Google apologises for tagging blacks as 'gorillas'
ADV
ADV
Google apologises for tagging blacks as 'gorillas'

Google apologises for tagging blacks as 'gorillas'

di lettura
(AGI) Rome, July 2 - The year 2015 will certainly not beremembered as one of the best for Google. After the GoogleGlass fiasco, the Mountain View company finds itself in a newfiasco with its Google Photo algorithms, which appears to haveproblems in differentiating people of African origins fromgorillas. The first incident occurred two days ago, followingthe complaints of a user, Jacky Alcine, who published onTwitter a screenshot showing the bizarre way Google Photoorganised its photos according to categories, i.e. skyscrapers,airplanes, automobiles and bicycles. So far so good. Then aphoto of the user, a black man, appeared with his friend, blackas well, with a tag saying 'gorilla'. "Google Photos, y'all(expletive) up. My friend's not a gorilla" was the text of thetweet. It was followed by many others in which Alcine, adeveloper by profession, proved that by writing 'gorilla' inthe search function of her library, photos of the man with hisfriend continued to appear. Two hours later, they were includedin a conversation with Yonatan Zunger, Chief Architect ofSocial Media for Google. His team managed to find a solution tothe problem in about an hour. The following morning, however,two more photos of Alcine and her friend appeared tagged as'gorillas'. Zunger could only remove the 'gorilla' tags fromthe Google Photo database while searching for a permanentsolution to the problem. The company apologised to the user andpromised to look into the issue more thoroughly. "There isstill clearly a lot of work to do with automatic imagelabelling, and we're looking at how we can prevent these typesof mistakes from happening in the future," said a Googlespokesperson. According to the Wall Street Journal, theincident is related to the bad functioning of the Google Photoalgorithms of facial recognition and the need to be moreaccurate. "We need to fundamentally change machine learningsystems to feed in more context so they can understand culturalsensitivities that are important to humans," said Babak Hodjat,chief scientist at Sentient Technologies, anartificial-intelligence company. "Humans are very sensitive andzoom in on certain differences that are important to usculturally," Hodjat told the Wall Street Journal. "Machinescannot do that. They can't zoom in and understand this type ofcontext.". .
ADV