4 ms·

Face recognition will solve that for you.

andybak 8y ago
Good luck guessing ethnicity for Iran. There will be many who look less middle-eastern than your average "Californian with a tan". You'd have the same problem with most near-east or Mediterranean countries. Being at the crossroads for a few millennia will do that to a place.
- some1else 8y ago
  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.
  - pjc50 8y ago
    > 99.999% efficient at racial profiling Next on HN: Falsehoods Programmers Believe About Race
- purerandomness 8y ago
  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.
- pjc50 8y ago
  State face recognition doesn't have to be any more accurate than other forms of institutional, armed, racism.
paganel 8y ago
Or just plain name checking. One of the personal projects of mine was to put on a map the nationalized buildings that used to be owned by Jewish owners (I live in a former Easter-European communist country, we did nationalize a lot of stuff immediately after WW2 ended). That was pretty simple: I just got the nationalization order/law from back in the day (it included the names of the building owners and the addresses), I inputed it in a database and then I matched those owners' names against a public database containing only jewish names, meaning this one [1]. It was a lot more difficult to exactly geo-locate those buildings on a map (a lot of street names had changed in the last 70 years) than it was to guess the ethnicity of a lot of people on that list. Again, this was part of a personal project that I had built in order to better know the history of the local Jewish community, but I could see the same "technique" (meaning just plain name matching against a known database with positive ethnic identification) being used for nefarious reasons in other circumstances and places. [1] https://www.yadvashem.org/ https://www.yadvashem.org/

Good luck guessing ethnicity for Iran. There will be many who look less middle-eastern than your average "Californian with a tan". You'd have the same problem with most near-east or Mediterranean countries. Being at the crossroads for a few millennia will do that to a place.

some1else 8y ago
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.
- pjc50 8y ago
  > 99.999% efficient at racial profiling Next on HN: Falsehoods Programmers Believe About Race
purerandomness 8y ago
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.
pjc50 8y ago
State face recognition doesn't have to be any more accurate than other forms of institutional, armed, racism.

paganel 8y ago

Or just plain name checking. One of the personal projects of mine was to put on a map the nationalized buildings that used to be owned by Jewish owners (I live in a former Easter-European communist country, we did nationalize a lot of stuff immediately after WW2 ended). That was pretty simple: I just got the nationalization order/law from back in the day (it included the names of the building owners and the addresses), I inputed it in a database and then I matched those owners' names against a public database containing only jewish names, meaning this one [1]. It was a lot more difficult to exactly geo-locate those buildings on a map (a lot of street names had changed in the last 70 years) than it was to guess the ethnicity of a lot of people on that list. Again, this was part of a personal project that I had built in order to better know the history of the local Jewish community, but I could see the same "technique" (meaning just plain name matching against a known database with positive ethnic identification) being used for nefarious reasons in other circumstances and places. [1] https://www.yadvashem.org/ https://www.yadvashem.org/