In this position paper, we propose the idea that emergent and evolutionary aspects of semantics, which are complementary to the problem of semantic detection, are foundational to multimedia computing. We show that media rich social networks reveal certain implicit assumptions in concept learning about semantics, including semantic stability, emergence, and stability of context. We study the problem of semantic evolution in the context of media rich networks - (a) since meaning is an emergent artifact of human activity, it is crucial to study how human beings interact with, consume and share media data. (b) The ready availability of large scale social interaction datasets of blogs including sites such as Flickr and YouTube, allows us to instrument the relationship between media and human activity at a scale not available to earlier researchers. We have identified three initial problem areas critical to evolutionary aspects of semantics - community discovery, information flow and semantic diversity. We shall present examples of research problems addressed in each of the three areas.