At the recent Automate 2024 trade show, Alphabet spinout Intrinsic announced a partnership with Nvidia aimed at leveraging generative AI to improve robotic programming tasks.
During the announcement, the two companies revealed their integration of Nvidia’s robotics solutions into Intrinsic’s platform to enhance the capabilities of a robotic arm.
The robot arm underwent training solely on synthetic data and demonstrated proficiency in various grasping tasks.
Intrinsic CEO Wendy Tan White explained that synthetic data generated from foundation models could streamline future robot deployments, making units more adaptable to diverse tasks while reducing development costs.
Nvidia’s Vice President for Robotics and Edge Computing, Deepu Talla, emphasized that with the latest AI foundation models, companies can program robot configurations capable of generalizing and interacting with various objects in real-world environments.
Launched in 2021 under Alphabet’s X moonshot program, Intrinsic focuses on developing advanced robotics solutions for industrial applications, particularly in manufacturing.
Intrinsic has collaborated with Siemens to broaden access to robotics and is partnering with Google DeepMind to accelerate robotics programming.
At Automate 2024, Intrinsic and DeepMind jointly created an AI-based robot motion planner, enabling engineers to program multiple robots simultaneously.
Using synthetic data, Intrinsic and DeepMind instructed four robots to weld a car, resulting in improved performance compared to traditional programming methods.
DeepMind and Intrinsic collaborated to enhance robot dexterity by combining data and model knowledge to train robots on human input using remote devices, thereby translating human skills to robotic tasks.
