Thanks Cade Metz and the The New York Times for covering RFM-1 – Covariant's big step towards the ChatGPT moment of robotics.
At a product level, RFM-1’s language understanding provides a chatbot-like interface to easily communicate with robots in natural language; at a more fundamental level, RFM-1, powered by a large transformer and pre-trained on large amounts of multi-modal data, is similar to LLMs in its scalability and generality.
Additionally, RFM-1 can do a lot of what existing LLMs can’t do: by generating and understanding videos, it has a deeper understanding of the physical world; by understanding tens of millions of robot interactions with the world, it knows how to take the most robust actions.
We are excited to roll out RFM-1 to production robots soon and accelerate the journey of building robotics foundation models.