Tech media outlet The Decoder (September 18) published a blog post reporting that Nvidia senior scientist Jim Fan predicts that the robotics field will usher in a "GPT-3 moment" in the next few years.
About Jim Fan
Jim Fan received his Ph.D. from the Vision Lab at Stanford University, where he studied under the supervision of Prof. Feifei Li. His research interests include multimodal fundamental models, reinforcement learning, and computer vision, and he has interned at well-known organizations such as Google Cloud AI, OpenAI, and Baidu Silicon Valley AI Lab.
Jim Fan currently leads AI-related research at NVIDIA, and his team is working on "Project Groot," the company's efforts to create a basic model for humanoid robots.
Research breakthroughs in the next two to three years
Jim Fan predicts that in the next 2-3 years, there will be major breakthroughs in the research of basic models of robots, but he also admits that it will take longer for robots to enter daily life.
In an interview with Sequoia Capital, Fan said he expects a "GPT-3 moment" in robotics - a breakthrough in foundational robotics models that rivals GPT-3's impact in language processing.
IT Home translates its views as follows:
Bringing robots into people's daily lives is not just a matter of technology. Robots need to be affordable and mass-produced, and we need hardware security as well as privacy and regulatory safeguards.
The world is built around the human form, right? Our restaurants, factories, hospitals, and all the equipment and tools – they are all designed for the human form and the hands.
He argues that, in theory, a capable humanoid robot can perform any task that a human can do, he predicts that the ecosystem of humanoid robot hardware will be ready within two to three years.
NVIDIA's robotics-related research
Nvidia uses a combination of three data types when developing robotic AI: internet data, simulated data, and real-world robot data. Dr. Fan highlights the strengths and weaknesses of each approach and believes that their combination is the key to success.
Nvidia is working on technologies such as "Eureka," which uses language models to generate reward functions for bot-trained robots to automate processes.
In addition to the real world, Fan's team is also working on AI agents for use in virtual environments such as video games. He identified similarities between these domains and worked to develop a unified model that would control both virtual and physical agents over the long term.