Read on Twitter
Of course computers cant be racist, but they also can’t do anything with initialization from a user. In the case of facial recognition tech, the algorithms that form the basis of the technology were not designed with representative population data in mind.
The algorithms across the board are significantly less accurate when analyzing people with dark skin. The consensus in the field is that facial recognition is by far the least accurate bioinformatic technology. It consistently produces false matches.
A technology doesn’t exist without an implementation of it. In Detroit, there is significant racial bias in the implementation of facial recognition tech when census data is considered. As an aside, these sets remind me of “redlining” maps.
The question is then, “Why would a city implement an unreliable technology meant to partially automate the policing of the populations that it is the worst at identifying?”. This question becomes rhetorical upon even a brief consideration of the history of policing in the USA.
In summary, technology doesn’t exist without implementation. If the implementation is racist, then yes, the technology is currently racist. This is a drop in the bucket of the problems with policing, but facial recognition is insidious because its proprietors present it as fair.
