Waste can be mitigated by designing the power supplies in the form of interchangeable packs, which can be moved from one generation of flying machine to the next. Risk can be mitigated with good design principles, overall, and advances usually aren't made by people who are afraid of handling risk. That only leaves performance as the real issue. If my quadcopter's chemical battery allows for 1 hour of flight with a load of 1 pound, then how do I increase the time or the load by a significant factor, when required? Unless there is some major breakthrough in chemical batteries, it seems that nuclear is the only real option here. Therefore: Can we build, using today's technology, a nuclear battery (or even a reactor) which can power quadcopters and mechanical bees [1]?
I use the combined example of quadcopters and bees because, perceptually, they appear to be on the same physical level (as opposed to a rock versus a molecule, for instance), but they require very different design decisions as to how they are powered. Chemical processes used for a quadcopter won't easily translate to a bee (unless we are using something exotic, such as ATP manipulated by nanotech), whereas nuclear processes could easily transition between the scales occupied by the quadcopter and the bee.
NOTE TO SELF: Write a children's technology coloring book called "The Quadcopter and the Bee", and after some initial success, turn it into a series of eyewatering, yet thought-inducing, books and movies, thus making lots of money, which would in turn be used for your "Nuclear Batteries Are Us" startup.
[1] http://news.ycombinator.com/item?id=3606394