Title: Sexual Dimorphism in Yudkowsky's Sequences, in Relation to My Gender Problems
Date: 2021-03-21 11:00
Category: commentary
-Tags: autogynephilia, bullet-biting, cathartic, Eliezer Yudkowsky, epistemic horror, my robot cult, personal, sex differences, Star Trek
+Tags: autogynephilia, bullet-biting, cathartic, Eliezer Yudkowsky, epistemic horror, ideology, my robot cult, personal, sex differences, Star Trek
Status: draft
> _I'll write my way out
> Is it cheating if you deliberately define your personal identity such that the answer is _No_?
-To which I now realize the correct answer is—_yes!_ Yes, it's cheating! Category-membership claims of the form "X is a Y" [represent hidden probabilistic inferences](https://www.lesswrong.com/posts/3nxs2WYDGzJbzcLMp/words-as-hidden-inferences); inferring that entity X is a member of category Y means [using observations about X to decide to use knowledge about members of Y to make predictions about features of X that you haven't observed yet](https://www.lesswrong.com/posts/gDWvLicHhcMfGmwaK/conditional-independence-and-naive-bayes). But this AI trick can only _work_ if the entities you've assigned to category Y are _actually_ similar—if they form a tight cluster in configuration space, such that using the center of the cluster to make predictions about unobserved features gets you _close_ to the right answer, on average.
+To which I now realize the correct answer is—_yes!_ Yes, it's cheating! Category-membership claims of the form "X is a Y" [represent hidden probabilistic inferences](https://www.lesswrong.com/posts/3nxs2WYDGzJbzcLMp/words-as-hidden-inferences); inferring that entity X is a member of category Y means [using observations about X to decide to use knowledge about members of Y to make predictions about features of X that you haven't observed yet](https://www.lesswrong.com/posts/gDWvLicHhcMfGmwaK/conditional-independence-and-naive-bayes). But this AI trick can only _work_ if the entities you've assigned to category Y are _actually_ similar in the real world—if they form a tight cluster in configuration space, such that using the center of the cluster to make predictions about unobserved features gets you _close_ to the right answer, on average.
The rules don't change when the entity X happens to be "my female analogue" and the category Y happens to be "me". The ordinary concept of "personal identity" tracks how the high-level features of individual human organisms are stable over time. You're going to want to model me-on-Monday and me-on-Thursday as "the same" person even if my Thursday-self woke up on the wrong side of bed and has three whole days of new memories. When interacting with my Thursday-self, you're going to be using your existing mental model of me, plus a diff for "He's grumpy" and "Haven't seen him in three days"—but that's a _very small_ diff, compared to the diff between me and some other specific person you know, or the diff between me and a generic human who you don't know.
Same principle when Yudkowsky wrote, "If I fell asleep and woke up as a true woman [...] I don't think I'd call her 'me'". The claim is that psychological sex differences are large enough to violate the Lipschitz condition imposed by our _ordinary_ concept of personal identity. Maybe he was wrong, but if so, that cashes out as being wrong _about_ how similar women and men actually are (which in principle could be operationalized and precisely computed, even if _we_ don't know how to make it precise), _not_ whether we prefer the "call her me" or "don't call her me" conclusion and want to _retroactively redefine the meaning of the words in order to make the claim come out "true."_
-Do people ever really recover from being religious? I still endorse the underlying psychological motivation that makes me prefer the "call her me" conclusion, the _intention_ that made me think I could get away with defining it to be true. Now that I don't believe that anymore—
-
-[...]
+Do people ever really recover from being religious? I still endorse the underlying psychological motivation that makes me prefer the "call her me" conclusion, the _intention_ that made me think I could get away with defining it to be true—even while I don't believe that anymore.
While [the Sequence explaining Yudkowsky's metaethics](https://www.lesswrong.com/tag/metaethics-sequence) was being published (which a lot of people, including me, didn't quite "get" at the time; I found a [later précis](https://www.lesswrong.com/posts/zqwWicCLNBSA5Ssmn/by-which-it-may-be-judged) easier to understand), I was put off by the extent to which Yudkowsky seemed to want to ground the specification of value in [the evolved design of the human brain](https://www.lesswrong.com/posts/cSXZpvqpa9vbGGLtG/thou-art-godshatter), as if culturally-defined values were irrelevant, to be wiped away by [the extrapolation of what people _would_ want if they knew more, thought faster, _&c._](https://arbital.com/p/normative_extrapolated_volition/).
