Skild AI raises $300M to build an off the shelf brain for robots

AI startup Skild AI announced it has raised $300M in a round led by Lightspeed Venture Partners, Coatue, SoftBank Group, and Jeff Bezos.

The company, which has been operating in stealth mode, is building what it calls a General Purpose Brain for various robots performing both simple and complex tasks in different environments such as industrial settings and households.

This approach contrasts with the current, more common method where each robot is built for a specific purpose and given intelligence specifically designed to perform assigned tasks.

With Skild’s model, individuals and companies building robots would not need to create a new model from scratch. This will help speed up time to market, foster innovation in use cases, and reduce costs as they can use the model as the brain for their robots.

Skild trained its model using various data types, including text, images, and videos. It built this dataset using mixed data collection techniques, such as operating robots manually, collecting data from these actions, and having the robots use the model to carry out tasks, collect data, and learn by trial and error.

Perhaps the most unique learning method Skild used is what its cofounder Deepak Pathak calls "Curiosity-driven learning," where they instill artificial curiosity into robots by rewarding the system for producing outcomes that occur when it cannot predict the results of its actions. The more uncertain the system is about predicting the effects of its actions, the more curious it gets to explore.

The combination of these training approaches incentivizes the AI to navigate more scenarios and collect more data.

Skild’s competitors include heavyweights such as OpenAI, Figure AI, and Covariant (an OpenAI spin-off).