“I was amazed at how my data can answer all my questions like a magic box that gives whatever you want.”
For this week’s ML practitioner’s series, Analytics India Magazine (AIM) got in touch with Saishruthi Swaminathan, Technical Lead and Advisory Data Scientist at IBM. Saishruthi is an active ethical AI practitioner and advocate based out of California and has been a consistent contributor to the field of Ethical AI through open-source contributions. As of today, her work has reached more than 25,000 people around the world and got them exposed to Ethical AI concepts.
AIM: What got you interested in AI/ML, especially AI ethics?
Saishruthi: I did my undergraduate in Electronics and Instrumentation Engineering, and Master’s in Electrical Engineering, specializing in Data Science. Throughout my academic phase, I was figuring out what I liked and wanted to become. It was clear that I wished my passion and profession to be the same. So, I experimented with everything from machines to machine learning. When I moved to the USA to pursue a Master’s degree, I was hesitant to take programming as a primary specialization. That unknown fear made me opt for a networking specialization. Do you know what? Everything seemed abstract to me. I had a hard time understanding what was going on.
My internship was my turning point. I got to work with the AI innovation team in a startup I joined as an intern, and my tasks were around data and AI. For the first time, I understood what my data was trying to convey. I was amazed at how my data can answer all my questions like a magic box that gives whatever you want. Isn’t it amazing? That’s how I started my AI journey and made data my best friend.
Life took another turn when I started experiencing unfair treatment in my job search phase. Instead of looking at my profile, people began making judgments and assumptions based on my gender, race, educational background and looks. For a moment, I thought, what if we use this biased data to train the AI system? Tomorrow if this becomes a product, how risky can it be? Isn’t it scary? The same fear drove me to explore more about AI Ethics.
AIM: Can you discuss your contributions to the field of AI ethics?
I followed a three-step process – Educate, Empower, and Bring a Change.
Educate
I want to start educating people about the importance of building trustworthy AI systems and bring awareness about how our unconscious bias can affect lives around us.
As of today (September 2021), I have educated over 20,000 people worldwide about building trustworthy AI systems. Some of the event highlights are All Things Open-Diversity and Inclusion (Global), OpenUP summit (Asia), AI for Social Good (USA), IBM Think (Global), IBM Developer Series (Global), Data Science Seed (Italy), Linux Foundation Open-Source Summit (Europe), IBM Canada Technical Meetup, Girls Who Code (USA), Women Who Code (USA), AI Engineering Industry Experts Series (Global), and AI for India (India).
Empower
Create tools that can help detect and mitigate bias in data and AI models – focusing on one pillar of trustworthy AI.
As a next step, I want to make the open-source toolkit available for detecting and mitigating bias accessible to all communities. Programming language should not be a barrier to accessing the tools. I was contributing to the AI Fairness 360 toolkit in Python and then thought, what if we expand to other communities like R? This thought led me to create the AI Fairness 360 in R programming language. It’s been a year since we released the package, and we see that package usage is increasing day by day. Also, I have contributed to the AI Factsheets, an open-source project that aims to increase AI transparency. These are the critical contributions I have made to the field and a step towards my empowerment goal.
Bring the change
I wanted to create a recruiting platform that is free from bias and purely skill-based recruiting. I pitched this idea in the 2021 Silicon Valley Business Competition and reached the semi-finals (as Top 10). On a mission to take the work around the world and bring this to life.
AIM: Can you briefly talk about how one can/should venture into AI ethics?
Saishruthi: AI Ethics is a broad space. I want to answer this differently. If you want to enter the data science field, ethics should be a part of the routine. Data scientists who will live with the data day-in and day-out should be aware of the ethical concerns of their system. Often, I come across materials explaining data science methodology without an ethics component. I have also been in a situation where my attendees in the workshop have requested to skip the ethical design part as it was considered only for experts, not beginners. My question has always been – when do you teach good manners to your kids? At their young age or when they are old?
On the other hand, if you want to get into the AI Ethicist role, I suggest asking yourself, why do you want to become an AI Ethicist? Often we step into the field because of its popularity. Being brave is a quality that is required as you will be in a place to see beyond profit margin; once you have clarity on why the next step is – to start acquiring the necessary skills. We can develop technical skills required through different platforms but qualities we hold matter for the role. I started reading articles and research papers in the AI ethics space, understanding problems and consequences. I typically map my learning into an image that I can easily understand. Every time I approach a problem, I try to implement my learnings. I map out the value system of people involved, policies and governance, and unintended consequences of my proposed design. I would highly encourage community interaction. Be open to getting into the community and hearing the issues. Learning together and creating a safe space is a crucial aspect.
AIM: Can you tell us about your current role at IBM?
Saishruthi: I feel grateful that I get an opportunity to be a part of efforts across the organization. I lead initiatives across different parts of the organization like open-source, advocacy, client engagements, ethics board, research, and product. Let me tell you all a factor that I enjoy the most in my current role – no days are identical. Every day I get to innovate, be a thought leader in the trustworthy AI space, get new ideas to the table, and bring change for social good.
AIM: How do you see the landscape of AI/ML ethics evolving in the future? What are the current challenges persisting in AI ethics?
Saishruthi: Areas I feel will evolve – fairness, explainability, transparency, privacy, security, and robustness. Before getting into challenges, I would like to highlight that the system can only function well if we all participate. Responsibility is on all our shoulders, and asking the right questions before using technology is essential. Now, coming to the challenges: How can we cover people’s beliefs around the world? What if we cover only specific perspectives of people in the world? Can we create data that does not represent actual bias that exists in the world?
AIM: Which domain of AI do you think will come out on top in the next 10 years?
Saishruthi: I want to answer this question as someone who practices and advocates for trustworthy AI. AI will play a more significant role in our lives in the future. As you all can see, it is entering into every single field that you can imagine. However, the domain which can stand the test of trustworthy AI pillars and values will stay. I will keep sharing about this in my LinkedIn space. So, let’s connect there. In our journey towards trustworthy AI, machines are indirectly teaching and making us aware of the bias that exists in the world.
