Fundamentally there is uncertainty associated with any physical measurement which is usually proportional to the magnitude being measured. As long as floating point is << this uncertainty results are equally predictive. Floating point numbers bake these assumptions in.