Then you get into the difficulty of "what is a human being?" and "what is harm?"

To say nothing of muddying the waters a little (would you classify Florence ambrose as human? Sam starfall even?) with 'non classical humans'

and yes, Coding such values would be a right bastard, Mostly because in order for a robot to understand the first directive, you have to nominate 'harm', 'inaction', 'action' and 'human being' to the point where it would be simpler to leave out everything _but_ what constitutes harm and let it go at that..