Summary of What’s Next in AI: NVIDIA’s Jensen Huang Talks With WIRED’s Lauren Goode
Summary of Jensen Huang's Discussion at Siggraph
In the conversation at Siggraph, NVIDIA's CEO Jensen Huang discusses the evolution of NVIDIA and its significant role in both computer graphics and generative AI. He highlights the historical milestones of NVIDIA, from its origins in computer graphics to its pioneering work in accelerated computing and deep learning. Key points include:
- Generative AI and Computer Graphics: Huang emphasizes that Siggraph has transitioned from a focus solely on computer graphics to incorporating generative AI and simulation. He illustrates NVIDIA's journey with a cartoon that highlights milestones like the invention of programmable shading and the development of the first programmable shading GPU.
- Deep Learning Breakthroughs: The introduction of AlexNet in 2012 marked NVIDIA's first major engagement with AI, leading to a pivot in the company's focus towards deep learning. This shift enabled NVIDIA to develop the DGX-1, the first computer designed specifically for deep learning.
- RTX Technology: Huang discusses the introduction of RTX in 2018, which revolutionized real-time ray tracing in graphics, and how it enabled rendering at much higher speeds through techniques like Deep Learning Super Sampling (DLSS).
- Generative AI's Impact: He notes the rapid advancements in generative AI since the launch of ChatGPT, highlighting its ability to learn from various modalities (text, images, 3D, etc.) and its implications across industries, including robotics and self-driving cars.
- Future of AI Development: Huang introduces the concept of Software 3.0, where generative AI will facilitate the development of software by using pre-trained models and collaborative AI systems. He expresses optimism that future generative AI tools will become more controllable and accurate, addressing current issues like hallucinations in AI outputs.
- Energy Efficiency: The discussion touches on the energy demands of generative AI, with Huang asserting that NVIDIA's advancements in accelerated computing are essential for managing energy consumption while meeting computational demands.
- Omniverse Platform: Huang explains how NVIDIA's Omniverse allows for collaboration in 3D content creation, integrating various data modalities and enabling users to generate controlled images and environments.
- Physical AI: He introduces the concept of Physical AI, which involves developing robots that can interact with the physical world, requiring sophisticated simulation and training methods.
- Open USD: Huang emphasizes the importance of open standards like Open USD for ensuring that digital creations can be preserved and shared across different platforms and tools.
Speakers
- Jensen Huang (NVIDIA CEO)
- Lauren Goode (WIRED)
Notable Quotes
— 14:27 — « The big breakthrough of ChatGPT was reinforcement learning human feedback, which was the way of using humans to produce the right answers or the best answers to align the AI on our core values. »
— 23:37 — « Well, that's what tools do. We invent tools here. This conference is about inventing technology that ultimately ends up being a tool. »
— 24:41 — « Imagine the stories that you're going to be able to tell, with these tools. »
— 49:04 — « AI doesn't care where it goes to school. »
— 58:04 — « In a lot of ways, open USD is the HTML of virtual worlds. »
Category
Technology