Raw LLM Responses

I assume this video-essay was at least in part prompted by the arrest of Robert Williams by the Detroit Police Department in March 2019 after the security guard at a store indicated that a picture which was included in a photo line-up corresponded to the thief. _ The picture was included by police using software provided by a company called Dataworks Plus, when they searched public sources for faces (because that's how the software works) similar to that of the man who appeared in the stores security camera footage. That's the important part. *Similar* . _ It found pictures in his social media. It wasn't (for example) confining its search to police databases because it is not limited to those. _ Mister Williams did not rob the store and was nowhere near it at the time of the offence. The security guard misidentified his picture as the offender five months after the offence took place in October 2018. - The police chief talked about the error at a press conference 6 days ago, going on to state that the software appeared to have an error rate of 96%. _ From other sources I am aware that the image software used by Google is unable to recognize a difference between images depicting human subjects and images depicting non-human subjects on some occasions. Specifically it misidentifies people of African descent as gorillas. _ Facial recognition software used by Microsoft, IBM and Face++ routinely misidentifies the gender of the subject in all groups except light skinned males, with darker skinned women being misidentified as males as much as 35% of the time (third link). _ https://arstechnica.com/tech-policy/2020/06/detroit-police-chief-admits-facial-recognition-is-wrong-96-of-the-time/ https://www.theverge.com/2020/6/24/21301759/facial-recognition-detroit-police-wrongful-arrest-robert-williams-artificial-intelligence https://www.nytimes.com/2018/02/09/technology/facial-recognition-race-artificial-intelligence.html