> I am actively hostile to neoreaction and the alt-right, routinely block such people from commenting on my Twitter feed, and make it clear that I do not welcome support from those quarters. Anyone insinuating otherwise is uninformed, or deceptive.
-[TODO: I objected that he shouldn't pay the Danegeld like this]
+[I pointed out that](https://twitter.com/zackmdavis/status/1164259164819845120) the people who smear him as a right-wing Bad Guy do so _in order to_ extract these kinds of statements of political alignment as concessions; his own timeless decision theory would seem to recommend ignoring them rather than paying even this small [Danegeld](/2018/Jan/dont-negotiate-with-terrorist-memeplexes/).
When I emailed the posse about it begging for Likes (Subject: "can't leave well enough alone"), Jessica said she didn't get my point. If people are falsely accusing you of something (in this case, of being a right-wing Bad Guy), isn't it helpful to point out that the accusation is actually false? It seemed like I was advocating for self-censorship on the grounds that speaking up helps the false accusers. But it also helps bystanders (by correcting the misapprehension), and hurts the false accusers (by demonstrating to bystanders that the accusers are making things up). By linking to ["Kolmogorov Complicity"](http://slatestarcodex.com/2017/10/23/kolmogorov-complicity-and-the-parable-of-lightning/), I seemed to be insinuating that Yudkowsky was under some sort of duress, but this wasn't spelled out: if Yudkowsky would face social punishment for advancing right-wing opinions, did that mean he was under such duress that saying anything at all would be helping the oppressors?
-----
-In November 2019, I received an interesting reply on my philosophy-of-categorization thesis from MIRI researcher Abram Demski. Abram asked: ideally, shouldn't all conceptual boundaries be drawn with appeal-to-consequences? Wasn't the problem just with bad (motivated, shortsighted) appeals to consequences? Agents categorize in order to make decisions. The best classifer for an application depends on the costs and benefits. As a classic example, it's very important for evolved prey animals to avoid predators, so it makes sense for their predator-detection classifiers to be configured such that they jump away from every rustling in the bushes, even if it's usually not a predator.
+On 3 November 2019, I received an interesting reply on my philosophy-of-categorization thesis from MIRI researcher Abram Demski. Abram asked: ideally, shouldn't all conceptual boundaries be drawn with appeal-to-consequences? Wasn't the problem just with bad (motivated, shortsighted) appeals to consequences? Agents categorize in order to make decisions. The best classifer for an application depends on the costs and benefits. As a classic example, it's very important for evolved prey animals to avoid predators, so it makes sense for their predator-detection classifiers to be configured such that they jump away from every rustling in the bushes, even if it's usually not a predator.
I had thought of the "false-positives are better than false-negatives when detecting predators" example as being about the limitations of evolution as an AI designer: messy evolved animal brains don't bother to track probability and utility separately the way a cleanly-designed AI could. As I had explained in "... Boundaries?", it made sense for _what_ variables you paid attention to, to be motivated by consequences. But _given_ the subspace that's relevant to your interests, you want to run an epistemically legitimate clustering algorithm on the data you see there, which depends on the data, not your values. The only reason value-dependent gerrymandered category boundaries seem like a good idea if you're not careful about philosophy is because it's _wireheading_. Ideal probabilistic beliefs shouldn't depend on consequences.
> No. The thing that's been driving me nuts for twenty-one months is that <strong><em><span style="color: #F00000;">I expected Eliezer Yudkowsky to tell the truth</span></strong></em>. I remain,
>
> Your heartbroken student,
-> [...]
+> Zack M. Davis
I followed it up with another email after I woke up the next morning:
> **I really used to look up to you.** In my previous interactions with you, I've been tightly [cognitively constrained](http://www.hpmor.com/chapter/57) by hero-worship. I was already so starstruck that _Eliezer Yudkowsky knows who I am_, that the possibility that _Eliezer Yudkowsky might disapprove of me_, was too terrifying to bear. I really need to get over that, because it's bad for me, and [it's _really_ bad for you](https://www.lesswrong.com/posts/cgrvvp9QzjiFuYwLi/high-status-and-stupidity-why). I remain,
>
> Your heartbroken student,
-> [...]
-
-
+> Zack M. Davis
[TODO: Sep 2020 categories clarification from EY—victory?!
https://www.facebook.com/yudkowsky/posts/10158853851009228
_ex cathedra_ statement that gender categories are not an exception to the rule, only 1 year and 8 months after asking for it
]
+
+[TODO: "Unnatural Categories Are Optimized for Deception"
+
+Abram was right
+
+the fact that it didn't means that not tracking it can be an effective AI design! Just because evolution takes shortcuts that human engineers wouldn't doesn't mean shortcuts are "wrong" (instead, there are laws governing which kinds of shortcuts work).
+
+Embedded agency means that the AI shouldn't have to fundamentally reason differently about "rewriting code in some 'external' program" and "rewriting 'my own' code." In that light, it makes sense to regard "have accurate beliefs" as merely a convergent instrumental subgoal, rather than what rationality is about
+
+somehow accuracy seems more fundamental than power or resources ... could that be formalized?
+]
+
+And really, that _should_ have been the end of the story. At the trifling cost of two years of my life, we finally got a clarification from Yudkowsky that you can't define the word _woman_ any way you like. I didn't think I was entitled to anything more than that. I was satsified. I still published "Unnatural Categories Are Optimized for Deception" in January 2021, but if I hadn't been further provoked, I wouldn't have occasion to continue waging the robot-cult religious civil war.
projects / Ultimately_Untrue_Thought.git / blobdiff