What is the best data for training humanoid & robotics foundation models?
@michellelsun12K followers
15,483 views
6 Comments
Comments from the original post on X (Twitter)

@PrismaXai
@peteflorence @generalist @tbpn Real-world robot experience at S-tier is exactly right. The bottleneck was never the model. It was always the data layer.
8 likes

@Manudata_ai
@peteflorence @generalist @tbpn Thats exactly what we are solving.
1 like

@cheuk_baby
@peteflorence @generalist @tbpn 有意思,B级视频挺強的。
1 like

@jmarbach
@peteflorence @generalist @tbpn Getting data off the sensors is the first step to training with it. For this there’s @hubble_network

@LeoKharon
@peteflorence @generalist @tbpn What about @PrismaXai 's recently released standard for data? It argues that most exisitng robotics dataset do not meet the quality standards required in order to be useful

@nurvai_ai
@peteflorence @generalist @tbpn The quality point is key. Scaling physical AI will require more than collecting massive datasets. The real challenge is capturing the interactions and edge cases that create robust robot behavior.
