/optimizing-styles/ More patient than death. Wed, 03 Apr 2019 04:01:42 +0000 hourly 1 https://wordpress.org/?v=5.9.2 /optimizing-styles/#comment-26 Tue, 10 Jan 2017 19:56:14 +0000 /?p=33#comment-26 The advantage of explicit approximations of the one true way is that explicitness makes engineering easier. It’s easier to fuse them with other styles. Thinking things through in terms of the work they are doing in producing value of information and outcome probability redistributing is a good way to progress through switching between styles depending on the situation, through blending them fluidly, through using their parts in new ways, toward the one true style.

]]> /optimizing-styles/#comment-25 Tue, 10 Jan 2017 19:39:52 +0000 /?p=33#comment-25 The law-chaos axis in the Dungeons and Dragons alignment system is a mix of values and style.
That’s why “Chaotic Lawful Neutral Good” is a totally sensible thing.
http://www.facebook.com/yudkowsky/posts/10153021749629228

]]> /optimizing-styles/#comment-22 Tue, 10 Jan 2017 07:49:17 +0000 /?p=33#comment-22 This extends to epistemics as well. There are things one could plausibly mean by “pessimism” and “optimism” that are specialities, rather than biases. Like if there are a bunch of highly uncertain distributions which could be good or bad, and you know things that can truthfully narrow down a certain class of them which happens to be bad, you are a pessimist. (If you’re an ideal Bayesian, and you know you’ve been handed a speciality like this out of a bag of equal good-finding and bad-finding specialities, then your total estimate of good/bad stays the same, but the good is pushed out to and spread out across the other very uncertain distributions, which you may not talk about as much, since you (truly) don’t know as much, making that optimism less noticeable.)

]]>