The problem with the technology is that it doesn't do anything other than what they show in those videos. Including animation. Those things can't move. And there's a fair bit of instancing going on in those videos.. it's not a million different rocks, it's a million copies of the same rock.
They're also ignoring the fact that poly-count is actually pretty low on the priority list when it comes to realism and "good graphics" in games these days. It's all about shaders and lighting to make the surface of the objects look real and behave like real objects. Games like Crysis and Battlefield 3 don't look the way they do because of how many polygons they have.
Even if this were magically implemented across the board, it would only serve to increase the effective polygon count, and the overall look of games wouldn't really change that much.
There are other things at work these days, like tessellation in DX11, that vastly increase the polygon count and the smoothness of objects, and results in a barely noticeable increase in the overall look of the game.
Dig around the 'net, you'll see a lot of feedback from actual developers that are pointing out all the problems that Mr. Unlimited isn't mentioning.
Yes, it's an intriguing idea, but the general consensus is that it won't really work within a game development environment because of its limitations.