Why would we NEED to face scan people? Why the disparity? Are they announcing they INTEND to pardon melanin defense? I remember hearing this on news bumper once or twice, then blip memory holed.
Comments (2)
sorted by:
Why would we NEED to face scan people? Why the disparity? Are they announcing they INTEND to pardon melanin defense? I remember hearing this on news bumper once or twice, then blip memory holed.
I have worked with facial recognition as part of my career, so I can answer some questions. . .
In almost all use cases, the algorithms the make up facial recognition focus on picking out various features of the face and comparing the distance and ratios between them. It is easier for the algorithm to pick out these features on a bright image than a dark image.
The facial features of a black person are just less visible on most cameras, especially lower resolution, low quality cameras that would be used for security/surveillance purposes.
Less distinguishable/visible features means that there is less data for the facial recognition algorithm to work with. Therefore, it is significantly less effective at recognizing/distinguishing black people from other black people.
Almost all facial recognition software uses machine learning to either derive or tune the underlying algorithms. The accuracy of these algorithms directly depends on the quantity, quality, and variety of data presented to it for "training". Thus far, this training is usually done using datasets built for research and proof-of-concept purposes, which often includes more easily recognizable (white) faces than those not as easily recognizable. Also, the provided images are usually from volunteers, which, for whatever reason, tend to be white.
Now, there is definitely room to improve these algorithms and come up with facial recognition technology that works to identify everyone without disparity, but people are scared of this- and rightfully so in most cases.
If you want to look at it from a law enforcement perspective. . .it's a damned if you do, damned if you don't situation.
How many times has a criminal walked free because someone didn't recognize them? In almost all cases, the ability of law enforcement to identify an individual is a subjective process. Facial recognition would be a tool that provides objective results.
For example, I could take a few frames from the security footage of a bank being robbed and compare them to the faces of people in a lineup. Instead of relying on subjective witness testimony that is prone to error and emotions, I could simply use a facial recognition tool to compare them. With proper vetting/certification of the software, it would be damn near impossible to question that identification in court
However, it could be used by corrupt governments or organizations to further their own agendas. . .imagine, you appeared on camera during the rally in DC, well, you were immediately identified and put on the list. . . that is where people should be scared.
Mass surveillance really hasn't been a thing, because there is not enough people watch everything 24/7. . .however, if you turned that over to AI. . .manpower is no longer a problem.