We at SonarSource are currently actively looking at levering machine learning techniques to enhance our products. Automatic classification of issues as false positives is indeed one application, but there also are others such as automatically inferring what the best practices are in general (from a lot of projects) or within a particular project. This exploration is still in its early stage, so we are not yet able to communicate about specific roadmap points nor timeline.
Your link does not seem related to SonarQube. There are many products out there called “sonar” (including ourselves previously) and this was a key reason for us to rename our own product form “sonar” to “SonarQube” in the first place: to avoid confusion.