Code Like a Girl meets world leader in AI and Chatbot tech, Kriti Sharma
December 11, 2018 by Penny Ivison
Kriti Sharma is an AI technologist and world leader in chatbot technology. She oversees bot development and artificial intelligence at London tech company Sage and invented Pegg, the world’s first accounting chatbot for entrepreneurs and startups. She’s also an Obama Foundation Civic Leader and Google Grace Hopper Scholar. She dedicates a lot of her time writing and speaking on the importance of acting now to eliminate bias in AI, as well as focussing a big chunk of her work life to this and to solving world problems with tech. All this by the humble age of 28. Kriti generously agreed to chat with us even though she’s in the midst of completing research papers (on top of her demanding job) because she’s so passionate about encouraging general diversity in the tech field. We think you’ll love what she had to say.
Code Like a Girl: You’ve carved out an extremely successful and purpose driven career in tech that many would desire to emulate. So what is it like day-to-day? Can you describe a typical work day for Kriti Sharma?
Kriti: I spend a lot of time these days working on humanitarian problems and making sure that AI is being used to solve some of the most difficult, challenging social development issues we face. I talk to organisations who have expertise in areas like women’s rights, equality, health care - finding out what the frontline issues are. I try to pinpoint where humans are not able to scale their efforts and are not able to help all the people all of the time and I identify where technology could play a role in that. My role is very end-to-end. It’s not just a small component of technology that I focus on. Once I identify how technology could play a role, I then work with a team of engineers to build the solutions.
I also spend a lot of time talking to robots these days, because - why not? And because robotics and AI is creating completely new ways for people to interact with machines, I spend a fairly large amount of time on identifying machine biases. And identifying potential risks to society if we don’t build AI well. A common example of this is voice assistants like Siri and Alexa having female voices and stereotypically female personalities. So I spend time designing what systems should look like to reflect a diverse society. How we can train machines to be good and to have values, especially self-learning machines.
Code Like a Girl: The biggest players in the technology space: Google, Amazon and Apple already have female voice assistants. I don’t see them suddenly turning Alexa into Alex or Siri into Steve, so how do we tackle this?
Kriti: It’s important to first look at why this has happened. My research tells me that a big chunk of it is that we don’t have diverse people building these products. You need diverse people on teams and not just in terms of gender or race but also life experience and backgrounds. I often say that the most important technology of our lifetime (which is going to be more transformational than smartphones or the internet) i.e. AI - should not be left to geeks like myself to design. We need people with very different skill sets.
The tech component of products is important, but it’s equally important to ensure that you are solving the right problems with it. And it’s not just the algorithms, but about the values we teach it. To do that you need people from the (so called) non technical backgrounds like the humanities. In my team for instance I have a conversation designer whose background is creative writing. She writes the personality of the robots and sets the tone for the human/robot interaction. We need artists, we need people with emotional intelligence, problem solving, judgement - all the skills that machines are not very good at but humans excel in.
There is also the issue of biased algorithms. For example, if you are using AI in the hiring process and the AI is matching candidates for the job. If historically it’s always been a man who has been up for the senior position, the AI can get skewed and believe that a man is more likely to be good at the role, just because of the circuit patterns, but we can teach machines to block these biases.
A good mini test for young people to do is to do a google image search for CEO. The first 9 or 10 results will be a white man. If you do the same for personal assistant you will get a white woman. These are the kinds of stereotype re-enforcements that we want to avoid when creating new digital products. I do a lot of work fixing things like this but also creating diverse data sets that the machines are learning from.
Code Like a Girl: Do you ever worry that it is too late to make a difference?
Kriti: I have hope. There are some good examples as well. Google Assistant has a male or female voice option. There’s some improvements. There’s also the ___ report that outlines the attempts that have been made to make the personality of these female assistants stronger. At first they might have been very apologetic and even if a user abused them they would say sorry, but because there has been a lot of backlash and challenge from the industry and society, we are starting to see some improvement. Another example is with facial recognition - white women have a 1% error rate and darker skinned women have around 35% and the facial recognition algorithms failed to recognise Michelle Obama and Oprah Winfrey. Now Microsoft and other companies have promised to focus on improving these algorithms and fix them. However I do acknowledge that it’s much harder to fix something later on that to get it right from the beginning.
