Varun Nair
@_varunnair
Scaling Physical AI | Prev @Coinbase @UCBerkeley EECS
All views are my own.
596 Following 1.9K Followers
“If every change requires collecting a new dataset and running another round of post-training, you’re signing up to repeat that work for as long as the robot is deployed. This is not scalable.”
Adaptation Cost has always been the right metric for generalization. You must minimize it through diversity.
Show more
This month both @DynaRobotics and @perceptroninc have independently proven
Scaling diverse pretraining reduces the need for embodiment specific data closer to the robot action space.
