I think I have just enough language to _start_ to talk about what it would mean. Since sex isn't an atomic attribute, but rather a high-level statistical regularity such that almost everyone can be cleanly classified as "female" or "male" _in terms of_ lower-level traits (genitals, hormone levels, _&c._), then, abstractly, we're trying to take points from male distribution and map them onto the female distribution in a way that preserves as much structure (personal identity) as possible. My female analogue doesn't have a penis (because then she wouldn't be female), but she is going to speak American English like me and be [85% Ashkenazi like me](/images/ancestry_report.png), because language and autosomal genes don't have anything to do with sex.
-The hard part has to do with traits that are meaningfully sexually dimorphic, but not as a discrete dichotomy—where the sex-specific universal designs differ in ways that are _subtler_ than the presence or absence of entire reproductive organs. (Yes, I know about [homology](https://en.wikipedia.org/wiki/Homology_(biology))—and you know what I meant.) We are _not_ satisfied if the magical transformation technology swaps out my penis and testicles for a functioning female reproductive system without changing the rest of my body, because we want the end result to be indistinguishable from having been drawn from the female distribution (at least, indistinguishable _modulo_ having my memories of life as a male before the magical transformation), and a man-who-somehow-magically-has-a-vagina doesn't qualify.
+The hard part has to do with traits that are meaningfully sexually dimorphic, but not as a discrete dichotomy—where the sex-specific universal designs differ in ways that are _subtler_ than the presence or absence of entire reproductive organs. (Yes, I know about [homology](https://en.wikipedia.org/wiki/Homology_(biology))—and _you_ know what I meant.) We are _not_ satisfied if the magical transformation technology swaps out my penis and testicles for a functioning female reproductive system without changing the rest of my body, because we want the end result to be indistinguishable from having been drawn from the female distribution (at least, indistinguishable _modulo_ having my memories of life as a male before the magical transformation), and a man-who-somehow-magically-has-a-vagina doesn't qualify.
The "obvious" way to to do the mapping is to keep the same percentile rank within each trait, but take it with respect to the target sex's distribution. I'm 5′11″ tall, which [puts me at](https://dqydj.com/height-percentile-calculator-for-men-and-women/) the 73rd percentile for American men, about 6/10ths of a standard deviation above the mean. So _presumably_ we want to say that my female analogue is at the 73rd percentile for American women, about 5′5½″.
-You might think this is "unfair": some women—about 7 per 1000—are 5′11″, and we don't want to say they're somehow _less female_ on that account, so why can't I keep my height? But if we refuse to adjust for every trait for which the female and male distributions overlap (on the grounds that _some_ women have the same trait value as my male self), we don't end up with a result from the female distribution.
+You might think this is "unfair": some women—about 7 per 1000—are 5′11″, and we don't want to say they're somehow _less female_ on that account, so why can't I keep my height? The problem is that if we refuse to adjust for every trait for which the female and male distributions overlap (on the grounds that _some_ women have the same trait value as my male self), we don't end up with a result from the female distribution.
The typical point in a high-dimensional distribution is _not_ typical along each dimension individually. [In 100 flips of a biased coin](http://zackmdavis.net/blog/2019/05/the-typical-set/) that lands Heads 0.6 of the time, the _single_ most likely sequence is 100 Heads, but there's only one of those and you're _vanishingly_ unlikely to actually see it. The sequences you'll actually observe will have close to 60 Heads. Each such sequence is individually less probable than the all-Heads sequence, but there are vastly more of them. Similarly, [most of the probability-mass of a high-dimensional multivariate normal distribution is concentrated in a thin "shell" some distance away from the mode](https://www.johndcook.com/blog/2011/09/01/multivariate-normal-shell/), for the same reason. (The _same_ reason: the binomial distribution converges to the normal in the limit of large _n_.)
