This is such a fascinating topic because the field of robotics outside of industrial applications is still so nascent. I'm not here to weigh in, just to drop off some useful tidbits for other people interested in this ethical gray area as well.

This theme of "terminators" and "killer robots" has been really prevalent in the field lately because of the DARPA Robotics Challenge[1], the latest in DARPA's Grand Challenges, which I'm currently participating in on one of the Track B teams. Many people see the break away from bomb-squad bots and factory floor robotic arms in to humanoids as a really scary thing, and the DRC seems to amplify that (if a robot can hold a sawzall, it can hold a rifle).

Just last month in Atlanta, the IEEE Humanoids conference took place. Dr. Ronald C. Arkin[2] gave a talk exactly on this topic from an ethics perspective, titled "How to NOT build a Terminator", and it was exceptional. It was a plenary talk and not a paper, so I can't really find any record of it to share. Unfortunate.

Tangentially, the lab that I work in (the robotics group at the Florida Institute for Human and Machine Cognition[3][4]) has employed and continues to employ a principle that we call "co-active design"[5] where we actively work to keep a human in the loop at all times; we're definitely not looking to build a "killer robot". It's an interesting design problem that overlaps a lot with UI and UX, popular topics here on HN.

And lastly, a shameless plug for the field itself; a lot of people don't realize just how software-oriented robotics research (especially humanoids, where the fun problems are) is. A lot of people are stuck on it being a hardware endeavor. While it's true that a chunk of robotics falls in the mechanical engineering domain, there's plenty of room for hackers from tons of different disciplines to get involved. There's interesting people solving interesting problems, from the Open Source Robotics Foundation (the Willow Garage spin-off) to private groups like Boston Dynamics, yet so many people still see the field as a black-box that only opens up for the hardware inclined. I could see a talented group with the right hacker mindset doing some really interesting stuff in robotics with the right impetus and execution.

1:http://theroboticschallenge.org

2:http://en.wikipedia.org/wiki/Ronald_C._Arkin

3:http://ihmc.us

4:http://robots.ihmc.us

5:http://www.jeffreymbradshaw.net/publications/20101008_Coacti...