For the same reason, it's "machine learning" rather than "non parametric statistics".
and while I agree the term is really loaded, ML is different from statistics in its approach and the discoveries it made. For example, ML showed that some techniques are much more effective at scale (large datasets) than suspected by most statisticians, as recognized by "hard core" statisticians like Wasserman.
(Note table near start of article, particularly entries for "large grant" and "nice place to have a meeting.")
Traditionally statistics is more geared towards hypothesis testing and analytics. While Machine Learning is about making prediction.
Only few methods such as K nearest neighbors are non parametric, rest all methods such as SVM's, decision trees are all parametric.
So using it as a jargon is a bad idea.