Google, along with other photo management programs, has a facial recognition feature that helps you store your digitized pictures by person. In my case, for instance, any picture that has Mary Ann in it can be found by simply clicking on her name. I’m not sure how it does it, but it recognizes photos of me as a child pretty well, although not so good on baby pictures. It really is fantastic and has amazing accuracy, especially if you have a lot of photos of that person. The more photos, the better the recognition.
As you know, I am a white male of Irish lineage. Turns out that white males are the primary subjects of most facial recognition programs, so the accuracy is pretty good. For people of color, the recognition isn’t as good yet. For females, even worse. For African-American females, it is the worst. It is a work in progress and will get better over time.
Now I’ve ranted in the past about license plate readers used by the cops to collect taxes. I don’t care much for the cops being able to spot my possibly expired tag from 500 feet away as I’m going the opposite way at 70 miles per hour, but cameras and computer software are getting pretty accurate, even with speed and distance factored in. So any such use of sophisticated technology is inevitably going to benefit society if the police can effectively do their jobs better, right?
Yet a number of California cities are banning their police departments from using facial recognition software. The innovative idea is that the police body cams could be equipped with the software and help an officer have knowledge of the person they are dealing with. As an officer, wouldn’t you want to know that the person you are facing is currently wanted for murdering an officer? Of course you would. But… what if the software is wrong? What if the person you are facing is John Citizen, not Bob “The Killer” Jones? Your heightened sense of alarm causes you to employ defensive tactics that result in the death of Mr. Citizen. Tragic, right? Especially when it comes out that the software has a significant error rate on African-American males, which Mr. Citizen happens to be.
Tech companies, including behemoth Microsoft, admit that facial recognition software is not ready for reliance by police agencies, no matter how many television shows claim otherwise. I recently saw a show where they took a grainy surveillance photo of a man’s profile and “enhanced” it such that it was the spitting image of the guy. Sure. On the bad spectrum, Google Photos once tagged two African-Americans as “gorillas,” and the Chinese use street surveillance cameras to single out Muslims for mass roundups. On the good spectrum, a facial recognition photo helped catch a mass shooter in Washington, D.C., and helped Indian authorities find 3,000 missing children in just four days. As with most anything, the software has good and bad uses.
So in the end, one must weigh the benefits of the good against the costs of the bad. For now, I concur that law enforcement needs to be restrained in its use of facial recognition software except for limited investigatory purposes.