The reason that 3D cards are being accessed for 2D GUIs now is to handle some of the transparencies, shading, or vector intensive graphics in today’s user interface.
The same functions of the graphics card that speed up rendering of polygons in 3D games can be used (with proper coding) to speed up 2D buttons, windows, and other interface elements. The same can be said for the drop shadows MS places behind their cursor and windows.
But who knows? Maybe they will come up with new interface elements that use lots of texture maps, transparencies, fog, and anti-aliasing. But it will still probably all be within the bounds of today’s 2D windowing environment.