After GTC, NVIDIA Rides the Generative AI Wave at SIGGRAPH

After the GTC 2023 in March, NVIDIA CEO Huang Jensen once again took to the stage at the Los Angeles Convention Center for the SIGGRAPH (Special Interest Group on Computer Graphics and Interactive Techniques) conference to announce the future of interactive entertainment and technology. 

During the hour-long keynote, he revealed new NVIDIA products and research focusing on generative AI, computer graphics, and the company’s role in OpenUSD developments. The organisation was recently formed to standardise the open universal scene description language for building 3D-enabled products and services.

This year, the Association for Computing Machinery (ACM) is celebrating the 50th SIGGRAPH and the program reflected on half a century of advancement and the current generative AI wave. Here are all the major announcements and updates from the event. 

AI Workbench

The software leader unveiled NVIDIA AI Workbench, a unified workspace for developers to effortlessly sculpt and fine-tune pretrained generative AI models on a PC or workstation — then scale them to virtually any data center, public cloud or NVIDIA DGX Cloud.

The workspace removes the complexity of getting started with an enterprise AI project. Accessed through a simplified interface running on a local system. With Workbench, users can customise and run generative AI in just a few clicks. It also allows them to pull together all necessary enterprise-grade models, frameworks, SDKs and libraries from open-source repositories like Hugging Face, GitHub and NVIDIA NGC using custom data and the NVIDIA AI platform into a unified developer workspace.

NVIDIA Enterprise 4.0 Upgraded

Jensen also proudly introduced the latest iteration of its enterprise software platform, NVIDIA AI Enterprise 4.0. This release grants businesses access to the essential tools for seamlessly integrating generative AI. Simultaneously, it promises the security measures and API stability that underpin dependable production deployments.

New features like NVIDIA NeMo, NVIDIA Triton Management Service and NVIDIA Base Command Manager Essentials are the latest add-on to the suite. NVIDIA AI Enterprise 4.0 will be integrated into partner marketplaces, including Google Cloud and Microsoft Azure. Furthermore, it finds a natural abode with NVIDIA cloud partner Oracle Cloud Infrastructure.

But the spotlight falls on NeMo — an application framework that helps companies curate their training datasets, build and customise large language models (LLMs), and run them in production on a grand scale. With organisations from Korea to Sweden using it to customise LLMs for their local languages, the framework will become the go-to solution for industries.

“Before NeMo, it took us four and a half months to build a new billion-parameter model. Now we can do it in 16 days — this is mind blowing,” CTO of Writer, Waseem Alshikh stated in the press release.

Hugging Face — Training Cluster as a Service

The collaboration hub for the machine learning community Hugging Face will be giving developers access to NVIDIA DGXTM Cloud AI supercomputing to train and tune advanced AI models. The collaboration will bring one-click access to NVIDIA’s multi-node AI supercomputing platform.

As part of the strategic alliance, HF will offer a new service — called Training Cluster as a Service. This avant garde service will simplify creating new and custom generative AI models for enterprises. Powered by NVIDIA DGX Cloud, the service will be available in the coming months.

Omniverse 

The company boasted its major strides with releases announced of its NVIDIA Omniverse platform for developers to enhance 3D pipelines with the (OpenUSD Universal Scene Description) framework and generative AI. Cesium, Convai, Move AI, SideFX Houdini and Wonder Dynamics are now connected to Omniverse via the native software platform — OpenUSD. 

NVIDIA's vision extends beyond software as it is also collaborating with global systems manufacturers to forge RTX workstations optimally configured for Omniverse experience. Powered by NVIDIA RTXTM 6000 Ada Generation GPUs and fortified with NVIDIA AI Enterprise and OmniverseTM Enterprise software, these workstations are designed to develop and create content. 

The release is now available in beta to download for free and coming soon to Omniverse Enterprise.

With the software ripple extending, the company also introduced three new desktop workstation Ada Generation GPUs—RTX 5000, RTX 4500, and RTX 4000—each bearing the signature NVIDIA prowess.

Within this panoramic overhaul, other updates have been made to Omniverse Kit, Omniverse Audio2Face, and Omniverse USD Composer. This follows the recent announcement of NVIDIA joining Pixar, Adobe, Apple and Autodesk to form the Alliance for OpenUSD.

Users can get early access to OpenUSD resources through the NVIDIA OpenUSD Developer Program.

Shutterstock x NVIDIA’s Picasso – A New Stroke

Building on the existing partnership the visual-content provider Shutterstock is providing features to a part of NVIDIA AI Foundation Models, Picasso, to let artists enhance and light 3D scenes based on simple text or image prompts, all with AI models built using fully licensed, rights-reserved data.

The companies have introduced an additional layer for artists — 360 HDRi Video. This feature empowers artists to create and personalize environment maps as per their creative visions (even in panorama!).

Renewed GPU Prowess  

The technology will provide professional visualization workflows, like real-time rendering, product design, and 3D content creation. With the fusion of OVXTM servers and the state-of-the-art L40S GPU, NVIDIA sets the stage for immersive experiences.