Nvidia’s Rev. Lebaredian leads simulation and multifaceted efforts for robotics development. Lebaredian’s career parallels Nvidia’s evolution from graphics to AI to robotics. Nvidia’s strategy includes anticipating technology trends, building tools, and enabling innovation.
Thank you for registering!
Access your favorite topics in a personalized feed on the go. Download the app
By clicking “Sign Up”, you agree to our Terms of Service and Privacy Policy. You can opt-out at any time by visiting our settings page or by clicking “unsubscribe” at the bottom of the email.
The year Nvidia was founded, the World Wide Web went online and “Jurassic Park” was released in theaters. The 1993 blockbuster film opened a dam for computer graphics in movies, and soon Reverend Lebaredian entered the movie business. He brought the fictional gorilla Mighty Joe Young to life and helped build the technology that brought the children’s classic Stuart Little to life. Currently, he works in the field of 3D for the biggest names in artificial intelligence.
Mr. Lebaredian is Nvidia’s Vice President of Omniverse and Simulation Technologies, reporting to CEO Jensen Huang. Simply put, Reveredians are responsible for building robot brains.
Lebaredian explained that if Nvidia wants to make the technology a reality, it essentially needs to build it first, but that doesn’t mean it will sell everything it builds. For example, Nvidia’s latest LLM NVLM, released in September, is very competitive in performance benchmarks. However, it is licensed for research purposes only, and commercial use is not permitted.
Mr. Lebaredian’s career trajectory is similar to that of the company. Graphics technology eventually transformed into simulating self-driving cars and, in the past seven years or so, has also been used to simulate robots.
It turns out that these technologies haven’t changed much in the modern era of machine learning. “AI is essentially bridging this divide between computing and the physical world, and it’s essentially robotics,” Lebaredian told Business Insider.
Nvidia veterans are building virtual worlds that are true to the real world for robots to use to practice and learn tasks. Simulation and Nvidia’s Omniverse artificial environment are key to creating thinking, perceptive, safe, and productive robots. Still, the stakes are higher than in the movies.
“The important thing here is that the simulations we use need to match the real world as much as possible so that what we learn in the virtual world can be transferred to the real world. Even if you learn using physics, it doesn’t mean you’ll behave appropriately once you get into the real world,” he said. Simulations allow billions of iterations and many mistakes in a safe environment where humans, robots, and property are not harmed.
embrace ambiguity
Nvidia has a habit of seeing big technology trends arrive early, building a suite of tools, and waiting for the rest of the world to come along. Lebaredian described what has happened in large-scale language models over the past decade.
“All these computers were built by us. No one asked for them, everyone was skeptical. And we went and trained the largest model in the world at the time. ” Lebaredian said. This model is called Megatron and Nvidia launched it in 2021.
“There is a direct connection from Megatron to GPT,” he said. The release of ChatGPT spurred the current generative AI boom. The company is now trying to recreate that method using robots.
Related articles
“I’ve been working on robotics simulations with the expectation that eventually we’ll have the technology to create a proper robot brain,” Lebaredian said. Generative AI was that moment. But robotics hasn’t had its ChatGPT moment yet.
The robot equivalent of Megatron is called Project Groot, and it is a basic model of robotics that the company announced in March. It’s not even commercial, Lebaredian said. The goal, he said, is for other companies to jump in as quickly as possible and build on what Nvidia has to offer.
Bringing the physical world into generative AI raises the bar well beyond language generation, and while skepticism is to be expected, Nvidia is essentially following a tested playbook, Lebaredian said. said.
“I saw Jensen make such a far-reaching bet. There’s a lot of ambiguity as to when it’s going to happen or not, and trusting first principles that it’ll get there, kind of We’re just navigating, ‘We don’t really know where it is, but we want to be there when it happens,’ Lebaredian said. “So we invest as much as we can without dying during that period until we are ready.”
Have a tip or insight to share? Contact BI Senior Reporter Emma Cosgrove at ecosgrove@businessinsider.com or use the secure messaging app Signal at 443-333-9088.
