And still down-playing themselves?
So, assuming this is actually done with image processing (and yes, it's normal to capture a larger frame and calculate a shift within it), why have they decided to go with "optical" image stabilization? Optical solutions fix camera shake, but do nothing to handle subject movement, like the bouncing Nordic woman, which is why professional sports photographers still have large aperture lenses. In extreme cases, as here, keeping the subject static in the frame would result in the background bouncing around, due to the change in perspective from the moving camera position (something Canon have tried to fix in a stabilized macro lens, but not for riding a bike).
There *is* a lot of research into stabilizing/removing blur independently from separate bits of the image - some was presented at SIGGRAPH this year, and Adobe explicitly stated that their work on this was the reason that they'd not yet released their camera-shake-removal technology (demoed recently) in Photoshop. But "optical" it's not.
If it's image processed, you may as well own up to it. Of course, if there really a stabilization element in there, I take it all back - but it's quite possibly not the best solution.
These things always look good in demos. I'll reserve judgement until a real world test, although I don't think my DSLR is going anywhere.