(no title)
eldavojohn | 3 years ago
Use race as a dimension for something and that ends up as a value in a vector that packs a human into a discreet set of pigeonholes. Then take many of those and stack them and you've got a matrix ready for things like principal component analysis or CNN training.
You might say "oh come on, that hasn't been done since WWII by IBM" and you'd be wrong. It still happens today with things like calculating insurance premiums and approving bank loans. And your response might be "no way, nobody records someone's race" and while that might be technically correct, we frequently harvest things like income and interest in products that are highly correlated with spacefic races (some innocuous others much less innocuous). This can be harvested through cookies in websites like facebook or they can be self reported income on credit card applications.
You can disagree that it's the same as saying "matrix multiplication is racist" but that is just a boiled down way of saying "we are very good at hiding racism in our algorithms and then acting super surprised when someone points them out and our defense is that we just did some math."
orangecat|3 years ago
Yes, and it is not at all clear that this should be considered racism.