Shoveling data into an ML library is relatively straightforward compared to understanding what that library is doing and why.