Humanoid Robots, AI, And Simulation Technology: Dinis Guarda Interviews Rev Lebaredian, Vice President At NVIDIA Omniverse

In the latest episode of the Dinis Guarda Podcast, Rev Lebaredian, Vice President of Omniverse and Simulation Technology at NVIDIA discusses the impact of NVIDIA Omniverse, the role of AI, stimulation in the Physical World, humanoid robots and OpenUSD in shaping the future.  The podcast is powered by Businessabc.net, Citiesabc.com, Wisdomia.ai, and Sportsabc.org.

Rev Lebaredian is a technology expert, entrepreneur, and business leader who is the Vice President of Omniverse and simulation technology at NVIDIA. He leads the development of NVIDIA Omniverse, a platform that combines rendering, physics simulation, and artificial intelligence to create realistic virtual worlds. He has also worked with Warner Brothers and Disney, started his own software business, and supports creative learning as an advisor to the Tumo Centre for Creative Technologies.

During the interview with Dinis Guarda, Rev Lebaredian shared the impact of NVIDIA’s work and said:

“We introduced CG, the progenitor of CUDA, by making our hardware programmable for computer graphics, and this opened up the possibility of programming it for other types of simulation. Very quickly, people in the scientific community realised they had supercomputers available to them in the form of gaming PCs. Instead of them contorting their algorithms to fit our computer graphics pipeline, we gave them a proper platform to take advantage of all this computing.

The way we deal with this complexity is by simulating how the system will be used while we’re developing it. At NVIDIA, we’ve been doing this with our chips from the very beginning. Before sending a design to be fabricated, we simulate its functionality using supercomputers. By the time it’s fabricated, we are confident it will work because the simulation has already proven it.

We understood we were building something special. Either we were going to completely change the world and become the most important company in the world, or we were going to die. Those of us who stayed bet that we weren’t going to die—we were going to change the world. And that’s what happened."

NVIDIA Omniverse: Transforming AI and simulation

NVIDIA Omniverse is a platform designed to improve 3D workflows by enabling real-time collaboration and simulation. It is built on Universal Scene Description (OpenUSD) and uses NVIDIA's RTX rendering and AI technologies to connect different 3D tools. The platform supports advanced simulations for creating virtual environments and digital twins, which are useful in industries like film, robotics, and autonomous vehicles. 

NVIDIA Omniverse also includes AI-driven features, such as tools for generating realistic animations. A recent update introduced 3D workflows, making the platform even more powerful for industries.

During the interview with Dinis Guarda, Rev Lebaredian explains the origin of NVIDIA Omniverse and its fundamental technologies: physical AI and simulation:

"Almost a decade ago, we decided to start building all the simulation technologies necessary to do this, and the computers for it, because we knew it was coming. When the AI explosion happened, the next wave would be physical AI—essentially robotics. To enable physical AI, we needed simulators and simulation computers that could represent the real world inside a computer. That’s where Omniverse came in.

Omniverse is loosely like an operating system that consists of a bunch of functionality to assemble virtual worlds and simulate them. By itself, it’s not actually a simulator or application—it’s more like an operating system with functionality others can build on top of or incorporate into their applications and technologies. We work with almost all tools providers and simulation developers in the ecosystem to make this happen."

 Rev Lebaredian also discusses why Omniverse is essential for physical AI:

"We realised the only way to generate the data—these life experiences for physical AIs—is by creating simulators. By turning the physical world into a software problem, we can generate unlimited data safely, with diversity, and at speeds impossible in the real world. The only limitation is the amount of computing we have available."

Talking about Omniverse’s role in the next Industrial Revolution Rev said :

“We believe the next industrial revolution is on its way. It’s about having autonomous agents—robots and robotic systems—operating in the physical world, alongside us, doing things humans possibly can’t. These systems will transform factories, warehouses, supply chains, transportation, cities, and buildings. To build that future, we need two fundamental technologies: AI and simulation.”

Humanoid Robots: The next era of AI and robotics

As the interview continues, Rev Lebaredian explains that robots are more than humanoid figures from science fiction. He said: 

"The term robot applies to a much more general set of things than the humanoid forms we often imagine. A robot, or a robotic system, is essentially an autonomous agent that operates in the physical world. It perceives the world through sensors, makes decisions and plans, and then acts to modify the world. It does this in a continuous loop: perceiving, reasoning, and acting.

Robotic systems are increasing everywhere now. For example, in our NVIDIA headquarters, the turnstiles at the reception area perceive you, decide if you have access, and then act to let you through. These systems are becoming more common, but we haven’t seen robots deployed at scale for several reasons.

The spaces we’ve built—retail stores, factories, homes, kitchens—are all designed for humans. They’re configured for beings that are roughly our size, with two arms and two legs, able to navigate stairs and ramps. The most useful general-purpose robots are likely humanoids because they can operate in these human-designed spaces and do useful things.

Since ChatGPT, there’s been a massive acceleration in R&D investment in humanoid robots. Companies like Tesla and many others are pursuing this because now we have the technology to make a general-purpose robot brain. Humanoids are the logical form for general-purpose robots because they fit seamlessly into human environments."

