How We Came Up With Your Result
The Aryan Recognition Tool was build on top of a model that uses FaceNet, a face recognition system debuted by Google in 2015. Facenet continues to be the most accurate implementation in an open source sense. Multi-task cascaded convolution neural networks (MTCNN) does rapid face detection and face alignment (literally, tilting the position of the face so it’s upright for the computer to see). An improved “loss” function is used to realize face verification and recognition with high accuracy.
One of our inquiries was what would happen if we tried to detect Aryan-ness or Aryan beauty–if this was indeed a conception of true beauty and pedigree human crafted by genocidal leaders…?
How would we go about doing that? Does it mean we can detect who’s more likely to be racist? Jussst kidding.
How we came up with the “percentage Aryan”:
- We scraped the internet for pictures of SS leaders and Nazis, including concentration camp guards and mobile killing squad members. This became part of our “Aryan” data set.
- We found face recognition models that would represent “non-Aryan” faces.
- From each Aryan face in our sample, we created an “embedding”—a mapping of values which represent a particular face. It’s really just like squeezing a ton of messy visual information into several thin straws of numbers.
- We find the distance between your face’s embedding and that of a particular “Aryan” and turn it into a percentage.
By now you can tell there was no scientific basis to any of the above except for the very particular sample set we chose.
The above was done in a hyperbolic way to exemplify how attempts at phrenology through the inhumane usage of technology for surveillance and racial profiling continues today.
Sound familiar?
Faception (circa 2016), a Tel-aviv based startup, purports to be able to detect extroversion and criminality.
