We just launched a new series on IT Conversations: Money:Tech. You might think "I don't care about financial services" (especially now), but they have some interesting, relevant problems. The first show illustrates that well.
In Data and Capital Markets, Michael Stonebraker discusses why traditional relational databases don't work for many of the problems that financial systems face. Along the way he talks about the power of linguistic abstraction and gives the reason that Oracle, DB2, Sybase, and other "elephant vendors" products run 30-100 times slower than the best solution in a range of problem spaces. For anyone who's interested in data, there's plenty of meat here.