Take faces. People are [verifiably very good at recognizing sex from (hair covered, males clean-shaven) photographs of people's faces](/papers/bruce_et_al-sex_discrimination_how_do_we_tell.pdf) (96% accuracy, which is the equivalent of _d_ ≈ 3.5), but we don't have direct introspective access into what _specific_ features our brains are using to do it; we just look, and _somehow_ know. The differences are real, but it's not a matter of any single, simple measurement you could perform with a ruler (like the distance between someone's eyes). Rather, it's a high-dimensional _pattern_ in many measurements you could take with a ruler, no one of which is definitive. [Covering up the nose makes people slower and slightly worse at sexing faces, but people don't do better than chance at guessing sex from photos of noses alone](/papers/roberts-bruce-feature_saliency_in_judging_the_sex_and_familiarity_of_faces.pdf).
-Notably, for _images_ of faces, we actually _do_ have transformation technology! (Not "magical", because we know how it works.) AI techniques like [generative adversarial networks](https://arxiv.org/abs/1812.04948) and [autoencoders](https://towardsdatascience.com/generating-images-with-autoencoders-77fd3a8dd368) can learn the structure of the distribution of facial photographs, and use that knowledge to [synthesize faces from scratch](https://thispersondoesnotexist.com/)—or [do things like](https://arxiv.org/abs/1907.10786) sex transformation, as demonstrated by the likes of [FaceApp](https://www.faceapp.com/), the _uniquely best piece of software in the world_.
+Notably, for _images_ of faces, we actually _do_ have transformation technology! (Not "magical", because we know how it works.) AI techniques like [generative adversarial networks](https://arxiv.org/abs/1812.04948) and [autoencoders](https://towardsdatascience.com/generating-images-with-autoencoders-77fd3a8dd368) can learn the structure of the distribution of facial photographs, and use that knowledge to synthesize faces from scratch (as demonstrated by [_thispersondoesnotexist.com_](https://thispersondoesnotexist.com/))—or [do things like](https://arxiv.org/abs/1907.10786) sex transformation (as demonstrated by [FaceApp](https://www.faceapp.com/), the _uniquely best piece of software in the world_).
-If you let each pixel vary independently, the space of possible 1024x1024 images is 1,048,576-dimensional, but the vast hypermajority of those images aren't photorealistic human faces. Letting each pixel vary independently is the wrong way to think about it: changing the lighting or pose can change a lot of pixels in what humans would regard as images of "the same" face. So instead, our machine-learning algorithms learn a [compressed](https://www.lesswrong.com/posts/ex63DPisEjomutkCw/msg-len) representation of what makes the tiny subspace (relative to images-in-general) of _faces in particular_ similar to each other, and that [latent space](https://towardsdatascience.com/understanding-latent-space-in-machine-learning-de5a7c687d8d) is a lot smaller—say, 512 dimensions.
+If you let each pixel vary independently, the space of possible 1024x1024 images is 1,048,576-dimensional, but the vast hypermajority of those images aren't photorealistic human faces. Letting each pixel vary independently is the wrong way to think about it: changing the lighting or pose changes a lot of pixels in what humans would regard as images of "the same" face. So instead, our machine-learning algorithms learn a [compressed](https://www.lesswrong.com/posts/ex63DPisEjomutkCw/msg-len) representation of what makes the tiny subspace (relative to images-in-general) of _faces in particular_ similar to each other. That [latent space](https://towardsdatascience.com/understanding-latent-space-in-machine-learning-de5a7c687d8d) is a lot smaller (say, 512 dimensions), but still rich enough to embed the distinctions that humans notice: [you can find a hyperplane that separates](https://youtu.be/dCKbRCUyop8?t=1433) smiling from non-smiling faces, or glasses from no-glasses, or young from old, or different races—or female and male. Sliding along the [normal vector](https://en.wikipedia.org/wiki/Normal_(geometry)) to that [hyperplane](https://en.wikipedia.org/wiki/Hyperplane) gives the desired transformation: producing images that are "more female" (as the model has learned that concept) while keeping "everything else" the same.
