“The Gato model can process inputs including images, text, video, actions, and proprioceptive sensor data from robotics tasks.”