The American technology company Nvidia, a developer of the most popular chips for AI, said that in the first half of 2025 it will release the latest generation of compact computers, Jetson Thor, for humanoid robots – a comprehensive solution with a number of components from individual chips to full-fledged training and control software for robots.
Nvidia will release comprehensive software for controlling humanoid robots in 2025
Author: Ekaterina Alipova
According to Nvidia Vice President of Robotics Deep Talla, a revolution in the field of artificial intelligence and the advent of “physical AI” is just around the corner, which is what he is talking about reported Financial Times. And at this “turning point” for the market, the company wants to be among the leaders.
As digital giants like Amazon, Microsoft and Google strive to reduce their dependence on Nvidia (by opting for AMD, for example), the leader in chip production is looking for new directions for development. And robotics is one of the key ones, although it has not yet been separated into a separate division within the corporation.
Nvidia, whose value has exceeded $3 trillion, positions itself as an investor in the field of “physical AI”: in February, along with Microsoft and OpenAI, it invested in a startup for the production of humanoid robots Figure AI, valuing it at $2.6 billion.
Talla believes that the biggest obstacle to the development of humanoid AI-powered robots has already been passed: thanks to large language models, robots can be trained using simulated environments, which solves the problem of “the gap between simulation and reality.”
“Over the last 12 months… (this gap) has been closed enough that we can do experiments with simulation combined with generative artificial intelligence, which we couldn’t do two years ago,” says Talla. “We provide a platform that allows all of these companies to do any of these things.”
Nvidia offers tools for three phases of robotics development: basic model training software that comes with Nvidia’s “DGX” system; modeling of real conditions on the Omniverse platform; as well as hardware for the “brain” of the robot.
According to US research agency BCC, the global robotics market is currently valued at approximately $78 billion and is projected to reach $165 billion by the end of 2029.
Amazon has already deployed Nvidia’s robotics simulation technology in three of its US warehouses, and Toyota and Boston Dynamics are among other customers using Nvidia’s training software.
Cover photo: Unsplash
Source: rb.ru