-[TODO: separating hyperplane / Face editing with Generative Adversarial Networks: https://youtu.be/dCKbRCUyop8?t=1433 ]
+Two-dimensional _images_ of people are _vastly_ simpler than the actual people themselves in the real physical universe. But _in theory_, a lot of the same _mathematical principles_ would apply to hypothetical future nanotechnology-wielding AI systems that could synthesize a human being from scratch (this-person-_didn't_-exist-dot-com?), or do a real-world sex transformation (PersonApp?)—and the same statistical morals apply to reasoning about sex differences in psychology and (which is to say) the brain.
-[...]
+Daphna Joel _et al._ [argue](https://www.pnas.org/content/112/50/15468) [that](https://www.pnas.org/content/112/50/15468) human brains are "unique 'mosaics' of features" that cannot be categorized into distinct _female_ and _male_ classes, because it's rare for brains to be "internally consistent"—female-typical or male-typical along _every_ dimension. It's true and important that brains aren't _discretely_ sexually dimorphic the way genitals are, but as [Marco del Guidice _et al._ point out](http://cogprints.org/10046/1/Delgiudice_etal_critique_joel_2015.pdf), the "cannot be categorized into two distinct classes" claim seems false in an important sense. The lack of "internal consistency" in Joel _et al._'s sense is exactly the behavior we expect from multivariate normal-ish distributions with different-but-not-vastly-different means. (There aren't going to be many traits where the sexes are like, _four_ or whatever standard deviations apart.) It's just like how sequences of flips of a very Heads-biased and very Tails-biased coin are going to be unique "mosaics" of Heads and Tails, but pretty distinguishable with enough flips—and indeed, with the right stats methodology, [MRI scans can predict sex at 96.8% accuracy](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6374327/).
-The same moral applies to sex differences in psychology. I'm a pretty weird guy, in more ways than one. I am not prototypically masculine. Most men are not like me. If I'm allowed to cherry-pick what measurements to take, I can name ways in which I'm more female-typical than male-typical. (For example, I'm _sure_ I'm above the female mean in [Big Five Neuroticism](https://en.wikipedia.org/wiki/Big_Five_personality_traits).) But "weird" represents a much larger space of possibilities than "normal", much as [_nonapples_ are a less cohesive category than _apples_](https://www.lesswrong.com/posts/2mLZiWxWKZyaRgcn7/selling-nonapples).
+Sex differences in the brain are like sex differences in the skeleton: anthropologists can tell female and male skeletons apart (the [pelvis is shaped differently](https://johnhawks.net/explainer/laboratory/sexual-dimorphism-pelvis), for obvious reasons), and [machine-learning models can see very reliable differences that human radiologists can't](/papers/yune_et_al-beyond_human_perception_sexual_dimorphism_in_hand_and_wrist_radiographs.pdf), but neither sex has entire _bones_ that the other doesn't, and the same is true of brain regions. (The evopsych story about complex adaptations being universal-up-to-sex suggests that sex-specific bones or brain regions should be _possible_, but apprently evolution didn't need to go that far. Good news for antisexism!—relatively speaking.)
-If you _sum over_ all of my traits, everything that makes me, _me_—it's going to be a point in the _male_ region of the existing, unremediated, genderspace. In principle, you could define a procedure that maps that point to the female region of configuration space in some appropriately structure-preserving way, to compute my female analogue who is as authentically _me_ as possible while also being authentically female, down to the proportion of gray matter in her posterior lateral orbitofrontal cortex and—the love of a woman for a man.
