I don't see how that issue with floating point (that all programmers should be familiar with) sensibly leads to a conclusion that floating point is just inappropriate for computing things, including probabilities.