Raw LLM Responses

__Im sorry that I’m copying my comment from another article but it’s relevant here Facial recognition AI poses more danger to African Americans. Why? Because the images of black people used to train these systems rely **disproportionately* on pictures of mugshots.** In several case studies, these bots have higher false positives for finding African American criminals. That is, given a mugshot of a criminal then deployed on a CCTV cam, the bots will be more likely to tag the wrong black person since the model *overfits* (i.e becomes lazy or “memorizes”, as we say in machine learning, to specific features) African Americans to images of mugshots. Hence, brown skin+ certain hair texture/facial features = suspected criminal. Contrasted with whites, often pictured in varying scenes, these bots, when used by police departments have much less false positives. Source: I’ve thoroughly researched this issue, even wrote a whole damn paper on it. Main source , for anyone who cares https://www.perpetuallineup.org Highlights: >**One in two American adults is in a law enforcement face recognition network.** > Face recognition may be least accurate for those it is most likely to affect: African Americans.