Robots are tools, like land mines and cruise missiles are tools. There is nothing more complicated than that. They are amoral.

If you want to have the kurzweillian discussion about robots as smart as humans, then you can't program the ethics, just like humans are too smart to program ethics.

Papers like this are just completely premature. A discussion about practical & intelligent safety systems for robotics would be a better way to describe the intention.

I have just read the introduction, and it does seem that the author is looking far ahead, but I don't think it's too early to reflect on the ethical aspects of lethal robots.

Science moves faster than the field of ethics (e.g. cloning) and the US has already lethal robots on the battlefield. They are not autonomous yet, but we're certainly going there. When the robot 'assists' the decision by providing additional data (say an sensor detects a hidden weapon on an otherwise civilian-looking guy), it already skews the decision process. As the author mentions, there will be a point when the military will start to think that their robots are better at making the shoot/not shoot decisions...

That's not how the military works.

At least for the next decade, humans will always be making the decisions. I know. I've met with people in the military about weaponizing robots.

The real safety addition comes from software aided aiming and added situational awareness. Imaging trying to manipulate a gun remotely with lag, vs. just saying "fire" to an auto-tracking gun. It's a world of difference.

But this is why I call it a "safety system". It has nothing to do with ethics. It's like that IR beam that makes sure your garage door doesn't close on someone. Useful, important, and nothing to do with ethics.