Amazon-backed Skild AI unveils universal robot brain for any machine

Amazon-backed Skild AI unveiled Skild Brain, a foundational AI model designed to operate on nearly any type of robot, from assembly-line machines to humanoids. The launch positions the startup to address robotics’ unique data scarcity challenge while advancing the broader push toward versatile humanoid robots capable of diverse tasks beyond single-purpose factory automation.

What you should know: Skild Brain enables robots to think, navigate, and respond more like humans through advanced spatial reasoning and adaptability.
• Demonstration videos showed Skild-powered robots climbing stairs, maintaining balance after being pushed, and picking up objects in cluttered environments.
• The model includes built-in power limits to prevent robots from applying unsafe force during human interactions.
• Unlike traditional single-purpose factory robots, this technology aims to create machines capable of handling diverse tasks across different environments.

The big picture: Robotics faces a fundamental data problem that sets it apart from other AI applications, requiring innovative training approaches.
• “Unlike language or vision, there is no data for robotics on the internet. So you cannot just go and apply these generative AI techniques,” CEO Deepak Pathak told Reuters.
• Skild trains its model on simulated episodes and human-action videos, then fine-tunes it using data from every robot running the system.
• Deployed robots feed data back into Skild Brain to continuously sharpen its skills, creating what co-founder Abhinav Gupta calls a “shared brain.”

Who else is involved: The two-year-old startup has assembled significant backing from major tech players and venture capital firms.
• Skild raised $300 million in a Series A funding round that valued the company at $1.5 billion.
• Investors include Amazon.com, SoftBank Group, Menlo Ventures, Khosla Ventures, Sequoia Capital, and Amazon founder Jeff Bezos.
• The company has hired staff from Tesla, Nvidia, and Meta, with co-founder Gupta previously founding Meta’s robotics lab in Pittsburgh.

Key details: Current clients include LG CNS, the IT solutions arm of LG Group, along with unnamed partners in logistics and other industrial applications.
• Unlike software that can scale quickly, robotics requires physical deployment, which takes time.
• However, Skild’s approach allows robots to add new capabilities across different industries more rapidly than traditional methods.
• “The shared brain allows robots to add new capabilities across different industries quickly,” said Raviraj Jain, partner at investor Lightspeed Venture Partners.