+Maybe this should just look like supplementary Statistics Details brushed over some basic facts of human existence that everyone knows? I'm a pretty weird guy, in more ways than one. I am not prototypically masculine. Most men are not like me. If I'm allowed to cherry-pick what measurements to take, I can name ways in which my mosaic is more female-typical than male-typical. (For example, I'm _sure_ I'm above the female mean in [Big Five Neuroticism](https://en.wikipedia.org/wiki/Big_Five_personality_traits).) ["[A] weakly negative correlation can be mistaken for a strong positive one with a bit of selective memory."](https://www.lesswrong.com/posts/veN86cBhoe7mBxXLk/categorizing-has-consequences) But "weird" represents a much larger space of possibilities than "normal", much as [_nonapples_ are a less cohesive category than _apples_](https://www.lesswrong.com/posts/2mLZiWxWKZyaRgcn7/selling-nonapples). If you _sum over_ all of my traits, everything that makes me, _me_—it's going to be a point in the _male_ region of the existing, unremediated, genderspace.
-[...]
+Okay, maybe I'm _not_ completely over my teenage religion of psychological sex differences denialism?—that belief still feels uncomfortable to put my weight on. I want to believe that there are women who are relevantly "like me" with respect to some fair (not gerrymandered) metric on personspace. But, um ... it's not completely obvious whether I actually know any? Most of the people in my robot cult (and much more so if you look the core of old-timers from the _Overcoming Bias_ days, rather than the greater Berkeley "community" of today) are male. Most of the people in my open-source programming scene are male. These days, [most of the _women_ in my open-source programming scene are male.](/2017/Aug/interlude-vii/) Am I not supposed to _notice_? I could _assert_ that it's all down to socialization and self-fulfilling prophecies—and I know that _some_ of it is. (Self-fulfilling prophecies [are coordination equilibria](/2020/Jan/book-review-the-origins-of-unfairness/).) But I can't assert _with a straight face_ that all the gaps will vanish after the revolution, because _I've read the literature_ and can tell you several observations about chimps and [congenital adrenal hyperplasia](/images/cah_diffs_table.png) that make that seem _unlikely_.
+
+I want to speculate that the nature of my X factor—the things about my personality that let me write the things I do even though I'm [objectively not that smart](/images/wisc-iii_result.jpg) compared to some of my robot-cult friends—is a pattern of mental illness that could realistically only occur in males.
+
+I was once told by a very smart friend (who, unlike me, is not a religious fantatic), "Boys like games with challenges and points; girls like games with characters and stories."
+
+I said, "I like characters and stories! I think."
+
+He said, "I know, but at the margin, you seem suboptimally far in the challenges and points direction. But that's fine; that's what women are for."
+
+And what evidence could I point to, to show him that he's _bad and wrong_ for saying that, if he's not already religiously required to believe it?
+
+_Alright_. So _in principle_, you could imagine having a PersonApp that maps that point to the female region of configuration space in some appropriately structure-preserving way, to compute my female analogue who is as authentically _me_ as possible while also being authentically female, down to her pelvis shape, and the proportion of gray matter in her posterior lateral orbitofrontal cortex, and—the love of a woman for a man. What is she like, concretely? Do I know how to imagine that?
+
+Or if I can imagine it, can I _describe_ it in this blog post? I am presently sorrowful that [(following John Holt)](https://www.greaterwrong.com/posts/S8ysxzgraSeuBXnpk/rationality-quotes-july-2009/comment/DtyDzN5etD4woXtFM) we all know more than we can say. I have mental models of people, and the models get queried for predictions in the course of planning my social behavior, but I don't have introspective access to the differences between models. It's easier to imagine people in hypothetical situations and say things like, "That doesn't sound like something she'd _do_, but _he_ would" (and be correct), than to say exactly it is about her character and his that generated these predictions (such that [my words would paint a picture in your head](https://www.lesswrong.com/posts/YF9HB6cWCJrDK5pBM/words-as-mental-paintbrush-handles) that would let you make your own predictions about her and him without having met them)—just like how you're better at recognizing someone's face, than at describing their face in words in enough detail for an artist to draw a portrait.
