WomenOfAI Summit 2020: Promoting Diversity & Inclusion in AI
If you were to ask your daughter about AI, what do you think she’d say? Not too long ago, AI was just a thing of science fiction, but today, AI’s real-world applications range from self-driving cars and search algorithms to creative art and medical diagnosis. But one of AI’s current problems is that it is only as strong as the people who are building it. A lack of diversity in AI has led to flawed and biased systems that perpetuate class and gender imbalances within AI applications.
This problem is what led Erica Lee to create WomenOfAI.org, a global nonprofit with 10,000 members that is focused on building up women to tackle AI’s diversity problem. On May 30th, 2020, WomenOfAI.org hosted its first AI Summit, with speakers from Andreessen Horowitz, Zillow, Salesforce, Google TensorFlow, SamsungNEXT, XOKind, Lean In, Intel, and Women in Product to discuss some of AI’s applications and how to increase diversity and inclusion in the industry.
Technical Sessions
The panel began with Andressen Horowitz partner Frank Chen discussing some of the modern applications for AI in industries such as banking, logistics, and medicine. Chen showcased a handful of their portfolio companies that are leveraging machine learning to solve problems that were previously unsolvable under regular conditions, such as the growing network of AI logistics solutions including DeepMap, Applied Institution, and Waymo, along with Zipline, a drone delivery program that delivers medical supplies to doctors in Rwanda who would otherwise have to travel to a single location in order to receive blood samples in order to treat their patients.
His final example was the Crisis Text Line, a nonprofit that utilizes machine learning to assist crisis counselors by analyzing patterns in language that are most indicative of a real crisis to improve the accuracy of their response.
After Frank’s session, the next technical talk was by the nonprofit founder, Erica Lee, who also serves as the Director of Machine Learning Engineering at Zillow. Erica gave a special lunch session on her work at Zillow Offers, the business line within Zillow Group focused on buying and selling homes. Erica went through the user experience of what it’s like to buy and sell a home, with an overview of the machine learning applications and infrastructure needed to deploy ML at scale for the company.
During the lunch break, two hiring managers from hosted a Meet-N-Greet to discuss their company cultures and open roles at their organizations. Afterwards, XOKind CTO & Co-Founder Yinyin Liu shared how their emphasis on user experiences led to the name of the organization, which focuses on building an “extraordinary kind” of AI that is focused on the human elements, also reminiscent of “XOXO” - or hugs and kisses. Currently, they are working on using AI to improve people’s quality of life by serving up users with more relevant travel options and eliminating the need to switch from platform to platform in order to make their leisure travel plans.
Abigail Hing Wen (Senior Director of AI at Intel & Bestselling Author) talked about some of the creative applications of AI, including the use of computer vision in filmmaking, as well as other storytelling and worldbuilding tools that are being developed to help authors and other artists navigate through the difficult process of production.
Yujie Li, Applied Scientist at Zillow, talked about how their company is using AI and computer vision to create an interactive 3D map for house buyers to view when they are looking to buy a home on the website. Yujie explained that Zillow’s technology is deployed in a free iPhone app that walks landlords through the process of creating a series of panoramic images in order to create these maps. They’re also incorporating a product called Zillow Floorplans into the app which complements the 3D tours by outlining the dimensions and layout of the house with complex classification algorithms.
Many of the speakers emphasized the importance of building a strong team in order to build and distribute these applications. Salesforce Senior Director Sarah Aerni explained that it’s important to have a diverse and multi-talented team that’s building these solutions, and that it’s equally important to communicate and continue to ask questions to avoid falling into traps that result in a negative experience for the people who are using them.
Diversity and Inclusion Sessions
WomenOfAI.org COO Jamie Lee led a panel featuring Women of Product cofounder Yardley Pohl, SamsungNEXT Head of D&I Jiun Kimm, as well as Raena Saddler, VP of People Operations at Lean In. Together, they discussed many of the difficulties faced by women in tech, as well as the importance of addressing the issues with diversity in the workplace.
