FoodaiThe startup of robotics from California from California, collected $ 405 million in many previously undisclosed rounds to develop “basic personified AI models”-the brain of robots designed to assist help every part, from humanoids to 4 times to self-propelled cars, adapt to latest environments.
The company announced financing on Wednesday; The last round collected $ 314 million in August and was jointly run by Bezos Expeditions, Prism and Temasek. Other supporters of Fieldi include Khosla Ventures, Intel Capital and Canan Partners.
Unlike traditional artificial intelligence, which processes text or images, embodied AI refers to AI, which controls physical works moving through real environments. Fieldai builds “field foundations models”, which are personified general models of AI rooted in physics. This approach gives robots the opportunity to quickly learn and adapt to latest environments, while being aware of the risk, the founder and general director of Fieldi Ali Agha said in an interview with TechCrunch.
“The mission is to build one robot’s brain, which can generalize in various types of robots and a variety of a set of environments,” Agha said. “To get there, you need to manage risk and safety when you go to these new environments. It was a basic gap in robotics that traditional models and traditional approaches have never been designed to manage this risk and security.”
Agha said that the key to obtaining robots for secure learning in latest environments is so as to add a layer of physics to those AI models. This add -on gives robots a second set of information from which decisions might be made – especially in the latest environment – in contrast to the response to every part that the model says to do traditional LLM.
He added that although a small amount of hallucinations AI is not harmful in some circumstances, it could be for robots working in dangerous environments or with people.
“Suddenly you start to have this feeling, as I know, and if I don’t know something, or if I make a decision how I’m sure,” said Agha. “Once [the] The network begins to access this, it begins to make many safer decisions. Not only does it spit out: “Hey, here is the next type of motion”, but he tells you how certain it is, and you as a customer you can define this risk threshold, and [the] The robot will be reactive to this. “
TechCrunch event
San Francisco
|.
October 27-29 2025
Agha has been working on this concept in various roles in places for many years, from NASA to Massachusetts Institute of Technology (myth). He decided to begin Fieldai when he achieved a technological breakthrough, which allowed one brain to work in differing types of robots to perform each the same and individual activities.
From the moment the company was launched in 2023, Fieldai provided contracts in industries, including construction, energy and urban delivery. The company refused to reveal customers by name.
Financing will support research and development, while helping the company increase production in order to implement its models for its clients and further extend the range abroad.
Agha compares Fieldai’s approach to human evolution. “You evolve to be able to perform various tasks in different environments and you can learn quickly, [and] We think it is a need for robotics. Yes, you can definitely optimize for one specific case of use, but this is not the market we are going. “