CSE 390 (for CSE 142)

This is my course review for CSE 390 at the University of Washington. I took the course in autumn 2014 with Stuart Reges (as part of CSE 142), and in winter 2015 with Adam Blank (as part of CSE 143). Below I will talk about my experience from autumn 2014; for my experience from winter 2015, see CSE 390 (for CSE 143).

CSE 390 was by far the most enjoyable course of the quarter. The university doesn’t seem to have strict guidelines for how professors conduct the course, so I suspect the course differs significantly depending on the professor with whom one takes the course.

Basic structure

As Reges discussed during the first meeting, the course sessions were divided roughly in two parts: the first half would be a discussion based on a book we would read throughout the quarter (for us, it was Our Final Invention: Artificial Intelligence and the End of the Human Era), while the other half would be a discussion about a topic related to computer science chosen by Reges.

He said to the effect that during his years at Stanford, what he remembered most from his one math class was a TA (or professor?) who would talk about all of the ideas that interested him—so that this course is an attempt for Reges to “recreate the experience”.

The book

Barrat’s book was a bit basic for me, since I had already been reading LessWrong for several years prior to the course. The content seemed a bit repetitive, and often the arguments seemed to be based solely on Barrat’s opinion (which seemed dubious, since he is apparently a documentary maker, and not an expert in AI). Reges, as well as the class, seemed to find the content repetitive as well. Although Reges did say that superintelligence is something we have to be concerned about, I got the impression that the class was unconvinced by the idea of a superintelligence.

Topics that were discussed

As best as I can remember (I forgot to keep a written list), here are all the topics discussed in the second half of each session, roughly in chronological order.

Other aspects

Perhaps what made the course entertaining was a number of anecdotes Reges told about famous people he had interacted with (including Peter Thiel and Steve Jobs).


  1. Watch out for this proof though; Reges might neglect the fact that some reals can be represented in multiple ways in the decimal system. I pointed this out to him, and he admitted that he likes the subset proof better.

  2. For this reason, he argued vehemently that most people get the wrong message from Chinese whispers, namely that humans are bad at copying messages. What the message should be, according to Reges, is that in principle, i.e. given enough motivation to try, humans can succeed at the game.