Facial recognition is the plutonium of AI
Plutonium, in other words, is an apt material metaphor for digital facial recognition technologies: Something to be recognized as anathema to the health of human society, and heavily restricted as a result.
Our faces are fundamental to our humanity—something we cannot choose, nor change. Something by which we are recognized and known to others. An aspect of our selves we cannot change. Pervasive deployment of facial recognition technology is an assault on individual freedoms and privacy, and can only be used to reinforce outdated morés and attitudes.
Stark (who works for Microsoft's Research Group) makes a forceful argument that numerical analysis of our faces, something so fundamental to our selves, ◊strong{should not be done}, and that such analysis ◊em{inherently} reinforces racist classifications of humanity, regardless of the intentions of the system's designers or implementors. I think, for ACM readers, there is an extra coddling needed to introduce the concept of privilege in a way palatable to white male engineer readership who believe they are "above" such concepts and believe they exist in a postracial utopia. Stark does an excellent job.
The article does mention at the end possible ◊em{medical} applications, such as a facial recognition device for those with #prosopagnosia (aka "face-blindness"). Just as there are medical applications for nuclear energy.
As a lighter note, however, I wonder about the application of facial recognition for photography—when done locally and in private as Apple's photo management software does. The silly, human reason that I want this is that it's a convenient way to sort a lot of photos quickly. My ability to produce mass quantities of photographs has far outstripped my ability to sort and organize them, and I appreciate the help. Unfortunately in this networked age, "local" deployment of Facial Recognition tech is just a small bit of scripting glue away from "Facial Recognition as a Service"…
#◊site/notes