[Filligree]

Consider that a lot of ML programs sit at 100% CPU use (on a single core), and 50-60% GPU use (if you're lucky).

The common use of Python is a huge issue.