Key Takeaways from NVIDIA GTC 2023
In this blog
NVIDIA GTC 2023 came at a high point of artificial intelligence (AI) conversations in society, and the event didn't disappoint, contributing directly to the most relevant topics in AI.
GTC's primary target audience is developers and integrators focused on new product and service announcements. More than 250,000 people attended the conference that offered more than 650 sessions.
NVIDIA CEO Jensen Huang opened the keynote with a prophetic statement: "AI and accelerated computing's time has come."
As a Platinum Sponsor at NVIDIA GTC 2023 and NVIDIA's Americas 2023 AI Solution Provider of the Year and NPN Networking Partner of the Year, we share our key takeaways from the event.
Key takeaway 1: Generative AI will continue to create more industry-specific assistants to drive productivity
Your workforce will expect co-pilots for everything, and generative AI will create more co-pilots across industries. Chances are you will need help building and controlling these tools as well as aligning them to your business drivers.
NVIDIA is enabling a faster path to value for specific industries, e.g., pharma, chip manufacturing, automotive and creative visual industries. One example is BioNeMo, a cloud AI service for pharma R&D; it helps researchers accelerate drug discovery and development through pre-trained AI, process standardization and easy access with strong GPU-based capabilities.
Another example is CuLitho, which expands the limits and speed of chip manufacturing design and production environments.
Omniverse is NVIDIA's multi-purpose Simulation and Digital Twin Platform. NVIDIA Omniverse uses digital-twin simulation, backed by several physics-driven connections, to emulate physical environments for process improvements. Additionally, Omniverse acts as a highly collaborative platform using the Universal Scene Description (USD) framework.
New connectors announced at GTC expand industrial digital twin capabilities across multiple industry segments. For example, Omniverse is used by Volvo, GM and BMW to optimize manufacturing, reducing costs and test changes in an immersive visualization environment.
New connectors for Siemens Xcelerator portfolio and Rockwell Automation are now available to integrate with applications using the Universal Scene Description (USD) framework. Additionally, NVIDIA's Omniverse Cloud now provides a SaaS version, initially offered from Microsoft Azure.
Together, Omniverse with USD creates an industrial metaverse to test ideas, optimize processes, and develop new manufacturing techniques without disrupting existing facilities and people.
New systems and services broaden the community that can leverage Omniverse. New workstations, laptops, large-scale computing systems and new cloud-based offerings enable more options for Omniverse users. DGX Cloud provides a pre-provisioned setup experience for AI workloads with NVIDIA AI software. Omniverse Cloud is a SaaS Service for ephemeral to longer proof-of-concept or technology assessment projects.
Expanding options for Omniverse platforms, NVIDIA announced next-generation NVIDIA RTX workstations, RTX 500 Ada generation laptop GPU, 3rd Generation of OVX (large-scale digital twin computing platform using NVIDIA L40 GPUs and Bluefield-3 DPUs).
Spinning up a GPU Supercomputer is now as easy as instantiating a VM in your cloud. However, multi-cloud environments can quickly become highly complex, e.g., more data movement challenges, governing yet another cloud and cost management.
NVIDIA's GTC featured several stories from practitioners about their real-world impact. WWT was proud to share our story about the work we did with Freeport-McMoRan.
The joint WWT and Freeport-McMoRan session reviewed risks in mining operations, which typically occur in remote, harsh environments with limited connectivity. For example, wet ore in an underground mine causes blockages and spillages throughout the mine system leading to severe downtime and possible injuries.
Identification of wet ore requires texture characterization in real time to take mitigative actions. After extensive consultations and ideation, a vision-based solution that ran on the edge was developed. The solution included waterproof Smart Cameras with NVIDIA GPUs installed to run models in real time.
For other data work with Freeport, see Democratitizing Data to Reach New Levels of Productivity.
Leveraging NVIDIA DGX Cloud, NVIDIA AI Foundations Services offer allows companies to build, operate and refine large language and generative AI models. Three services offerings are:
- NVIDIA NeMO Service: A cloud service helps developers make large language models (LLMs) more relevant to their specific use cases.
- NVIDIA Picasso Service: Offers advanced text-to-image/video/3D processes and integrated with NVIDIA Edify models. Adobe, Getty Images and Shutterstock are early adopters of this service.
- NVIDIA BioNeMo: BioNeMo speeds drug discovery AI model training and inference engine development. The new release accelerates the time-consuming processes while enabling use of companies' own proprietary information. The service includes pre-trained AI models to further speed drug development projects.
NVIDIA AI Foundations Services are finding increased adoption across industries.
Related NVIDIA news releases:
- AI Foundations
- NVIDIA Launches DGX Cloud
- NVIDIA Unveils Large Language Models and Generative AI Service to Advance Life Sciences R&D
Computational Lithography is the process of imprinting patterns onto semiconductors to be used in computing platforms. Current processes are at their limit to increase the speed of chip development.
NVIDIA, working with semiconductor industry leadership ASML, TSMC and Synopsys, have developed a solution that can increase chip manufacturing by 40x.
NVidias cuLitho library runs on the latest generation NVIDIA HopperTM architecture GPUs. This is another example of the industry-specific solutions NVIDIA develops with key partners from the industry such as ASML, TSMC and Synopsys.
Related NVIDIA news releases:
NVIDIA, in conjunction with Quantum Machines, announced a new system for researchers working with quantum-classical computing.
Quantum computing is a computer science area based on the principles of quantum theory. It addresses problems that are overly complex for traditional computing models using the principles of quantum theory.
One example where quantum computing can apply is the classical application of solving a traveling salesperson's optimal path. Other use cases include cybersecurity, financial modeling and traffic optimization.
For the first time, the combination of accelerated computing with quantum computing provides a high-performance, low-latency platform. This powerful platform enables researchers to speed solutions to some of the most compute-intensive compute problems now and in the future.
Also announced is the adoption of CUDA Quantum into companies Anyon Systems, Atom Computing and QMware.