First (Past the) Post
Today* I embark on the long-planned 30 minute writing project! The rules of the game are very simple: write for 30 minutes a day in some externally visible setting. Right now, this setting is a shared page on Workflowy because of bandwidth constraints: soon, it will move to its own Hakyll location.
The goal is to write consistently enough to learn what I like writing about, and eventually kick off a bona-fide blog that isn't merely a collection of journal entries. (Of course, each 30 minute could be a journal entry.) Enough with the formalities! Right now I am on the bus home, reading about predictive processing and free energy. I started here:
https://slatestarcodex.com/2019/03/20/translating-predictive-coding-into-perceptual-control/ Scott Alexander writes that it's helpful to understand free energy and predictive coding (which sounds like uses a lot of interesting math) as perceptual control. He then described perceptual control in a really simple way, as just figuring out how to move yourself towards a set-point. THEN Scott linked to an article purporting to explain the free energy framework mathematically and said it was an example of how confusing said framework was. I click on the article: https://medium.com/@solopchuk/tutorial-on-active-inference-30edcf50f5dc get a quarter way through, and realize poor Scott has been had! I am so offended by the mathematical casualness and carelessness of the author (not his fault specifically, typical of the ML community) that I have to go dig up the original paper: http://bdl.kaist.ac.kr/lecture/2011springbis525/10%20(2010)%20The%20free-energy%20principle-%20a%20unified%20brain%20theory.pdf By the way, the author of this thing (Karl Friston) has the highest academic statistics I've ever seen (h-index over 200). I start reading the paper, and... here we are! I'm about to get off the bus. Well, this has been the inaugural session of the 30 minute project. Stay tuned!
*actually written on Apr 6