TFA lists short run arguments against autonomous robot killers, but a long run argument is that with them, the arms race will get so out-of-control that it will entirely consume the worlds' economies. Once both "sides" have them, they'll just fight amongst themselves, and both sides will have to deploy better and more numerous bots. Since no humans will be dying, there'll be much less pressure to restrict the scale of the wars. With R/C killers, there's still some limit on how many killers can operate per human (necessary to make the kill decision). Even if they do everything autonomously except kill, that still imposes some kind of limit.
What happens when all the world is destitute? Chaos, unrestricted warfare, etc. So banning these things isn't a bad idea.
The biggest problem I see with enforcing a ban is discerning between R/C killers and fully autonomous killers. R/Cs will become so sophisticated that they'll do everything themselves up to the point of pulling the trigger. From an outside observer, would it be clear whether a human ordered it to execute or an additional single line of code did it?
"Today there are no victims of fully autonomous weapons..." Really? How do you know? Author should have said "...as far as I know."