where s**t = stat

biological sex, gender identity, presentation, and sexuality are all distinct concepts BUT they are reliably better than random guessing to predict one another.

conceptually organizing any of these four into vectors is difficult (even if you assume cis binary people on the kinsey scale, what is the latent space of gender presentation?), BUT it is safe to say they are mutually independent even if highly correlated

let's call this dataset the SGRUB dataset (Sexuality, GendeR and Uhhhhhhh 🅱️resentation)

big questions:

• can we learn a lower-dimensional manifold that surmises this dataset?
• can we identify human-understandable principle components of this dataset?
• how does geometric deep learning even WORK hahaha
• why do none of my fellow ML researchers or LGBT grad students want to help me develop GENDER THEORY 6.0?
• investors: i will incorporate the blockchain and quantum computing into this if you please just fuckening fund me
• okay but fr i think "gender" as a label or feature is a (bad and human-interpreted) reduction from a higher-dimensional feature that could be calculated and recorded and i think that'd be neat.
• i feel like an outlier in the vector space of sgrub
• i feel far from the center of the cluster forming my voronoi cell of "amab nonbinary"
• i've strayed far from the whorfian guideposts and i find myself without words to express my identity and i feel i'm seen as being at odds with my fellow lgbt/queer folks
• and divvying up this linguistic landscape with further words feels like the wrong path
• like applying burn ointment to a cut wound
• or adding mayonnaise to cereal
• no matter how much you add it won't make the cereal better
• just like more words won't make it easier to express how i Feel about my identity
• and it's not that big a deal, it's not so important to me
• it's about as big a deal as someone trying to eat cereal with mayonnaise
• that is, kind of notable, and interesting to discuss, but that's it
• thank u for coming to my ted talk