Essentially, this question could also be asked: how many species do all of the paleontological specimens in collections (or collections records) represent?
I recognize that species is a temporal concept when thinking about differentiation across time; but, natural history collection catalog specimens to species level, so there must be a reasonable basis to make this distinction. Also, note that I'm not exactly interested in inferred species (i.e. if we have records of species A and species B, we might infer a last common ancestor, species C).