AI will reproduce and enshrine age-old biases—if we let it

Repeating the same (mostly white, mostly male) mistakes of the past.
Repeating the same (mostly white, mostly male) mistakes of the past.
Image: Reuters/ Eddie Keogh
We may earn a commission from links on this page.

We shouldn’t be worried only about whether robots will take our jobs, but about who is programming them—and with what values.

Genevieve Bell is a cultural anthropologist who’s spent the last two decades pondering the intersection of technology and culture, specifically focusing on the ethics of AI in her work at Intel.

“I think the gravest dangers are we take the world we live in now and make it the world in perpetuity moving forward,” Bell told me. “All the things about the current world that don’t feel right is what the data reflects, where women aren’t paid as much as men, where certain populations are subject to more violence, where we know that certain decisions get made in manners that are profoundly unfair. If you take all the data about the way the world has been and that’s what you build the machinery on top of, then you get this world as our total future. I don’t know about you, but I’d like something slightly different.”

Examples of how technology has failed women play out every day, leaving the tech industry in danger of continuing to repeat the same wrongs. For example, Bell pointed out that the microphone I was using to record our conversation was originally optimized to pick up male voices, leading women to sound distorted. Facial recognition technology works almost perfectly for light-skinned men but has extremely high error rates for darker-skinned women.

In one startling Reuters story, Amazon recruiting technology was unintentionally programmed to reject the résumés of female candidates. The AI in the company’s résumé-sorting program had “learned” from the past data that male candidates were preferable, simply because most of the résumés submitted came from men. It penalized résumés containing the word “women’s,” including those from candidates who had attended women’s colleges. Human resources executives have high hopes for AI to take bias out of the hiring process; the problem is, the AI itself is already biased.

The lesson is clear: Technology has the power to reproduce and enshrine age-old inequities. Or, if the population building the technology better represents the population it aims to serve—not just by gender but also by age, race, sexuality, education, class, et cetera—the optimists say it also has the power to challenge inequality.

Sam Altman, president of the seed incubator YCombinator, now spends half his time considering AI’s impact on society, through the nonprofit OpenAI, which he co-founded with a group of technologists including Elon Musk. Musk has warned of a sort of AI apocalypse, believing that superintelligent computers are one of the scariest threats to human civilization. Altman puts it more diplomatically.

“We need to think right now about how we want this deployed, how we are going to govern it, how we are going to make it safe and good for humanity,” he explains. “One of my hopes for AI is that it will help us amplify our best and stop our worst impulses. We have a lot of known psychological flaws. We have deep inequities in the world…The injustices, I think we will be able to address.”

The pessimists aren’t as generous. Former Googler Yonatan Zunger put it ominously, tweeting:

The vast majority of these “young engineers” are male. But women and men have different views and worries about emerging technologies. According to a study conducted by Intel, when it comes to AI, women are much more concerned than men about privacy, data, and security.

“We’ve just come through a year of women talking pretty explicitly about all the ways they have felt like they were on display, and visible and taken advantage of,” Genevieve Bell points out. Like me, Bell suspects Facebook and Twitter might be very different, and perhaps more civil, platforms had a diverse group of women had input in product decisions early on. And we are reminded every day that civil is one thing social media is not.

At the height of the Brett Kavanaugh hearings, social networks were again weaponized to hurl hateful messages on both sides. Both Ford and Kavanaugh reported they were harassed online and received death threats, and vicious threats against their families. And yet there were glimmers of technology also delivering unity and hope.

Ford might never have been granted her hearing if the tech-powered #MeToo movement had not been fresh, if women across the country had not felt the increasing sense of connectedness provided by social networks. One lawyer, Jeanne Christensen, who has represented sexual assault victims in lawsuits against tech companies like Uber, told me she expects many more women to come forward as a result of the bravery of Dr. Ford and so many others. “The sexual harassment that the whole world seems to be hearing about just now has been going on for years. The difference is that people are actually paying attention,” Christensen said. “I think that we are on the tip of an iceberg.”

Excerpted from BROTOPIA: Breaking Up the Boys’ Club of Silicon Valley by Emily Chang with permission of Portfolio, an imprint of Penguin Publishing Group, a division of Penguin Random House LLC. Copyright © Emily Chang, 2019.