Experts: To Solve AI’s Bias Problem, Hire Fewer White Men

When you can think of a nagging problem, there’s likely a researcher trying to build up an artificial cleverness to solve it somewhere. But while AIs themselves have become more diverse lately, the analysts creating the systems haven’t.

According to a fresh report out of NY University’s AI Now Institute, the AI industry continues to be dominated by white men – which “diversity crisis” is partially to be blamed for the biased AIs we’re viewing crop up around the world.

“To date, the diversity problems of the AI industry and the problems of bias in the systems it builds have tended to be looked at separately,” the research workers wrote. “But we claim that they are two variations of the same problem: issues of discrimination in the labor force and in system building are deeply intertwined.”

The AI Now team discovered that men comprise 82 percent of the authors at leading AI conferences and more than 80 percent of AI professors. The true figures don’t progress when you leave academia, either: women constitute just 15 and ten percent of Facebook and Google’s AI research staffs, respectively.

Black people appear to be even less represented in the AI industry – machine vision researcher Timnit Gebru told the AI Now team she was one of just 6 dark people at a 2016 machine learning conference attended by 8,500 people.

That insufficient gender and racial diversity isn’t just a problem for those looking to work in the AI industry, either – it’s also a problem for society all together.

“The urgency behind this issue is increasing as AI becomes built-into society increasingly,” Stanford researcher Danaë Metaxa, who wasn’t associated with the report, told The Guardian. “Essentially, having less variety in AI is focusing an increasingly massive amount power and capital in the hands of the select subset of individuals.”

The AI Now team’s report includes several recommendations for how AI companies can improve workplace diversity. Included in these are the creation of bonuses for employing underrepresented groups and the publication of comprehensive compensation reports divided by competition and gender.

The team shared suggestions for addressing biased AIs themselves also, such as getting the operational systems audited by indie parties and increasing transparency concerning where in fact the AIs are used.

“Every full day that goes on it gets more challenging to resolve the problem,” Tess Posner, CEO of AI4ALL, a non-profit centered on AI diversity, told The Guardian. “At this time we are in a fascinating instant where we can change lives before we observe how a lot more complicated it can get later.”

The AI Now team uncovered that men comprise 82 percent of the authors at leading AI conferences and more than 80 percent of AI professors. The real figures don’t improvement when you leave academia, either: women constitute just 15 and 10 % of Facebook and Google’s AI research staffs, respectively.

Black people seem to be even less represented in the AI industry – machine vision researcher Timnit Gebru told the AI Now team she was one of just 6 dark people at a 2016 machine learning conference attended by 8,500 people.

That insufficient gender and racial diversity isn’t simply a problem for those seeking to work in the AI industry, either – it’s also a problem for society altogether.