Covariant's Vision for RFM1 and AI in Robotics

RFM1, the robotic foundation model developed by Covariant, aims to extend AI advancements into the physical realm by combining general internet data with data rich in real-world interactions. This unique approach allows RFM1 to bridge the gap between software-only models and robots that operate in real-world scenarios. Covariant's utilization of embodied AI, starting from 2017, has enabled RFM1 to become an eight billion parameter transformer model trained on various modalities like text, images, videos, robot actions, and sensor readings, paving the way for robots to interact autonomously and effectively in diverse real-world settings.