Unlike video streaming libraries, there's no certain way for people to verify the outcome of an ML function is correct (e.g. Would this model work on other data

Unlike video streaming libraries, there's no certain way for people to verify the outcome of an ML function is correct (e.g. Would this model work on other datasets? Is the statistical inference correct?). You need at least some basic knowledge about the underlying math to avoid pitfalls.