Beyond being informative:
http://www.economist.com/node/21531433.. this illustrates one of the reasons Machine Learning, and the Stanford class, has been so successful lately.
From the article:
Another potential advantage for UAS is that future designs may be better able to survive in contested airspace than manned aircraft are. Without the need to accommodate crew, drones can be given strange radar-cheating stealthy shapes. They may also acquire “hyper-manoeuvrability”.
Now look at a video from the Stanford ML course intro. It is a small helicopter that is too small and fast to use usual piloting. Hyper-manoeuvrability at work.
To solve the problem, they used machine learning for weird capabilities such as flying upside down and other stunts. Initially deterministic methods were used to attempt these maneuvers. It was too difficult and error prone, for a craft of this size the decision speeds were just too great for current algorithms.
So ML approaches were taken .. producing this extraordinary capability.
And yes, they can make quieter, smaller and deadly versions. Ethicists in the military are deciding when these capabilities can target and kill with no human in the loop.
Its not too late to look into the Stanford class.
-- Owen
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at
http://www.friam.org
Locked