+
+As a _first-order approximation_, I do have a sister. I think the family resemblance between us is stronger than with either parent. We're about equally intelligent. (OK, plausibly she's smarter than me; [the SAT is pretty _g_-loaded](https://www.gwern.net/docs/iq/2004-frey.pdf) and her 1580 (out of 1600) solidly beats my 2180 (on [the out-of-2400 scale used between 2005 and 2016](https://en.wikipedia.org/wiki/SAT#2005_changes,_including_a_new_2400-point_score), such that 2180 proportionally scales down to 1453 out of 1600).) Our dark hair curls into helices with similar radius. We even have similar mannerisms, I think? She's 5′6½″.
+
+But in a lot of ways that matter, we are _very_ different people. When you compare representative outputs of what we've _done_ with our (roughly) similar intelligence—her chemistry Ph.D. from a top-10 university, my _batshit insane_ secret ("secret") blog about the philosophy of science and the etiology of late-onset gender dysphoria in males—it ... paints a different picture.
-Okay. Having supplied just enough language to _start_ to talk about what it would mean to actually become female—is that what I _want_? I mean, if it's reversible, I would definitely be extremely eager to _try_ it ...
+Of course same-sex siblings would _also_ be different pictures. (Identical twins aren't _duplicates_ of each other, either.) But the advantage of having a sister is that it gives my brain's pattern-matching faculties a target to [sight](https://en.wikipedia.org/wiki/Sight_(device)) against. As a _second_-order approximation, my female analogue is close to being somewhere on the vector in personspace between me and my sister. Not exactly on that line, because the line spans both the difference-betwen-siblings and the difference-between-sexes. But the angle between the line between me and my sister, and the line between me and my female analogue, is the [arctangent](https://en.wikipedia.org/wiki/Inverse_trigonometric_functions) of the difference-between-same-sex-siblings and the difference-between-sexes, which is small if sex differences are a lot larger than same-sex sibling differences (with respect to whatever metric on personspace we're using).
+
+(All this is in accordance with "Everything is a vector space" philosophy implied by this blog's [TLD](https://en.wikipedia.org/wiki/Top-level_domain). (If it turns out that something _isn't_ a vector space, I'm not sure I want to know about it.) I can hope that my description of the _methodology_ is valuable, even if your brain's pattern-matching faculties can't follow along with the same example, because you haven't met my sister and only know the aspects of me that shine through to the blog.)
+
+Okay. Having supplied just enough language to _start_ to talk about what it would even mean to actually become female—is that what I _want_? I mean, if it's reversible, I would definitely be extremely eager to _try_ it ...
I had said we're assuming away engineering difficulties in order to make the thought experiment more informative about pure preferences, but let's add one constraint to _force_ the thought experiment to be informative about preferences, and not allow the wishy-washy evasion of "I'm eager to _try_ it."
What if I can't just "try" it? What if the machine can only be used once? Come up with whatever frame story you want for this: maybe the machine costs my life savings just to rent for two minutes, or maybe the transformation process is ever-so-slightly imperfect, such that you can't re-transform someone who's already been transformed once, like a photocopy being a perfectly acceptable substitute for an original document, but photocopies-of-photocopies rapidly losing quality.
-In that case, if I have to choose ... I _don't_ think I want to be Actually Female? I _like_ who I am on the inside, and don't need to change it. So, I guess ... if I'm being honest ... I guess I _want_ to be a normal man wearing a female body like a suit of clothing.
+In that case, if I have to choose ... I _don't_ think I want to be Actually Female? I _like_ who I am on the inside, and don't need to change it. I don't _want_ to stop liking challenges and points as much as I do—and if I don't know enough neuroscience to have an _informed_ preference about the ratio of gray matter in my posterior lateral orbitofrontal cortex, I'm sure it's _probably fine_.
+
+At the same time, the idea of having a female body still seems like _the most appealing thing in the world_. If artificial superintelligence gives me BodyApp for a subjective year and tiles the _rest_ of our future lightcone with paperclips, that's _fine_; I will die _happy_.
+
+So, I guess ...
+
+If I'm being _really_ honest with myself here ...
+
+And I successfully make-believe that I can tell the truth with no consequences on my secret ("secret") blog even though at this point my paper-thin pseudonymity is more like a genre convention rather than providing any real privacy ...