They began by discussing the importance of having diversity and inclusion as core values in your organization, starting with your hiring processes. Each of the women discussed situations where they were faced with discrimination throughout their career, and shared strategies about how women can empower themselves to overcome a toxic work environment.
“With these negative remarks, how you react to them makes an impact. I always try to assume there is positive intent, and maybe ignorance. I try not to go into a defensive mode. Look at it from an educational perspective, make it a conversation, instead of an attack.”
One of the key themes of their discussion was the importance of having access to education and how to address the root causes behind bias and loss of agency. Many of the women in the group shared personal situations where they had been discriminated against and passed up for career opportunities.
Saddler discussed how gender bias starts becoming apparent for girls in middle school as they begin to lose confidence and participate less in class, but that it starts even earlier than that, and it’s up to parents to raise children to be strong enough to overcome it.
“There’s a lot of data that shows black girls remain confident throughout childhood because black parents tell their children that it’s going to be harder for them, so when they face those barriers, they’re ready for them.” —Raena Saddler
“I’m a mom of two young boys, so I have been very mindful [of the toys that my sons get]. They have tea-sets, they have dolls. They get girls’ toys and boy toys. I want to make sure that I’m raising my boys to respect women and understand their perspective as much as possible. As moms of boys, we have a responsibility to raise boys to become men who can be allies and understand the female perspective.” — Yardley Pohl
Another key topic was around imposter syndrome, and how difficult it is for women to feel like they’re respected or that their voice is being heard. The panel discussed how Imposter syndrome is something that will never go away, but when someone asks you, it’s often difficult for to find actual evidence or data that backs it up
Saddler introduced how Lean In is helping to address these problems through programs such as the 50 Ways to Fight Bias, a free program for allies to learn how they can address gender bias in common workplace situations, and Lean In Circle, a women’s support network.
“AI is a very male-dominated field, and it can feel isolating...but it’s also more common to feel that way than you think,” said Saddler.
The group also discussed how important it is for women to have people in a position of leadership as supporters, role models, and advocates.
Hing Wen later discussed the importance of learning from different cultures, and how reading fiction from other cultures can help you have empathy for people from a different background. Her book, Loveboat Taipei, is soon being turned into a film, and she has seen a lot of positive change that has already started to happen as a result of producers who are eager to share stories from more diverse backgrounds as well as the massive success of films like Crazy Rich Asians.
Laurence Moroney also addressed some of the problems that come up in machine learning as a result of having an insufficient amount of diversity in the industry. As the instructor of Coursera’s TensorFlow, Laurence has noticed a clear correlation with the viewership of his technical videos, where the more advanced the video, the fewer women are watching them, regardless of whether the presenter is a man or a woman.
Moroney further emphasized the importance of introducing more women and people from diverse backgrounds in order to architect these unbiased solutions. He gave the example of an AI that fails to interpret an African wedding from a row of wedding photos because it falls outside of what Westerners would typically identify as a wedding.
Laurence also shared his personal story of discrimination from being a young Irish man growing up in a culture of racism in the UK. While high demand for his technical expertise enabled him to secure work, he was glad when people started to see past his accent and were able to accept him as an equal.
“I bet that many people will find it hard to believe that an Irish person could be discriminated against. But I hope that the same thing can be said for women, 10-to-15 years from now.”—Laurence Moroney
“When I think of how difficult it is to be the only woman in the room, I pause and zoom out to think about how it's such a privilege to be working in AI. There’s so many women out there that don’t have education, that live in poverty, that live with violence every day, that are used for sex trafficking, and don’t have a voice, security or stability. Even if you’re not advanced in your career, if you have the time to mentor and give back, that’s what matters.” —Erica Lee
This article was written by Daniel Lipson in collaboration with WomenOfAI.org. WomenOfAI.org hosts many events throughout the year that are easily accessible for people at all levels of AI expertise and interest. For more about their nonprofit organization, you can visit their website at www.womenofai.org and join at www.womenofai.org/join. Follow on Twitter or LinkedIn.