During the CadenceLIVE Silicon Valley 2024 conference held on April 17 at the Santa Clara Convention Center in Silicon Valley, NVIDIA CEO Jen-Hsun Huang had an in-depth conversation with Cadence CEO Anirudh Devgan. Cadence, formed in 1988 through the merger of two companies, SDASystems and ECAD, is the world's largest provider of electronic design automation, semiconductor technology solutions, and design services.
This cutting-edge conversation covered several topics, including issues such as accelerated computing, the future of artificial intelligence (AI), and energy consumption. According to Jen-Hsun Huang, AI will bring revolutionary impact in three areas: data centers, robotics/autonomous driving, and life sciences, and humanoid robots may be as low as $10,000~$20,000 in the future.
At the same time, for the current widespread concern about AI energy consumption, he said that although AI will consume a lot of computing power, AI will revolutionize the way we deal with climate change, help use less energy, improve energy efficiency, etc.
Generative AI is difficult to achieve without accelerated computing
In the conversation, Jen-Hsun Huang emphasized the importance of accelerated computing for AI development. He cited the many benefits accelerated computing has brought to Millennium, Cadence's digital twin platform, as an example of how generative AI is likely to become a reality once accelerated computing is adopted. Without the transition to accelerated computing, generative AI would be difficult to realize.
Accelerated computing is not the same as general-purpose computing, he said. In general-purpose computing, you can create a processor that will run all the code, which is not the case with accelerated computing. According to him, accelerated computing is capable of delivering a 1000x X-factor, and on top of that, there is another 30x X-factor. And if you add generative AI, there is another 100,000x factor on top of that. He mentioned that design tools usually process once, but designers need to explore multiple times to find the best solution in multi-dimensional and multi-modal situations. And AI will help us delve into specific areas for exploration and optimization.
According to Jen-Hsun Huang, he has found that a small portion of the code in a program represents the majority of the tool's runtime. CFD (computational fluid dynamics), for example, may only use 3% of the code, representing 99.9% of the runtime, while the remaining 97% of the code can be rewritten with AI and accelerated computation, speeding up the application by a factor of 100,000.
As the coiner of the term "accelerated computing," he said, "If we don't move to accelerated computing, if we don't move to AI, the computer industry could experience an anti-Moore's Law, and the reason for that is very clear: the amount of work and the amount of computation that we're doing is growing, but CPU scaling has slowed down. and as a result, our computing costs will grow, not decrease."
Humanoid robots may be as low as $10,000-$20,000 in the future
When asked what industries NVIDIA is involved in that would be very exciting for it in the short or medium term, Jen-Hsun Huang expressed great interest in three industries: data centers/computing, robotics/automated systems, and life sciences.
Speaking about the area of robotics/automated systems, he said that whether it's cars or trucks, pizza delivery robots, or humanoid articulated self-connecting robots, these types of systems have a lot in common in that they need to have a lot of sensors, and more importantly, they need to be functionally safe. It is very important to design computers and validate them in a way that requires operating systems that are not ordinary types of operating systems.
He believes that the use of AI is so widespread that these systems will be readily connected to the cloud, to data centers, so that they can update the experience, report faults and new situations, and then download new models. "Suffice it to say, I love the whole automated systems space, it's a whole new category."
Jen-Hsun Huang mentioned in the conversation that humanoid robots could be much cheaper to build than people expect. "You can get a $10,000-$20,000 car, why can't you have a $10,000-$20,000 humanoid robot? Robots may well be able to perform in an environment designed for humans and be more flexible and versatile than humans."
When it comes to the topic of life sciences, he said that he wants to turn biology into an engineering field and that the scientific discovery process is very important, but it is sporadic.
In any case, he believes that digital biology is going to undergo a full-blown renaissance, with science and engineering getting closer and closer, and that it's a very complex field. "We don't talk about Schrödinger's equation in chip design because we change the transistor until we can avoid Schrödinger's equation. And in biology the Schrödinger equation is necessary. So we have a lot to innovate, and for the first time we have the necessary tools, the computational systems, and algorithms to help us deal with very large and very messy systems, and the convergence of data-driven approaches with the principled simulation approach that you were talking about before may give us an opportunity."
He emphasized in the conversation, "I think the market size for all three of these industries is going to be very large, and the market size for humanoid robots alone is large enough."
"Energy black hole"?AI will revolutionize the way people deal with climate change
Regarding the energy consumption of AI, Jen-Hsun Huang admitted that "the energy consumption of accelerated computing is very high due to the very large number of integrated computers."
However, he also said that any optimization of power utilization translates directly into higher performance, which is measurable because higher efficiency generates more revenue or translates directly into cost savings from buying something smaller with the same performance.
"AI can help people save energy." As an example, he said that a single investment in model training would benefit millions of engineers like him, and billions of people would enjoy the savings over the next few decades. The cost and energy savings should be considered vertically across the span. He believes that vertically, AI will revolutionize the way people respond to climate change.
Jen-Hsun Huang emphasized that by investing in accelerated computing, AI, and data centers, humans can design better, more energy-efficient products. "You design a chip once, but can ship it trillions of times faster; you build a data center that saves 6% of power, and the power saved can be used by a billion people for a whole day. So the energy we save the world by designing better software, chips, and systems will have a permanent benefit to society. On the one hand, AI consumes more power and data centers; on the other hand, through better product design, better computers, better cars, better phones, better materials, and so on, we will reduce the other 98% of power and energy consumption."
