A lot has been made of facial recognition in the last few years. It has been used in various ways to spot people in crowds, to identify users of phones to let them in the phone, to determine who people are in pictures on Facebook, etc.

The current reports are that the programs that do facial recognition are apparently not real reliable on identifying people and quite often mis-identify people.

One study said they showed a facial recognition program that was supposed to identify people who had mug shots and were wanted for some crime the photos of members of the USA Congress. Twenty-eight member of congress were identified as matching mug shots. The stories said these were all misidentification and a high number of wrong identifications. Of course, there are a number of people that would comment it probably really was the opposite and the program failed to match a number of Congress people to mug shots. In any event, it failed to make correct identifications.

Apparently one issue is who all the program has been trained to identify and then identifying similar type people. Apparently one of the ways Facebook has worked on getting more accuracy is by publishing posts that go viral wanting people to post pictures of themselves, like this spring wanting a high school picture and current picture of you.

Facial recognition works by using Artificial Intelligence (AI) and getting the machines to teach themselves as they look at large quantities of pictures of the same people to find ways to make unique.

Dwight