> I think it has potential to be the language of choice for non-neural-network stats/ML (that's python for the foreseeable future).

I think it will overtake python for NNet stuff also: https://julialang.org/blog/2018/12/ml-language-compiler