Code Like a Girl: What made you choose technology as a career?
Kriti: It was quite natural to me. I always loved maths and science and I didn’t have a computer when I was a teenager, so I read a book and built my own. It took all of my fear of technology away. I built my own robots and I thought this is cool, I could do more of this. And then when I was quite young, I got invited to a government research lab where I saw one of the early versions of an AI assistant. A scientist was talking to a computer to make it do something. It was just amazing to me. I was blown away. I thought - I would love it if I could have a machine like this in my lifetime. Little did I know that it would happen much sooner than I thought, and that these devices would be everywhere in our homes and in our phones. Where I grew up, in Rajasthan in India isn’t the most developed part of the world and we do have issues around access to resources. The school I went to had one teacher for 70 students. It made me wonder if technology could help with these issues. Access to doctors is an issue, access to a proper education or high quality teachers is an issue. As a kid I would think that it would be great if machines could be used to help with these things.
Code Like a Girl: Did you have any visible female role models who were working in technology to look up to as a child?
Kriti: Unfortunately not, but my mum was a journalist and a very strong one at that. In the 90s in India this wasn’t very usual or common. So I definitely had her as my role model as someone who would just go out and kick ass and do great work. In many ways that inspired me to just go out and do what I wanted to do. To do what I loved regardless of where I came from or what I looked like.
Code Like A Girl: You were invited to be civic leader at the inaugural Obama Foundation Summit last year? What was this like?
Kriti: It was one of the most transformational moments of my life. It was exciting, not just because President Obama and Michelle Obama spent time with us and they are amazing people (obviously). But it was also because of all the people who were in the room. They were all change agents doing incredible stuff around the world to make the world better. For example there were people fighting for women’s right to drive in Saudi Arabia, others giving sexual assault survivors the right tools and legal support and there were some people like myself who were trying to make the internet and the algorithms fairer to everyone.
Code Like a Girl: What do you see as your career path going forward?
Kriti: I see myself hanging out with robots and machines forever. I think over time I will spend more and more of my time using this technology to solve world problems. Whilst the internet has solved many problems in the last few years. More people are online now than ever before and access to information has improved. There are also challenges that have come with it. We need to build technology responsibly. Young people can easily get addicted to their phones and waste time rather than being productive. We have seen in recent times some real issues that have come with data usage. I see my role in the near future being more about responsible technology and ethical technology. These days I spend most of my time helping small businesses improve their productivity, helping entrepreneurs. I spend time on building tools for women who face domestic abuse, harassment and violence and also healthcare.
Code Like a Girl: At the moment women make up only 10% of enrolments into IT courses at Australian universities but across both genders enrolments area at a 20 year low, so it’s a bit of a problem in Australia. What would you say to any young person (especially female) who is considering this path.
Kriti: It’s super awesome to work in technology, especially at this time where access to internet has improved massively, you can reach people through technology very very quickly. More so than ever before. Most importantly, it’s super fun. A lot of the work I do involves creative problem solving. And you get so much satisfaction from this.
In London I run a program called Future Makers teaching kids to build AI. These are not your typical Code Club kids. They’re from a range of backgrounds and we ask them, why wouldn’t you work in AI or emerging technology?
The top three reasons we get are:
They don’t think they are smart enough
They are not getting the right teaching from schools because sometimes you have the dynamic of the digital native students with a non-digital native teachers. The students can know more than their teachers in some areas.
They think it’s not creative enough
All of these things are fixable. You don’t have to necessarily be academically super smart to work in computer science, you might have different skills and that is perfectly fine. With regards to the school curriculum, there’s so much available now outside of your school. There are open source projects you can follow, there are after school clubs you can go to. In my time we didn’t have any of this so I had to read books and build my own computers. Lastly, it’s a very creative field. Even if you’re a youtube star for example, you need to know how to use technology to build your audience and to reach them at the right time, you have to create intelligent and engaging content. A lot of those skills are gained by trial and error and being prepared to try things. That’s the same way that I got into technology.