khb writes: It seems that the UN has started a debate on whether to place limits or bans on robots that can kill without manual supervision. It seems that bombs are viewed as "kinder" than robots which might be programmed to achieve specific ends (e.g. destroy that bridge, kill anyone carrying a gun, etc.).
khb writes: Good designers provide error detection and (where feasible) correction at the lowest levels of the system. But how are such mechanisms comprehensively tested through the entire hw+sw stack? Often they aren't. Some lessons from the the Sun e-cache fiasco of 1999.