Harvard recently developed an improved manufacturing technique for similar robots based on self-deforming materials and origami construction techniques:
Maybe not for continual use, but if you had it land and charge then launch again, you'd greatly increase the deployment time, range, and utility of such a device.
Put a microphone or timed camera on it, and it would be pretty slick.
The time-lines seem unrealistic. While the mechanical design seems impressive, what about imaging, computing, and comms?
I.e. it's one thing to make a mini UAV flight demonstrator, but quite another to put a payload on it and have it transmit data over a usable range, for a useful length of time.
That's only a few hundred bucks worth of off-the-shelf equipment (and a fair few tens or hundreds of hours worth of acquiring the skills needed to make it all work together)
The original article is about a UAV that masses 20 g, whereas the links you mention have video from a Funjet, which is has a mass of 550g, according to this: http://www.barnardmicrosystems.com/L4E_arctic.htm#x3
So, that's 27x the mass -- a completely different problem in terms of engineering.
Depending on your situation, a a usable range, for a useful length of time might mean "around that corner 1 meter away, for 10 seconds" (As anyone who has played a first peson shooter will tell you).
I am sure that weight/size isn't the only parameter they're optimizing around. Likely battery life/flight range is the other primary parameter and optical & audio quality and transmission range secondary priorities they would be considering.
