I'd argue it's just "languages that are not Python struggle in this space". Python has an undoubtably unique ecosystem and gravity around ML, but that's more a unique advantage of Python than a weakness of every other language design
Scala does have Spark's ML capabilities. Spark's ML capabilities might not be the best, but they can be run on huge datasets. I'm guessing Dask has better model support, but isn't as scalable as Spark ML.
