Covariant, a robotics company founded by former OpenAI researchers, is revolutionizing the way robots interact with the physical world. Using the same technology development methods behind chatbots, Covariant is creating A.I. technology that allows robots to pick up, move, and sort items in warehouses and distribution centers.
By analyzing vast amounts of digital data, Covariant’s technology gives robots a broad understanding of the English language, enabling them to communicate with humans much like a chatbot. The company, backed by $222 million in funding, focuses on building the software that powers robots rather than the robots themselves.
The A.I. systems developed by Covariant learn from a combination of text, images, and sensory data, allowing robots to handle unexpected situations and respond to plain English commands. This technology, known as R.F.M. (robotics foundational model), has the potential to transform industries beyond warehouses, providing a roadmap for the deployment of robots in manufacturing plants and even on roadways with driverless cars.
While the technology is not perfect and still makes mistakes, researchers believe that as companies continue to train these systems on larger and more varied datasets, robots will rapidly improve in their ability to adapt to unpredictable scenarios. This shift from traditional robotic programming to learning from digital data signifies a new era where robots, like chatbots and image generators, become more agile and responsive in the physical world.
