Even Artificial Intelligence Is Sexist – Glamour

Sexism is so deeply ingrained in the way we think about the world, we've actually passed it on to our computers, according to a new University of Virginia (UVA) and University of Washington study . Artificial intelligence (AI) is more likely to label people who are cooking, shopping, and cleaning as women and people who are playing sports, coaching, and shooting as men.

UVA computer science professor Vicente Ordez got the idea for the experiment when he noticed that his image-recognition software was associating photos of kitchens with women. After training software using two photo collections that researchers use to create image-recognition software, including one supported by Facebook and Microsoft, he and his colleagues found that not only do these collections contain gender bias, but they also multiply that bias when they pass it on to the software. The program these photo sets produced actually labeled a man a "woman" because he was standing by a stove.

This isn't the only evidence we have that technology contains biases. In addition to image-recognition software, software that analyzes writing and speech also reflects hidden assumptions about gender, according to a study published earlier this year in Science. The researchers analyzed how computers interpreted words from Google News and a 840 billion-word data set used by computer scientists, and they found that machines linked male and man with STEM fields and woman and female with chores. The problem wasn't just with gender either: Stereotypically white names were more likely to be associated positive words like happy and gift. Another study published last summer found that when software based on Google News was asked, "Man is to computer programmer as woman is to X, " it responded with "homemaker."

Of course, computers don't make up these associations out of nowhere. They're reflecting our own biases back to us. But when they pick those beliefs up, these can take on a life of their own. The snafu that Google's image software made in 2015, when it mislabeled black people as gorillas , demonstrates this. Google image searches are another example: Search for hand and you get mostly white ones, while girl yields sexy photos and boy yields kids.

This tendency becomes even more problematic when AI is used to create robots that interact with people. Mark Yatskar, a researcher at the Allen Institute for Artificial Intelligence and an author of the new study, told Wired he could imagine a scenario where a robot asks a woman if she wants help with the dishes while handing a man a beer. "This could work to not only reinforce existing social biases but actually make them worse," he said.

The way artificial intelligence identifies words and images is based on the way people use them, so in order to promote a more egalitarian world, engineers would have to intervene in the creation of the software. And that's a possibility many are considering. Eric Horvitz, director of Microsoft Research, told Wired that Microsoft has a committee for this. "I and Microsoft as a whole celebrate efforts identifying and addressing bias and gaps in data sets and systems created out of them," he said. "Its a really important questionwhen should we change reality to make our systems perform in an aspirational way?"

Follow this link:

Even Artificial Intelligence Is Sexist - Glamour