Between the two videos, Mark is obviously making a pitch to various players, and a good one I find. But peeling away that, here's my reading of it:
Firstly, it avoids the chicken/egg problem concerning hardware, by sitting on top of the Android kernel and the Android HAL drivers, both already supplied by the manufacturer (Samsung, HTC et al) for a particular device. Meaning, as I read it, that any Android smartphone device, presumably with sufficiently capable hardware, is more or less automatically and instantly an Ubuntu smartphone device as well. That's pretty clever, particularly since (as we know) any system that relies on completely Open Source drivers for all the hardware, is in for a _much_ tougher battle. That aspect is of course regrettable, but has been a constant factor across all platforms for decades now.
Since the Android kernels are by definition Open Source, it should mean that the only bits on the entire system that would have to be closed source (for now), are the Android HAL drivers. Everything else could be Open Source. Whether that will be the case remains to be seen of course, but one can hope. I believe particularly camera apps can be tricky business in that regard.
Another thing that will be very interesting, is how big a portion of the entire Ubuntu package archive will be immediately and automatically installable on a device, or the extent of the changes needed to a package, for it to be installable and runable on the device. If that number turns out to be great, that obviously is a very interesting proposition, given the amount of great software sitting in that archive.
What will also be very interesting is to learn how easy it will be to use the (presumable) API, that an application should program towards, in order to use all the smartphone capabilities. And also which programming languages can be used with that API. The greater that number, and the lower the barrier the API presents, the more interesting it gets.
As someone who has keenly watched Android from it's inception, and develop mobile webapps for a living (amongst other things), I think, when watching the videos, that there's been some _very_ good thinking going into how to crack the hard nuts, of having a very powerful, capable and versatile system, on the very constrained interface that a smartphone presents. I have to admit that I'm impressed with the demos so far, and even think it looks surprisingly good, particularly considering this early stage. You wont know before you've used it for a week, but I see some fairly good innovation going on in those videos. Combining the best of what's been seen so far on various platforms, avoiding the worst pitfalls, and introducing some novell and smart twists.
It could turn out to be a dodo, but for now I intend to watch keenly what's going on with this. It looks very interesting, and has the _potential_ to be something huge, if people don't screw it up. We'll see...