Rev also explains the historical challenges of building robots:

"The reason we haven’t had robots at scale is that building a robust robot brain has been too hard. Programming them to handle unexpected situations in the real world took so much effort that it wasn’t worthwhile. That’s why robots have largely been limited to automotive factories, where the conditions are controlled and the capital costs can be justified.

Now we finally have the ability to create good robot brains. The ‘ChatGPT moment’ unlocked the imagination of what’s possible. We now understand that we can create a general-purpose intelligence capable of reasoning about many things, and this core technology is the foundation for a useful robot brain."

Then he talked about the future of robots in human spaces and said:

"Humanoid robots will be the first to be deployed at scale. It’s going to take time to reconfigure spaces optimised for specific robots. Eventually, we’ll see lights-out factories and warehouses with robots built specifically for those environments. But for now, most spaces are designed for humans, making humanoid robots the best fit for these existing spaces.

We’ve probably had the technology to build the mechanical and physical parts of humanoid robots for a while, but without the brain to unlock their capabilities, there was no point. Now that we have the capability to build a brain, we’re seeing an acceleration that’s going to change everything."

Rev then talked about superhuman-specific intelligence and its potential and said:

"What’s true is that for a long time now, we’ve had specific intelligence—not general intelligence—but specific intelligence that’s superhuman. When we first solved the image classification problem with AlexNet, it wasn’t quite superhuman. But within a year or so, image classifiers could distinguish subtle differences between breeds of dogs better than any human could. We’ve reached superintelligence in many specific areas, like transcription, language translation, and more.

This is what excites me—the fact that we have forms of intelligence that are specific but superhuman. As an engineer, it excites me because I start asking, what can we do with these things to make human life better?

A lot of the stuff that was just a dream or science for a while has now become practical engineering. We’re focused on the pragmatic—what can we do with these amazing new superpowers we’ve created? Let’s go put them to good use."

NVIDIA shaping the future: AI, Simulation, and OpenUSD

Rev Lebaredian shares NVIDIA’s vision for the future and how combining AI, simulation, and computing can bring the digital and physical worlds closer together:

"Our strategy is to take it one step at a time. We know it’s going to happen. We’re not sure exactly what the shape and form of it will be in 20 years, but we’re pointing in that direction and building all the technologies necessary to get to the next step."

In order to truly unlock the value of computing and benefit from it as a society, we have to figure out how to apply all of this computing to the physical world. Right now, we’re putting 100 trillion dollars into the physical world versus 5 trillion into the information world. That tells you what humans find most valuable."

Talking about NVIDIA bridging the physical and digital worlds through simulation Rev said:

“What we are focusing on with simulation is essentially making a bridge. We’re turning the physical world into a computing problem, so we can solve the challenges of the physical world with the speed and efficiency of software."

"We were willing to invest in this deeply for a long time on this conviction. Most other companies won’t or can’t do that kind of investment.

When machine learning, deep learning, this new kind of AI thing, appeared on our platform, we understood what it meant. On first principles, we went all in. We asked, ‘What are the things we need to build to bring about the future? How do we capitalise on this?’ The first thing was, that we needed to take these computers we were building and make them really big—supercomputers at the scale of data centers. Nobody was asking for it at the time, but we believed it was necessary.

We also believed that with this new form of software development—machine learning and AI—there were algorithms we could now develop that were previously out of reach. The most valuable ones would be the ones that interact with our physical world—algorithms that perceive the world around us and manipulate it. Essentially, robotics algorithms. The promise has always been there, but it’s just been out of reach.

If we’re going to create AIs that understand and manipulate the world around us, those AIs are going to have to learn from some examples. Where are those examples going to come from? It’s hard gathering them from the real world, so we should turn the real world into software—into a computing problem. Once we have it inside a computer, that data can feed our AI systems, and we can eventually solve that."

Rev discusses how NVIDIA is revolutionising physical simulation through OpenUSD:

"To simulate a factory, for example, you need a single data model that includes all its components—buildings, conveyor systems, robots, sensors, humans, and their properties. This has been a challenge because these data sources have historically been fragmented across different tools and formats. OpenUSD solves this by providing a universal way to represent all these layers of complexity in one model.

Last year, we formed the Alliance for OpenUSD, with founding members including Pixar, Disney, Apple, Adobe, Autodesk, and NVIDIA. Since then, over 30 companies have joined, including industrial leaders like Siemens, Trimble, IKEA, and Lowe’s. Together, we’re standardising how the physical world is represented digitally.

By combining OpenUSD with AI technologies and NVIDIA’s advanced computing capabilities, we now have the ability to perform rich, complete simulations of the physical world fast enough to be practical. This is necessary for building all the complex physical systems we need for the world around us."

Rev also shared the importance of simulation for the future and said:

“We finally have the basic capabilities to simulate the physical world effectively. This is crucial for designing, building, and operating everything from factories and warehouses to cities and transportation systems in a safer, more efficient, and sustainable way.”