AU 2024: Autodesk Offers Glimpses of the Future with Project Bernini

Earlier this month, in the sunlit Convention Center a few blocks away from San Diego's popular Gaslamp Quarter, Autodesk CEO Andrew Anagnost delivered the keynote to Autodesk University, the annual Autodesk user gathering. The event drew an estimated 12,000 in-person guests, according to Autodesk PR.

Considering the AI-related acquisitions Autodesk has recently made (such as that of Blank AI) and the AI-powered features it's been touting, the year's AU keynote is no surprise to anyone. “It's still really early, despite what a lot of technology companies are saying. The great sorting out … is just beginning,” said Anagnost. 

Bernini Paves the Way

For Autodesk, Project Bernini exemplifies how design programs like Autodesk Fusion might incorporate AI into the conceptual phase. The proof-of-concept technology developed by Autodesk Research allows users to generate 3D objects using text prompts, hand-drawn sketches, and point clouds. The application replicates in 3D the easy content-generating method pioneered by Generative AI text-to-image models like Midjourney and Dall-E, which are trained on NVIDIA GPUs.

Raji Arasu, Executive Vice President, Chief Technology Officer, Autodesk, revealed a few caveats with the preview technology: “First, Bernini is trained on publicly licensed data, so it's not available for commercial use. Second, it's open to the AI community, so we can make it smarter together.”

The Autodesk Assistant is also evolving, bolstered by AI- and language-driven features. Arasu said it “gives timely insights, whether in our docs, autodesk.com, or in our products. It works as a system that understands you, your industry, and the context of the product.” 

Later, during his meeting with the industry press, Anagnost said, “AI is going to move faster for manufacturing. First reason is, we've been in the cloud longer ... The other thing is, we think AI has a lot of value to add to the manufacturing space ... The manufacturing space in general has shown a great willingness to adopt digital technologies, if they provide real benefit.”

AI Transparency

Arasu revealed Autodesk plans to introduce an AI transparency card, similar to the nutrition and ingredient cards that accompany grocery items. “We think it's a great way to show at least the bare minimum of info people need to know about the AI used in our products,” said Arasu. 

During his meeting with the industry press, Anagnost said, “The training methods [used on Bernini] are data-independent, so we could go to customers A, B or C, and say, would you like to participate in helping us make this model better? There’ll be an evolution as we start partnering with certain customers to train the model on their data.”

This business model opens the possibility that many Autodesk software users in automotive, aerospace, and consumer goods may further refine the foundational model from Autodesk with their own proprietary data and industry-specific rules to develop variants of Bernini capable of generating designs that conform to a user’s brand aesthetics and compliance requirements. Such workflows benefit from NVIDIA RTX™ GPUs with Tensor cores for AI-powered applications found in Dell Precision workstations.

Hardware for AI

On the exhibit floor, Himanshu Iyer, Senior Industry Manager, Manufacturing, NVIDIA, showcased a new offering from the stock-art merchant ShutterStock. “Imagine you’re creating a virtual photoshoot taking place in an office environment, and you need to populate it with tables, chairs, computers, and typical office furniture. Shutterstock, with their  millions of licensed images , has launched a 3D AI Generator based on NVIDIA Edify where you can create the 3D models you need just by entering text or image prompts,” he said.

NVIDIA Edify is described as “a multimodal architecture for developing visual generative AI models for image, 3D, 360 HDRi, physically based rendering (PBR) materials.”

At the conference, Dell showcased its lineup of AI-targeted Dell Precision workstations. “It’s a combination of workflows. It’s never just one system,” said Ken Flannigan, AECO Alliances and Solutions, Dell. “We have from laptops to towers to rugged systems, with neuro-processing units onboard the chipset.”

For AI developers, Dell recommends the Dell Precision 5860 Tower workstation with NVIDIA RTX 6000 Ada Generation GPU; and the Dell Precision 7000 Series mobile workstations with NVIDIA RTX 5000 Ada Generation Laptop GPU.

For data scientists working in AI, Dell recommends the Dell Precision 7680 mobile workstation with up to NVIDIA RTX 5000 Ada Generation Laptop GPU; and the Dell Precision 7780 mobile workstation with NVIDIA RTX 5000 Ada Generation Laptop GPU.

Autodesk to Join the LA Olympics 28

The AU keynote wrapped up with the announcement that Autodesk has been selected as the Design and Make Platform for the LA28 Olympic and Paralympic Games. The company revealed, “A hallmark of the city, the LA Memorial Coliseum will be the first stadium ever to host three Olympic Games. Autodesk’s Design and Make Platform will be used to reimagine the iconic stadium for the LA28 Games' Track and Field events, while preserving its past, present, and future traditions.”

Anagnost said, “I can confidently say that LA will be ready. They are reinventing how the city can prepare for this moment and taking a completely unique approach to change—one that has never been done before.”

More Dell Coverage

Artificial Intelligence for Design and Engineering Workflows

In this white paper, learn how artificial intelligence and machine learning can improve design and simulation.

More NVIDIA Coverage