The whole concept of micro air vehicles (MAVs) tends to creep me out, mostly because I see them as having a lot of potential for abuse from both the police if you aren't doing anything wrong you have no reason to hide and the criminal element, which tends to have the same opinion, though more monetarily oriented.
But the science behind them fascinates me, which is why I found this article from IEEE Spectrum about research at Cornell to be fascinating. Basically, they're testing the use of a neural net to allow a MAV, equipped with a single camera (we don't need no binocular vision!), to 'read' an environment rapidly and find a safe path through obstacles. Another step towards truly autonomous vehicles. Scary, but fascinating.