Summary of NVIDIA GTC announcements

Transformative Moment in AI

Welcome to the 325 new members this week! This newsletter now has 42,711 subscribers

I think Jensen's GTC keynote was the best I have seen since Steve Jobs presented the iPhone. and what a week for Technology and AI! So much has happened, and I've written a full recap to keep you current.

Today Iโ€™ll cover the main announcements:

  • What is the GTC conference

  • The most powerful GPU platform to date

  • Project GR00T: Humanoids are here to stay

  • NIMs, the new Nvidia Inference Microservice

  • Partnerships and integrations

  • Our integrations with the Nvidia ecosystem at IBM

Letโ€™s Dive In! ๐Ÿคฟ

What is GTC

The Nvidia GTC, or GPU Technology Conference, is a global AI conference happening twice a year that brings together developers, researchers, and professionals interested in the latest advancements in AI, computer graphics, data science, and more. It features keynote speeches, technical sessions, workshops, and exhibitions, offering a great opportunity to learn about the future of AI and accelerated computing.

What is pretty cool about GTC is that the main Keynote is live-streamed on YouTube. As of the time Iโ€™m writing this, it has already 22 million views ๐Ÿคฏ

Link to the full GTC 2024 Keynote

Jensen is the new Taylor Swift

This conference has become one of the main tech events of the year. The expectation is similar to the Apple Keynotes in the 2000s and Jensen Huang, the CEO and Founder of NVIDIA is now compared to Steve Jobs.

This is a stunning visual: NVIDIA has filled the arena where the San Jose Sharks play to capacity for its GTC event. The excitement in the audience, predominantly from other Big Tech companies, is palpable.

New GPU Platform: welcome to Blackwell

NVIDIA recently introduced the Blackwell platform, marking a significant advancement in computing for AI. This platform allows organizations to deploy real-time generative AI with trillion-parameter models at a cost and energy consumption up to 25 times less than earlier models.

Main features:
- AI Superchip, 208B transistors
- 2nd Gen Transformer Engine: FP4/FP6 Tensor Core
- 5th Generation NVLink: Scales to 576 GPUs
- RAS Engine: 100% In-System Self-Test
- Secure AI: Full Performance Encryption
- Decompression Engine: 800GB/sec

Compute improvements over the years

The DGX GB200 system encapsulates this innovation with 36 of these GB200 Superchips, combining 36 NVIDIA Grace CPUs and 72 Blackwell GPUs into a singular supercomputing force. This setup, linked through the advanced fifth-generation NVIDIA NVLink, is engineered to offer a performance boost of up to 30 times over the H100 Tensor Core GPU for processing large language models.

Elevating the computational potential further, the DGX SuperPOD powered by Grace and Blackwell includes a minimum of eight DGX GB200 systems. This scalable framework can extend to include tens of thousands of GB200 Superchips through NVIDIA Quantum InfiniBand connectivity. With the potential to link 576 Blackwell GPUs across eight DGX GB200 systems via NVLink, this configuration is designed to support expansive shared memory spaces for pioneering AI model development.

NVIDIA Introduces Generative AI Microservices for Easy Deployment

NVIDIA announced the NVIDIA Inference Microservice, NIM, which offers cloud-native microservices designed to expedite the development and deployment of AI applications across diverse environments.

It offers the following:

  • Prebuilt containers and Helm charts for quick setup and deployment.

  • Industry-standard APIs for seamless integration and ease of use.

  • Domain-specific code and optimized inference engines, including Triton Inference Serverโ„ข and TensorRTโ„ข-LLM, for enhanced performance.

  • Support for custom models on the NVIDIA AI Enterprise runtime, enabling a broad range of development and deployment capabilities.

How we will create software in the future

NVIDIA introduced the concept of using Agents for future software development, moving away from traditional coding or cloning from GitHub. Instead, developers will orchestrate a team of specialized AIs, led by a SUPER-AI that devises and delegates tasks based on the project's requirements. This includes AIs with expertise in specific domains like SAP's ABAP or data manipulation with Pandas, each contributing to the project based on their specialization.

The process entails these specialized AIs collaborating to execute parts of a broader plan, with each AI focusing on its area of expertiseโ€”ranging from software development frameworks to data analysis tools. This collaborative effort culminates in a comprehensive solution, pieced together from the individual contributions of each AI.

This approach revolutionizes software development, enabling customized, complex systems with unmatched speed and efficiency. By leveraging the unique skills of various mini-AIs, developers can assemble software in unimaginable ways, fundamentally altering the landscape of software development.

The humanoid robots are here to stay

NVIDIA announced Project GR00T, a Foundation Model that enables these robots to learn and adapt through observation and practice, mirroring human learning processes. The project aims to revolutionize robotics by making them more adaptable and efficient in various tasks.

