Honestly, even if they were using the robot dogs as attack dogs, I'm a whole lot more comfortable with robots with auditable source code (or at least an auditab

Honestly, even if they were using the robot dogs as attack dogs, I'm a whole lot more comfortable with robots with auditable source code (or at least an auditable training procedure) than I am with actual dogs that are basically trainable black boxes, and certainly more than I am with humans, who are trainable black boxes that can rationalise their actions with plausible deniability.