Comment by necovek

Computer programs always fail in predictable — but usually unpredicted — ways.

Human bodies, not so much, mostly because we lack the capability to monitor, measure and emulate the behaviour of such complex systems. As such, we gain medical knowledge using statistics, usually covering most common "failure modes" first, but we increasingly learn that those are never as clear cut either as our observation technology improves (as it does with science otherwise too — eg. Newtonian mechanics is completely true up to some error bars and constraints achievable in that period).