Can you talk a little bit about your work as an AI researcher (past/present)?

My graduate degree specialization is in Machine Learning and I love that it is a domain that finds application in almost every industry today. This has allowed me to work with companies and groups that are involved with causes that I care about. I have been working for SkySpecs, a startup based in Ann Arbor, MI for the last 5 and half years. At SkySpecs, we are working towards building a digital platform for wind energy asset operation and maintenance management with the goal of reducing the overall cost of wind power generation. Reduced costs will help increase profit margins and thus promote further investment in renewable wind energy. From improving operational efficiency to predictive analytics, there is a lot of scope for application of AI in this space and the AI-focused teams at SkySpecs are working on solving some of these problems by leveraging domain knowledge along with Machine Learning skills.

Apart from SkySpecs, I also volunteered some of my time with Climate Mind, a non-profit group  that aims to build a system that presents to an individual the climate information based on their unique set of personal values with the goal of motivating them to take action. 

At Climate Mind, we are utilizing Natural Language Processing techniques to parse free text like media articles and create a Climate Change Knowledge Base. The Knowledge Base will connect higher level causes of climate change with lower level impacts on human life and with solutions to the same. 

Jumping from the field of work to my modus operandi as an engineer in AI space. Previously, I would spend most of my time on research and development of algorithms – for proof of concepts. More lately, thanks to the availability of easy-to-use AI frameworks and advanced libraries and models combined with team experience in this space, my focus has been on the application of AI models to real world data. Additionally, I advocate solving focused business problems and providing direct value to the organization by taking these projects beyond the proof of concept stage and productionalizing them.  

 

What are some of the issues women face in the AI field (in terms of education, getting jobs, career advancement, etc)?

I can speak only from what I have observed and I am sure my observation is just a small piece of the puzzle. 

To start with, I looked up some numbers. According to a research by WIRED and Element AI done back in 2018, 12% of leading ML researchers are women and according to another study by World Economic Forum and LinkedIn, a total of 22% of all jobs in the field of AI are held by women. So the numbers are pretty low – surprise surprise! Honestly, this story is pretty consistent throughout the engineering space. Now, human beings learn from examples. We are more likely to feel motivated when we see someone like us succeed, because that shows it’s possible for us too. So when women see very less representation of their gender in a field like AI, it creates a subtle psychological barrier to entry. Apart from this, there are other forms of barriers to entry for women depending on which country / geography one is from. In certain patriarchal societies, girls aren’t encouraged early on to take up technical degrees. In the west, thanks to popular media and marketing, computer education and technology have been stereotyped as a playing field for boys. The consequence of this is seen as skewed gender ratios in universities and further along in workplaces.

Even if women get higher education and find their place in the workforce, case studies and data do show that women have to work harder to prove their credibility than men, especially in STEM fields like AI. There is an inherent bias in quite frankly almost all of us that assigns lower expectations to women than men when it comes to technical capabilities. 

Moreover, there is the fact that women still for the most part have more household responsibilities towards kids and families and have a more challenging job managing the dual responsibilities of work and home. And then there is motherhood. Pregnancy and childbirth often put a speed breaker on the mother’s career growth. It’s not only important to have provisions for good maternity leaves but equally important to offer long paternity leaves, because if the father has to work, the responsibility of the child falls on the mother and her career has to take the back seat.

When because of all these reasons, it becomes harder for women to rise, you see fewer women in the leadership roles, which loops back as fewer role models for younger women to look upto to enter into this industry in the first place, and creates this cruel cycle of lack of representation.

 

What kinds of issues can a lack of gender diversity in AI cause?

Generally speaking, gender diversity in the workplace is important for several reasons and there is a lot of research on this topic. For example,

  1. When you have people  from different backgrounds come together, you are bringing a lot of different perspectives and creativity to the table. This kind of community is more likely to come up with an “out of the box” solution and innovative ideas. 
  2.  If your company has a diverse client base, having diversity in your office helps bring more insight and understanding for interacting with this diverse client base. 
  3. And this one I strongly believe in. Diversity attracts diversity. A company with diversity is perceived as more progressive, equal and global and is likely to attract talent from all over. For example, as a woman, it’s natural for me to be a little wary of working in a place that only has male employees. 

This is true for all fields by the way, and is especially true for AI since there is a known lack of gender diversity in this field. 

But, in the field of AI,  there is an additional concern that has become quite the news these days – and that is the issue of gender bias getting embedded in AI systems.

There is overwhelming evidence that gender biases are baked into some of the popular AI tools out there. This can be seen in examples like the infamous Face Recognition system by a well-known tech giant that performed poorly on dark skinned women. I remember reading another example about an automated speech to text system that performed much more poorly on the voice of female speakers than males. And not just women, but these issues affect other minority groups too in terms of bias in color of skin, accent etc. In all these cases, the models were built and likely optimized on male datasets. These tools often reflect the biases of those who were building the datasets. So having diversity in the AI room, may have helped catch some of these concerns with data early on. And in general, as I have mentioned before, a more diverse team is likely to identify and catch a wider variety of issues.

 

What can we do, as individuals and the field as a whole, to make AI more inclusive for all genders?

Ah, good question. I wish I knew the right answer for this one. But I guess if we are talking about gender equality, then I believe that the adoption has to start early in the life of a child. This could mean that media, brands, families, schools have to start promoting gender-equality and avoiding stereotypes early on. Boys can play with dolls, girls can play with robots. I know that there are a lot of independent movements happening across the world to help bring about this change – like Girls Who Code. I used to facilitate a local Girls Who Code chapter in Michigan sometime back. They are bringing coding principles to Middle and High School girls to get them exposed to STEM fields early on. There is another initiative called Society of Women Coders and their mission is to introduce technology to young girls from developing nations with the goal of  increasing their access to careers in technical fields, higher paying jobs and eventually financial stability.  Maybe we could have more such initiatives at the school level.

At the workplace, companies have to weave in inclusion efforts into their hiring culture keeping in mind that it is harder to find minority talent in a general application pool. There are a few strategies that companies can adopt to improve diversity while keeping the process merit-based. One example could be reaching out to targeted groups and communities to increase the diversity ratios in the application pool and only pulling from the application pool once certain diversity targets have been achieved. Another effort could be to propagate the importance of gender diversity throughout the company through consistent messaging like via  gender and bias training, setting up committees that track Diversity, Equity and Inclusion metrics, holding the leaders or managers accountable for them, and improving the representation of women in public forums and conferences. All that said, I think one of the most important pieces to achieving progress is to have male participation in this effort. Without the support from men and leadership, progress will be much harder to achieve and a lot of efforts may end up getting directed towards a lost cause.  

 

Are we on the right track to making AI a more gender-balanced field?

This one is hard to say since I don’t have a lot of direct reference from the past but I do believe that equality and discriminiation is taken a little more seriously in workplaces these days. Women empowerment has also gained some traction in society. That said, there is still a lot of scope for improvement and I do believe there still persists inequity when it comes to equal pay for equal jobs, female representation in executive leadership roles, and the rate of promotion among males and females. 

In the field of AI, we have some good initiatives, primarily backed by strong women leaders in the community that are working towards promoting gender diversity and representation in AI through groups like Women in Machine Learning, Women Leading in AI etc. I have seen some initiatives like disallowing All male panels, demanding an ethical AI board and audits to the hiring process. So it’s a slow improvement, that’s gaining some popularity but it needs support  – 

  1. The big tech companies need to set good examples for the rest of the community to follow, and 

  2. As I said before, we still need unequivocal support from the male leaders in society. Without these efforts, this is going to be a long battle.