Hacker News Reader

This holds up if you imagine the profiling as done by humans. However, neural networks already surpass human ability in some computer vision tasks. The features the trained models come up with can go way past what we might consider visual likeness. The algorithmic method might eventually become 99.999% efficient at racial profiling with the help of something as ridiculous as fingerprinting eyebrow strand distribution & orientations. The right way forward is to fully consider the ethical implications when shaping these policies, regardless of technological ability.

> 99.999% efficient at racial profiling

Next on HN: Falsehoods Programmers Believe About Race

State face recognition doesn't have to be any more accurate than other forms of institutional, armed, racism.

Oh, just correlate with a heuristic for the user's first and last name and their browser's language preference headers to improve the matching in case the face recognition algo is unsure.