+
+I guess I _want_ to be a normal man wearing a female body like a suit of clothing.
Is that weird? Is that wrong?
What's going on here? The sexologist [James Cantor speculates](https://youtu.be/q3Ub65CwiRI?t=281): mirror neurons.
+[TODO: I don't know enough neuroscience for "mirror neurons" specifically to not be https://www.lesswrong.com/posts/NMoLJuDJEms7Ku9XS/guessing-the-teacher-s-password , _however_, the real answer almost has to be _something_ in the vicinity of "something in the brain is getting confused between sex-target and sense-of-self"; that's _the only way to make sense_ of the "I love her, but also, I want to be her" qualia in otherwise-ordinary males that I feel and _lots_ of trans women self-report (including lots who claim it's an effect of "gender"/"gender identity" rather than a cause)]
-https://www.lesswrong.com/posts/NMoLJuDJEms7Ku9XS/guessing-the-teacher-s-password
+[TODO implication that departs from the standard narrative: being Actually Female would undermine my _reason_ for wanting a female body]
+[TODO implication that departs from the standard narrative: if I could get body-mods _without_ psych effects of HRT &c., that would actually be desirable]
-[in particular, being Actually Female would undermine my _reason_ for wanting a female body]
-[if I could get HRT without the psych effects, that would actually be an improvement]
-[the fact that I'm happy with my breasts is suggestive of body-mods still being positive, even if the desire is a confusion]
+[TODO concern: if the desire is just a confusion in male sexuality, wouldn't that imply that body mods aren't desireable? Like, maybe staight non-trans AGP men _think_ they want to mod female, but if they actually did it, they would get super-dysphoric because (and which proves that) they're perverts and not Actual Trans Women, which are a different thing. You might think so! But, _empirically_, I'm pretty happy with my 5.month-HRT breasts for about the reasons I expected. From what I hear from people braver than me, that experience continues down the slope of interventions]
-(The scintillating but ultimately untrue thought.)
+[TODO: but if you haven't _made_ all these fine mental distinctions using the everything-is-a-vector-space skill, you might interpret the qualia as simply (wanting to) "be a woman", or at least be attached to the idea even if you don't quite believe it]
-[but if you haven't made all these fine mental distinctions, you might think that you want to "be a woman", or at least be attached to the idea even if you don't believe it]
+(The scintillating but ultimately untrue thought.)
[...]
+[Back to Yudkowsky's "Changing Emotions"—]
+
> If I fell asleep and woke up as a true woman—not in body, but in brain—I don't think I'd call her "me". The change is too sharp, if it happens all at once.
In the comments, [I wrote](https://www.greaterwrong.com/posts/QZs4vkC7cbyjL9XA9/changing-emotions/comment/4pttT7gQYLpfqCsNd)—
_Why would I?_ In the English of my youth, "gender" (as a single word, rather than part of the phrase "gender role") was understood as a euphemism for _sex_ for people who were squeamish about the potential ambiguity betweeen _sex_-as-in-biological-sex and _sex_-as-in-intercourse. (Judging by this blog's domain name, I am not immune to this.) In that language, my "gender"—my sex—is male. Not because I'm necessarily happy about it (and I [used to](/2017/Jan/the-erotic-target-location-gift/) be pointedly insistent that I wasn't), but as an observable biological fact that, whatever my pure beautiful sacred self-identity feelings, _I am not delusional about_.
+Okay, so trans people aren't delusional about their [developmental sex](/2019/Sep/terminology-proposal-developmental-sex/); the claim is that their internal sense of their own gender should take precedence.
/2018/Jan/dont-negotiate-with-terrorist-memeplexes/
-
Men who fantasize about being women do not particularly resemble actual women! We just—don't? This seems kind of obvious, really? Telling the difference between fantasy and reality is kind of an important life skill?
projects / Ultimately_Untrue_Thought.git / commitdiff