I know @Quesito and @loujin are doing data visualisations based on MB data. I'm also in the planning stage of my own data visualisation project right now. See e.g., @loujin's project:
Some of the data also gets used for various academic purposes. I know a lot of the Turkish and Indian related Work attributes were added because some of the UPF-MTG students were doing analysis work on that (@alastairp might have a link to any publications or anything that came out of that).
Also, combined with ListenBrainz and AcousticBrainz data, we will hopefully see more interesting projects pop up too, combining 2 or 3 of the datasets together. (E.g., recommendation engines based on who produced the music?)
Speaking of, note that LB, AB, CritiqueBrainz, AcoustID, FanArt.tv, … are all essentially "3rd party" services building on top of MusicBrainz, utilising MusicBrainz's unique identifiers. (Cover Art Archive also, somewhat, but since the coupling between CAA images and MusicBrainz Releases happen in the MusicBrainz database itself, not really. A project similar to CAA could have easily been made decoupled from MusicBrainz though, similar to how LB, AB, CB, AcoustID, and FanArt.tv works: by internally using MBIDs as external identifiers. (In fact, FanArt.tv and CAA are basically functionally the same, just with differing project goals.))