Google has issued a fix for an error that put its image recognition software at the center of a heated controversy about racism in 2015. However, the fix is not really teaching the algorithm any lesson in the diversity of the human race.
The image recognition algorithm, which is at the core of the Google Photos app, came under fire for categorizing an image of a Black couple as ‘Gorillas’. Back then, the company apologized for the blunder, but even after three years, the fix does not really fix anything.
As Wired reports, the company had been working on a long-term fix. However, instead of teaching its algorithm what differentiates gorillas from people of color, Google simply removed the reference to Gorillas from its algorithm. To test, the publication used a collection of 40,000 images of animals, and the algorithm reported no results for search terms including ‘gorilla’, ‘chimp’, ‘chimpanzee’, and ‘monkey’.
In a separate test, it was found that the algorithm did not serve any results to the term ‘African American’. For terms such as ‘black man’, ‘black woman’ or ‘black person’, the results were only black and white colored images.
This workaround is clearly not a solution for a technology that is fast becoming the core of a range of services other than the humble Google Photos app. Image recognition technology is set to power most of the artificially intelligent operations such as self-driving cars, voice-controlled bots, and more.
In a statement issued to the publication, Google representative confirmed that the term ‘gorilla’ was removed from the search and image tags after the 2015 incident. The company representative further said that the image labeling technology is still at a nascent stage and is far from perfect.
Meanwhile, the Google Arts & Culture app is taking the social media by storm for its fun face matching technology that is keeping people busy by matching their selfies to famous museum portraits. The development is ironic given how the existing algorithm is unable to differentiate between humans and animals, and yet the technology is improving so as to perfectly find doppelgängers.