A lot of our presentations on algorithms had to deal with the negative uses of them. While doing research for my groups topic, risk assessment, the same idea kept coming up with how our personal biases can be unconsciously placed in the algorithms we create. I’d like to think that some algorithmic insensitivity Is purely by accident such as the relation between typing “gorilla” into Google and receiving imagery of a black software engineer at Google in 2015. Nobles writes about this in our reading, from this week, when she talks about this specific incident with Google’s auto-tagging and facial recognition software (Nobles 22).
I was curious what, if any, positive response Google has done in the last three years. Sadly, it has not been very substantial and from my minimal research it looks like all that Google has done is removed any imagery tagging of primates (specifically gorillas, chimps and monkey. The terms are now unable to be used in searches and as such have in a way become nonexistent to the AI that controls Google’s image searching abilities.
In January of this year, WIRED did a test seeing how Google Photo’s algorithms would label a collection of 40,000 images of animal stock photography. WIRED reports that the test proved that Google Photos is quite skilled in finding and labeling most of the creatures but that the program would respond with “no results” when searching with the terms of “gorilla,” “monkey,” “chimp”, and “chimpanzee.” Google’s imaging programs use machine-learning to automatically tag images into group with what the program finds in the composition. The program did however recognize “baboon,” “gibbon,” and “marmoset” just as long as “monkey” was not use in the search as well (Simonite 2018). In another test of 20 photos of chimps and gorillas gathered from primate nonprofits and using Google Photos some of the images could be found when using terms such as “forest,” “jungle” or “zoo.”
Google has confirmed that “gorilla” had been banned after 2015 and that “monkey,” and “chimp” are also banned as of now (Simonite 2018). A spokesperson for Google gives the reasoning that “Image labeling technology is still early and unfortunately it’s nowhere near perfect.” Thus, it is just easier for them to ban these tags. The fact that Google has refused to fix the problem and just erase the issue all together is incredibly problematic since this could have been a learning experience to fix the mistake instead of erasing it.
Sources
Noble, Safiya Umoja. Algorithms of Oppression: How Search Engines Reinforce Racism. New York University Press, 2018.
Simonite, Tom. “When It Comes to Gorillas, Google Photos Remains Blind.” Wired, Conde Nast, 18 Jan. 2018, www.wired.com/story/when-it-comes-to-gorillas-google-photos-remains-blind/.