Sounds good in theory, but...
The implementation will need to be very good.
The central contradiction as I imagine it is that normally you move your eyes to look at something. If moving your eyes also moves the thing then you're getting very unnatural feedback.
If they solve that by having the scrolling occur only when you look at certain areas of the screen — e.g. look at the bottom to scroll down, return your eyes to the main portion to stop — then you have to judge when to stop based on peripheral vision.
Supposing the people at Samsung have solved these problems then they deserve to be number one.