It needs time to determine if you are searching the web for any text with "* 2" excluding (-) "-2". I'm sure that over time with enough of the same calculation requests from different users it will become much faster at producing the correct response, due to machine learning.