We’re very good at spotting patterns in data, but applying simple mathematical models isn't always useful, argues Professor Michael Stumpf. (Extract 3 of 4 from the Imperial College Podcast 21 March 2012)