In 1993, the same year Nvidia was founded, the World Wide Web went live, and the groundbreaking film Jurassic Park hit theaters. The movie, with its pioneering use of computer-generated imagery (CGI), opened the door for CGI in films, leading Rev Lebaredian into the movie industry. He played a crucial role in developing the technology that brought the giant gorilla Mighty Joe Young and the animated mice in Stuart Little to life. Fast forward to today, and Lebaredian is now shaping the future of robotics at one of the most influential companies in artificial intelligence (AI).
As Vice President of Omniverse and Simulation Technology at Nvidia, Lebaredian is in charge of building the “brains” behind robots. His work revolves around creating the virtual environments where robots can learn and simulate tasks. For Nvidia, this means building advanced technology for AI and robotics, though not all of it is meant for commercial use. For instance, Nvidia’s latest large language model, NVLM, which was released in September, offers competitive performance but is licensed only for research purposes—not for direct sale or use.
Lebaredian’s career trajectory mirrors Nvidia’s evolution. What started with graphics technology has expanded to simulations for autonomous vehicles and, for the past seven years, robots. Today, Nvidia’s deep involvement in AI is bridging the gap between computing and the physical world—essentially advancing robotics.
“The key here is that the simulation we use must match the real world as closely as possible,” Lebaredian explains. “If a robot learns in a virtual world with cartoon physics, it won’t behave properly in the real world.” This is why simulations are vital: they allow robots to make mistakes and learn through billions of repetitions in a controlled, risk-free environment, without any harm to people, property, or robots themselves.
Nvidia has a track record of spotting significant tech trends early and building tools ahead of demand. Lebaredian recalls how the company developed powerful computing systems and models like Megatron, even when few people were asking for them. In 2021, Nvidia unveiled Megatron, which became a precursor to GPT (Generative Pre-trained Transformer), the architecture behind ChatGPT. The boom in generative AI followed, and now Nvidia is working to replicate that success with robotics.
Lebaredian has been preparing for this shift by developing simulations for robotics in anticipation of a major breakthrough. “I’ve been working on a simulation for robotics, knowing that the technology to create proper robot brains was coming,” he says. That moment arrived with generative AI, but robotics has yet to experience its own “ChatGPT moment.”
Enter Project Groot, Nvidia’s answer to the robotic version of Megatron. Announced in March, Project Groot is a foundational model for robotics that aims to help build advanced, capable robot brains. However, like other Nvidia technologies, it’s not intended for commercial use. Instead, the goal is to create a platform that others can build upon, spurring innovation across the field.
Bringing AI into the physical world may raise the stakes beyond language models, but Lebaredian is confident that Nvidia’s strategy—tested time and again—will lead the way. “I’ve seen Jensen take these kinds of risks before,” he says, referring to Nvidia CEO Jensen Huang. “There’s always ambiguity around when things will happen, but we invest based on first principles, believing that eventually, we’ll be there when the technology matures.” It’s a bet on the future, and one that Nvidia is making with the patience to ensure its place when the moment finally arrives.
Leave feedback about this