I've been reading for my comprehensive exams. I won't bore anyone with the details, but in Geography at CU, that means selecting three subtopics of geography developing three extensive reading lists in each. Then reading everything in the list and then being tested over the content. For various reasons, I've been blessed with relatively short reading lists. I have, combined, about 100 journal articles and book sections plus about 20 books.
The reading lists are designed specifically to give a feel for recent development in each subtopic grounded with the "classics". One of my areas is "Cartographic Generalization". That means (depending on how you ask) the process of creating abstractions geographic phenomena in reality or simply taking a set of such abstractions at one scale (say, 1:24,000) and creating a new set at a smaller scale (say, 1:100,000). Much of what I am reading in this subtopic is Geographers trying to apply squirrely Artificial Intelligence methods to a problem that is uniquely tied to our relation with the world around us. It's almost sad...
The process has really made me appreciate the continuum of things humans are good at versus things computers are good at. This continuum does not seem to be shifting (i.e., computers aren't taking the place of human) inasmuch as it is becoming better defined (i.e., the computer is becoming a better tool).
On my to-do list: Write a critique of automated generalization drawing from Dreyfus' What Computers Still Can't Do.