Having finished reading Chomsky 1992, we will now turn to the minimum description length (MDL) principle. MDL offers a conceptually simple way to view inductive inference (hypothesis formation) in any domain:
Choose the hypothesis that best describes the data
Seems like a no-brainer, right? Indeed, I think pretty much everyone says/thinks something that could be phrased along these lines. The meat of the MDL principle is that ‘best describing the data’ is defined in a particular way:
It seems that both votes are equally divided among the three options (with one person suggesting two topics).1 I suggest then that we begin with a deep read of Chomsky’s “Explaining Language Use,” before moving on to an overview of MDL. Although both MDL and Attribute Grammars received a vote each, I think that there is more synergy involved in the department with the topic of MDL.
Chomksy’s article appears in an issue of a philosophical journal dedicated to the philosophy of Hilary Putnam (one of the most important philosophers of the 20th century).
Course Info
Welcome to the colloquium! I plan to run the colloquium like a reading group. Before I list a few of the topics that I think would be fun to explore this semester, I will describe how I currently envision this course actually working, during the (hopefully short) era of online learning. While we could meet via Skype/Zoom, I’m not completely convinced about the added benefit of that. Instead, I’d like to try having a Faculty of Language-like blog, with an active comments section, where we can raise and discuss issues together.