Most people subconsciously and automatically sort people into two genders. But what happens when a machine is tasked to this? Alyx explores the the human side of artificial intelligence technology, and what it could mean for transgender people.
Through the AI Eye
Mirror, mirror, on my screen, which gender does my face seem?
There are dozens of facial recognition websites where you upload your selfie and tries to guess your gender, age and mood. This is a form of Artificial Intelligence (AI) machine vision, and it’s how the machine on the other end tries to understand who you are.
Sometimes I look in this digital mirror before I go outside, to either reaffirm my current trans DMAB gender presentation, or out of curiosity to see how far machines have come in understanding queer faces.
Most of these mirrors are not perfect. Their AI vision is blind to the ambiguity of nonnormative facial features, because they’ve been fed only a small range of faces to train on, rarely queer people. Some AI train only on celebrity and stock photos, where gender binaries are at extremes,like howhot.io. Others, like how-old.net train on a larger, diverse corpus which give more lucid face results.
Most people subconsciously sort strangers into two genders. For example, when I walk down Canal street in NYC, street hawkers try to sell me either handbags or watches depending on what gender binary I appear closer to that day. It’s an effective barometer to measure how others see me.
I want people to sort me into a gender closer to how I identify, since it makes my life easier (a little towards F but near the middleish [on a 2D plane]). I feel better when I upload my selfies to “How Old,” which I trust it the most. I still can’t get “How Hot” to sort me into my desired gender bucket.
AI facial recognition is here to stay, and will become more ubiquitous and influential. We have to ensure all types of queer people are represented by this sorting technology. This could become a safety issue in the future if a robot misgenders a trans person publicly because of the way light and shadow casts off their face, which is one metric used by AI to understand presented gender.
I wish someone would build a more diverse AI eye. One that reads both binary extremes and the broad spectrum of faces in the middle, so robots can understand fluctuating gender identities. When we upload our selfies and photos with friends, the machine will not only hold our memories but also see the gender tesseracts to understand us.
About Alyx
Alyx Baldwin is the co-founder, CTO of Kip and an award-winning designer / technologist in NYC. They love networks, artificial intelligence and cats.
Humans, Not Robots
Contrary to popular belief, Micah is not a robot. All of these pages are lovingly hand-crafted by one tiny little human. Send some love by donating a cup of coffee a month: patreon.com/neutrois.
genderqueer.me 
I’ve tried a couple of other face recognition sites, including rekognition.com and faceplus.com, on a bunch of photos, with interesting results when compared against each other. Most of the time they identified me as my preferred/presented gender, but also showed me that certain hairstyles (short, and/or extreme colours such as black or blonde) were probably best avoided, and prbably didn’t disguise a masculine face shape and jawline. It was an interesting, and slightly salutary exercise.