It all sounds a bit too much like the
plot to
Prey (the novel by Michael Crichton). Robotic Helicopters that
watch other helicopters and learn how to maneuver in the same way
as the teachers they watch.
Some students at Stanford built the helicopters. Currently, they
don't have a targeted application - but the suggestion of use in
the military is a bit chilling.
The tricky part comes from the fact that helicopters are hard to
fly. So you can't simply pre-program the moves. But the learning
allows for variation and adjusts, on the fly, in order to pull off
a 
particular trick correctly.
That is all fine and dandy until the learning helicopters, now
equipped with laser death rays for their military use, take a
"wrong turn" and decide to annihilate all of us.
Link:
http://news-service.stanford.edu/news/2008/september10/helicopter-091008.html
Link: My previous
post on Michael Crichton's Prey (in regards to fun AI animation
tests)