This has been exactly my feeling about what video models will do. Fundamentally there is a lot of missing information still. Even humans -- our brains and bodies honed by millions of years of evolutionary data for manipulation -- cant like one shot new objects without touch
This has been exactly my feeling about what video models will do. Fundamentally there is a lot of missing information still. Even humans -- our brains and bodies honed by millions of years of evolutionary data for manipulation -- cant like one shot new objects without touch
Source: Chris Paxton — X Robotics · Read original article ↗
Article text · Original source · English
This has been exactly my feeling about what video models will do. Fundamentally there is a lot of missing information still. Even humans -- our brains and bodies honed by millions of years of evolutionary data for manipulation -- cant like one shot new objects without touch
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported duration
30,000 hours
Reported duration
300–30,000 hours
View original evidence
Going from 300 to 30,000 hours did not narrow the gap, and a 4× larger model barely did either.
Open source S5
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
【引用 @JiahuaDong】[1/7] We trained a world model on 30,000 hours of first-person video.
Open source S3
Going from 300 to 30,000 hours did not narrow the gap, and a 4× larger model barely did either.
Open source S5
[1/7] We trained a world model on 30,000 hours of first-person video. It learned exactly where the hands go. What happens to the things they touch, much less so. Going from 300 to 30,000 hours did not narrow the gap, and a 4× larger model barely did either.View the quoted post on X
Source：Chris Paxton · x.com