Some key highlights:

1. ๐—Ÿ๐—ฒ๐—ฎ๐—ฟ๐—ป๐—ถ๐—ป๐—ด ๐—–๐—ฎ๐—ฝ๐—ฎ๐—ฏ๐—ถ๐—น๐—ถ๐˜๐—ถ๐—ฒ๐˜€ ๐—ผ๐—ณ ๐—š๐—ฅ๐Ÿฌ๐Ÿฌ๐—ง

โ†ณGR00T robots can learn complex tasks by observing humans, enhancing their ability to integrate into environments such as factories.

โ†ณThrough trial and error in specialized environments, GR00T robots refine their skills, making better decisions over time.

๐Ÿฎ. ๐—œ๐—บ๐—ฝ๐—ฎ๐—ฐ๐˜ ๐—ฎ๐—ป๐—ฑ ๐—”๐—ฝ๐—ฝ๐—น๐—ถ๐—ฐ๐—ฎ๐˜๐—ถ๐—ผ๐—ป๐˜€ ๐—ผ๐—ณ ๐—š๐—ฅ๐Ÿฌ๐Ÿฌ๐—ง

โ†ณGR00T could transform factories by enabling robots to quickly adapt to new tasks, making production more flexible.

โ†ณRobots equipped with GR00T could offer personalized care and companionship, particularly for the elderly and patients.

โ†ณGR00T-powered robots could navigate hazardous or inaccessible areas, such as disaster zones or extraterrestrial environments, improving safety and exploration efforts.

๐Ÿฏ. ๐—๐—ฒ๐˜๐˜€๐—ผ๐—ป ๐—ง๐—ต๐—ผ๐—ฟ ๐—ฎ๐—ป๐—ฑ ๐—˜๐—ป๐—ต๐—ฎ๐—ป๐—ฐ๐—ฒ๐—บ๐—ฒ๐—ป๐˜๐˜€ ๐˜๐—ผ ๐—œ๐˜€๐—ฎ๐—ฎ๐—ฐ ๐—ฃ๐—น๐—ฎ๐˜๐—ณ๐—ผ๐—ฟ๐—บ

โ†ณNVIDIA introduced Jetson Thor, a computing platform optimized for humanoid robots, featuring a modular architecture and high-performance capabilities.

โ†ณSignificant updates to the NVIDIA Isaac robotics platform include AI foundation models, simulation tools, and workflow infrastructure, facilitating robot development.

โ†ณThe Isaac tools suite, including Isaac Lab and OSMO, supports the development of foundation models across varied robot forms and environments.

๐Ÿฐ. ๐—–๐—ผ๐—น๐—น๐—ฎ๐—ฏ๐—ผ๐—ฟ๐—ฎ๐˜๐—ถ๐—ผ๐—ป ๐˜„๐—ถ๐˜๐—ต ๐—œ๐—ป๐—ฑ๐˜‚๐˜€๐˜๐—ฟ๐˜† ๐—Ÿ๐—ฒ๐—ฎ๐—ฑ๐—ฒ๐—ฟ๐˜€

โ†ณNVIDIA collaborates with leading companies like Agility Robotics, Boston Dynamics, and others to push the boundaries of humanoid robot technology.

โ†ณThe partnership focuses on investing in the necessary computing power, simulation tools, and machine learning environments to realize the vision of integrating robots into daily life.

โ†ณThese collaborations aim to address global challenges and drive innovation in robotics, emphasizing the importance of not working in isolation. GR00T represents a major leap towards artificial general robotics, with potential applications that extend beyond current limitations.

Our integrations with the Nvidia ecosystem at IBM

At IBM we are proud to assist clients with intricate business problems by integrating its deep knowledge of technology and industry sectors with Nvidia's advanced AI Enterprise software suite, which includes the latest NIM microservices and Omniverse technologies. This collaboration will speed up AI workflows for clients, improve the optimization process from use case to model, and foster the development of AI applications tailored to specific business and industry needs. Leveraging Isaac Sim and Omniverse, IBM is actively creating and deploying digital twin solutions for the supply chain and manufacturing sectors.

NVIDIA is driving innovation in several key sectors. In transportation, its technology is set to enhance next-generation vehicle fleets. The company is boosting healthcare through advanced imaging and speech recognition microservices, and digital biology. NVIDIA is also advancing robotics, telecommunications with a focus on 6G, and quantum computing, aiming to improve AI applications in network infrastructures and accelerate molecular simulations. These initiatives highlight NVIDIA's pivotal role in technological progress across industries.

It was an amazing GTC, and Iโ€™m very excited about the progress we are all making to move forward advanced in AI.

and thatโ€™s all for today. Enjoy the weekend folks,

Armand ๐Ÿš€

Whenever you're ready, learn AI with me:

The 15-day Generative AI course: Join my 15-day Generative AI email course, and learn with just 5 minutes a day. You'll receive concise daily lessons focused on practical business applications. It is perfect for quickly learning and applying core AI concepts. 17,000+ Business Professionals are already learning with it.

Reply

or to participate.