Your morality doesn't have to converge with that of your genetic twin who was raised in a culture with a different ideology/religion—maybe culturally learned values from childhood get "frozen" after reasoning ability kicks in, such that I would never see eye-to-eye with my genetic analogue who was raised with (say) a traditional Muslim view of relations between the sexes, no matter how much we debates and no matter how many new facts we learned.
-At the same time, while reflecting on one's own values and how to refine them in response to new information and new situations, the question of what your genetic analogue raised in a different culture would think is _relevant_ information. ["Anyone who gives a part of themselves to a[n ideology] discovers a sense of beauty in it."](https://www.lesswrong.com/posts/LRKXuxLrnxx3nSESv/should-ethicists-be-inside-or-outside-a-profession) However pure my heart was, my teenage religion got a lot of things _wrong_ on the empirical merits. When figuring out how to [rescue the spirit](https://arbital.greaterwrong.com/p/rescue_utility) of early-twenty-first century (pre-[Great Awokening](https://www.vox.com/2019/3/22/18259865/great-awokening-white-liberals-race-polling-trump-2020)) egalitarian individualism in light of the terrible discovery that [_none of this shit is true_](/2020/Apr/book-review-human-diversity/), it's instructive to consider how you would have formulated your values, if you had always known the true state of affairs to begin with (even if, as a point of historical fact, your mistaken beliefs had a causal role in the development of the values you're trying to rescue).
+At the same time, while reflecting on one's own values and how to refine them in response to new information and new situations, the question of what your genetic analogue raised in a different culture would think is _relevant_ information.
-When I introspect on the _causes_ of my whole gender ... _thing_, I see three parents in the causal graph: autogynephilia, being a sensitive boy rather than a macho guy, and my ideological commitment to antisexism (wanting to treat feminism as a religion, as a special case of egalitarianism as our state religion). The first two things seem likely to be more "innate", more robust to perturbations—
+When I introspect on the _causes_ of my whole gender ... _thing_, I see three parents in the causal graph: autogynephilia, being a sensitive boy rather than a macho guy, and my ideological commitment to antisexism (wanting to treat feminism as a religion, as a special case of egalitarianism as our state religion). The first two things seem likely to be more "innate", more robust to perturbations—but the the ideology thing mostly seems like a _mistake_—albeit a beautiful mistake that I formed my values in the context of—and the process of figuring out how to do better would benefit from looking at the space of _other_ possible mistakes.
-[...]
+["Anyone who gives a part of themselves to a[n ideology] discovers a sense of beauty in it."](https://www.lesswrong.com/posts/LRKXuxLrnxx3nSESv/should-ethicists-be-inside-or-outside-a-profession) However pure my heart was, my teenage religion got a lot of things _wrong_ on the empirical merits. When figuring out how to [rescue the spirit](https://arbital.greaterwrong.com/p/rescue_utility) of early-twenty-first century (pre-[Great Awokening](https://www.vox.com/2019/3/22/18259865/great-awokening-white-liberals-race-polling-trump-2020)) egalitarian individualism in light of the terrible discovery that [_none of this shit is true_](/2020/Apr/book-review-human-diversity/), it's instructive to consider how you would have formulated your values, if you had always known the true state of affairs to begin with (even if, as a point of historical fact, your mistaken beliefs had a causal role in the development of the values you're trying to rescue).
Suppose it _is_ true that female adults and male adults have distinct transpersonal values. At the time, I found the prospect horrifying—but that just shows that the design of male transpersonal values _contains within it_ the potential (under appropriate cultural conditions) to be horrified by sex differences in transpersonal values. If, naïvely, [I don't _want_ it to be the case that women are a different thing that I don't understand](/2019/Jan/interlude-xvi/), but that preference _itself_ arises out of—something like, the love of a man for a woman leading to, wanting to be aligned with women in the sense of AI alignment, and genuinely do right by them—which vision is threatened by the idea of the sexes having fundamentally different interests.
projects / Ultimately_Untrue_Thought.git / blobdiff