faced with the rise in power of generative AI, we still had to adapt,” however, indicates Suchit Jain, VP business development at Dassault Systèmes. And to show for example how SolidMacro works, a personalized GPT which allows you to create Solidworks macros to produce product descriptions.
Real 3D
If Dassault Systèmes appears fairly calm in the face of this new competition, with which it must deal, it is above all because classic AI models generate images in 2D, not in 3D. And even though they create 3D images, they are not realistic. “Currently, general-purpose models do not have the data necessary to create 3D images,” says Gian Paolo Bassi.
The model developed by Dassault Systèmes is different. Not so much on the form of the interface as on the precision of the data. The model works specifically for 3D design. The user asks the AI model via a prompt for a drawing or part model, and the model understands it and creates the desired file. The user also interacts with the AI as it works on the design.
The same goes for model data. Gian Paolo Bassi said the model is currently training on data. It focuses on design image data, not general images. “The only way to make 3D images realistic is to integrate as much specialized data as possible into the model,” he reiterated.
Go beyond model training
Gian Paolo Bassi admits, however, that the model is not yet ready for commercialization. “Our model needs to learn more than a general-purpose model to be able to produce images of sufficient quality to make a commercial product,” he explains, “We need to develop the ability to create products, not just images”.
In fact during a demonstration, Suchit Jain shows examples of GenAI to design chairs, where we clearly see that generative AI certainly works, but that it is not yet able to deliver a product ready to be manufactured.
Suchit Jain shows examples of GenAI for designing chairs. Further development is required to make this technology commercializable.
“Dassault Systèmes has long applied AI technology,” concludes Gian Paolo Bassi, “and we will release the 3D image foundation model as part of our next major product update.”
Source: “ZDNet Korea”