AIResearchers Comment on Robot In-Context Learning Video
Posts examine a robot demonstration of in-context learning for new tasks.
TLDR
Andrew Carr responded to Lerrel Pinto's post by suggesting 3D simulation of robot demonstrations to replace physical ones. Yoav Goldberg replied that the robot in the video squeezed too hard and dropped early, striking the side of the bowl. He added that semantic understanding appears incomplete and guessed the system had learned to pick the yellow item and drop it over the blue one.
Researchers Comment on Robot In-Context Learning Video
Posts examine a robot demonstration of in-context learning for new tasks.
