The challenge of alignment: it is virtually impossible to define a perfect objective, there is always a way to circumvent it. Human values are not uniform, let alone when expressed in a way that AI can understand.