Dr Kevin Knight is one of those guys that I think of when I ponder the word 'Brilliant' (he's not alone, ISI is chockablock with people I place in this category). It's inspirational to be working in the same department as someone who builds computer systems that do wacky things with language: automatically translate mysterious languages only spoken by a handful of people into English, translate rhyming poetry well, or by cracking famously unsolved codes.
His team of researchers are ISI are all a little scary-brilliant in their own way. So, as the organizer of AI seminars here at ISI, I had a feeling we were in for something good when he sent me a terse, efficient email asking to give a seminar with a one-line abstract:
"The Copiale cipher is a 105-page enciphered book dated 1866. We describe the features of the book and the method by which we deciphered it."
The talk itself was typically brilliant, a walk through of how his shared interest in amateur codebreaking with his dad lead him to apply statistical machine learning techniques to the problem of pulling out the structure of this complex 19th century code. Listening to Kevin give presentations is always wonderful since he just has this knack of making really complex things seem simple and leads you through complex mathematical analyses in a way that you can see what he's doing and why. So, after many iterations and insights; figuring out that the language of the Cipher was probably German, and finally making the connection that the Cipher was probably a code governing rules of initiation rites of a secret society, we all felt that we'd just listened to a Tintin story rather than a scientific talk.
Naturally, the good people at the NY and LA Times know a good thing when they hear it and put together a nice write-up of Kevin's work. So Congratulations to Dr Knight for some really fascinating, fun work and thanks for sharing with us!