The question he's asking is whether there is anywhere we can innovate in robotics without having a huge budget. There is generally one huge problem with traditional UAVs, that'd be the cost. So how chea and simple can a UAV be?
There are two functions in aerial robotics, stabilisation, just keeping the thing aloft, and then there is navigation. Traditionally auto pilots do both. But if you unlink the two problems then you can buy fairly cheap off the shelf stuff. If you use an off the shelf unit for stabilisation then you've turned a hard 3D problem into a simpler 2D problem.
Separating the two functions, giving the ailerons to the stabilisation unit, and the rudder to the navigation unit, means that you end up with crabbing into turns rather than cleanly banking into turns. It's not pretty, but it works. Which is okay...
He talking about the the different autopilots he's built, starting off with Lego, cell phones and moving on to Arduino boards. Pretty cool stuff.
Moving on to blimps. The blimp uses an ultra sonic sensor and a vertically mounted propellor to maintain altitude, and an infra-red sensors to allow it to follow people around.
Update: A short movie of the blimp demo where it uses infra-red to follow someone around the room.
Update: The minimum blimp demo at ETech got picked up by Boing Boing TV...