If I work with AI, I have an obligation to understand enough of the system to know when it should not be trusted.