General Robotics takes a new approach to building a robotic brain

General Robotics is pioneering a novel approach to constructing robotic brains, moving away from the prevalent practice of embedding intelligence directly into robots. The Redmond-based company, General Robotics, operates the GRID system, which hosts its robotic intelligence layer in the cloud, allowing for seamless delivery and deployment to remote robots.

GRID manages various elements such as models, simulations, robot representations, data, and agents, facilitating the development, testing, orchestration, and deployment of robotic skills to on-site robots. According to General Robotics CEO Ashish Kapoor, "Having one layer that looks after 50 to 100 robots is way more important than 50 or 100… robots that operate completely independently."

Kapoor emphasized that General Robotics is primarily focused on industrial models rather than the viral videos of sprinting or humanoid robots. The company's GPU-based GRID cloud provides a robust foundation for ingesting data and developing skills tailored to specific industries. Kapoor illustrated the functionality of GRID using a biotechnology lab scenario, where robots needed to pour liquids from beakers while interacting with transparent materials.

Traditional software modeling for such behaviors is lacking, but AI agents integrated through GRID create a "hybrid simulator" that unifies the pieces in the cloud. The robots are subsequently tested against this simulated model, streamlining what would typically take years into mere minutes.

While GRID offers numerous advantages, General Robotics faces challenges such as latency in cloud-to-robot communication. The round-trip time between the cloud and the robot is critical, necessitating a second loop—an inner control loop that operates at a millisecond level on the robot's edge. Additionally, companies with stringent regulatory and security requirements, such as automotive manufacturers, may be hesitant to store proprietary information in the cloud.

However, General Robotics remains optimistic about its system's adaptability, noting their ability to deploy across commercial clouds, on-premises servers, and even edge deployments. The company emphasizes ongoing improvements to its technology stack and actively seeks partnerships to address these challenges. As the robotics industry continues to expand, Morgan Stanley predicts that the global humanoid robot population could reach 1 billion by 2050, valued at $5 trillion, with 50,000 to 85,000 humanoid robots expected to be shipped this year alone.