Google Photos, a new service that launched last month, is Google’s effort to “unbundle” Photos from Google+ and turn it into a standalone service. One of the new features is auto-tagging with the use of advanced facial recognition algorithms. Computers aren’t perfect however.
Google Racism Discovered
Programmer Jacky Alciné noticed that the Photos app was tagging black people as gorillas, and shared the tagging results on Twitter:
Like I understand HOW this happens; the problem is moreso on the WHY. This is how you determine someone's target market.
— Jacky Alciné (@jackyalcine) June 29, 2015
To Google’s credit they responded very quickly. Chief Social Architect Yonathan Zunger replied to Alciné’s tweet, saying:
@jackyalcine Holy fuck. G+ CA here. No, this is not how you determine someone's target market. This is 100% Not OK.
— Yonatan Zunger (@yonatanzunger) June 29, 2015
Soon after, Zunger said that the entire gorilla label was removed from the Photos database. It’s only a temporary solution as they try to further refine the recognition technology. Apparently this isn’t the first time something like this has happened, as previously there was a bug where people of all races were being tagged as dogs.
Google has been developing artificial intelligence, machine learning and facial recognition for a long time, but even they can make mistakes. Google’s AI project, simply called Brain, was started in 2011 as a way to “crack the problem of artificial intelligence”. The Brain has improved since then, and has even trained itself to recognize a cat by watching 10 million YouTube videos of our feline friends.
By definition, a facial recognition system is “a computer application for automatically identifying or verifying a person from a digital image or a video frame from a video source.” One way to achieve this is by comparing a standard set of facial features from a database to select photos that are analyzed.
One way to do this is by building an artificial neural network with back-propagation algorithms, which is used as a way to train artificial neural networks. The ultimate goal is to find the best function that maps a set of inputs to its correct output. An example is to provide an image of an animal as the input, and the correct output would be the name of that animal.
Although Homo sapiens and the Great Apes do share an evolutionary common ancestor, I’m not sure if Charles Darwin ever dreamed of such an occurrence.