Every week, we talk about important data and analytics topics with data science leaders from around the world on Facebook Live. You can subscribe to the DataTalk podcast on iTunes, Google Play, Stitcher, SoundCloud and Spotify.
This data science video and podcast series is part of Experian’s effort to help people understand how data-powered decisions can help organizations develop innovative solutions and drive more business.
To keep up with upcoming events, join our Data Science Community on Facebook or check out the archive of recent data science videos. To suggest future data science topics or guests, please contact Mike Delgado.
In this #DataTalk, we talked with Beena Ammanath, the Founder & CEO of Humans for AI, a nonprofit organization focused on increasing diversity in tech leveraging AI technologies. Follow their work on Medium and Twitter.
Mike Delgado: Welcome to Experian’s Weekly Data Talk, a show featuring some of the smartest people working in data science. Today, we’re featuring Beena Ammanath, the founder and CEO of Humans For AI. She also serves as the Global VP of Data, Artificial Intelligence, and New Tech Incubation at Hewlett Packard Enterprise. Beena, thank you so much for joining us. It’s an honor to have you.
Beena Ammanath: Thank you for having me on your show.
Mike Delgado: Tell us a little bit about why you started Humans For AI.
Beena Ammanath: Back in the early ’90s, when I was in college, we had artificial intelligence as one of the courses, and nobody really wanted to do it because it was considered so futuristic. Self-driving cars? Never going to happen. Personalized marketing? Impossible to do. And if you look at it, it was just a few decades ago, in our own lifetime, and now a lot of those things are happening.
So that was one data point for me — to see how things we thought would never happen are happening. The other big data point is the internet and mobile. It started in the computer science world. It started in the tech industry, but it impacted everybody. It doesn’t matter what your profession is today. You have a phone. You are connected to the internet.
I see a very similar thing that’s going to happen with artificial intelligence. No matter what your job is today, it’s going to change, either significantly or slightly, depending on the actual role itself. But AI is going to be so pervasive. And we’ve seen it with internet and mobile, so it’s that history that gives me that ability to see what’s going to happen with AI, and I think we have an opportunity because we missed the wave with internet and mobile.
What I mean by that is, if you look at it, internet and mobile created new job types. Like iOS developer. There was nothing called an app iOS developer, an app developer or UX designer. It didn’t really exist, and AI is going to create a lot of new jobs.
So that was one part of it. The other part is knowing artificial intelligence and what we’re trying to do, which is really about automating work we do more routinely. I know AI is still in its early phases, but later on, we’re going to need deep domain experts who are going to drive the AI products for their professions.
Today, it’s computer scientists like me who will interview, say, a lawyer and then go build a legal AI product or something similar for healthcare. But five, 10 years from now, we’re going to need these domain experts actually driving the product design and the product, what the product should look like for their specific domain.
Knowing that and then combining it with my passion for getting more women and more minorities into tech, I just saw this opportunity where we know there are new jobs that are going to be created. How do we get more women and minorities to do those newer jobs? If we knew that iOS developer would be a new career path 30 years ago, we could have trained more women and minorities. Then this whole issue around lack of women in tech would not have existed.
So that’s the attempt with Humans For AI: to proactively train women and minorities to be part of tech by learning more about AI but staying within their domain.
Mike Delgado: I love that mission too, especially for people not in tech roles. Whenever I hear about AI positions, definitely those roles usually require somebody who has a coding background, background in computer sciences or statistics and mathematics. Your organization is looking to bring AI out of tech and introduce it to people who are in general roles, like in communications and the legal field, because, in your opinion, that’s how we’re going to see AI thrive in the future, with these domain experts. Right?
Beena Ammanath: Yes. I think today AI is still in its infancy, so we are really addressing the low-hanging fruit. We are going after solving very specific, narrow-use cases, using AI techniques. We are not going for the big game-changers with AI. When we reach the limits with narrow artificial intelligence is when we will start looking at truly expanding AI’s own capabilities in use cases, and that’s when you’ll need domain experts. That’s where you need a marketing person to tell you what AI products make sense from a marketing perspective.
Another data point to this is also that I have this friend who is a lawyer, and he’s always complaining about the mobile app somebody built for lawyers. He says, “That’s not how I work.” Especially with AI, you are now trying to capture the intelligence behind that profession.
So, you need professionals driving the design. Today it’s computer scientists, the data scientists, people who know R and Python and so on who are doing coding, but I’m not talking about replacing those jobs. It’s more on the product side, more on driving the vision around what AI should do for specific professions.
The other part to it is we also hear a lot of fear around AI, about AI going rogue. Part of it is because it’s not the domain experts who are driving it who know all the nooks and crannies of their profession, who know all the guardrails that need to be put in place so the intelligence doesn’t go rogue. The fear exists because we don’t have the domain experts deeply embedded in building today’s AI software.
Mike Delgado: So, when you talk about the different domain experts, Beena, can you talk a little bit about other areas that you see for growth? You mentioned the legal and healthcare fields. What other occupations do you see where people can become domain experts and help out with AI technology?
Beena Ammanath: By domain experts, I mean professionals. It can be accountants. It can be finance people. The way we do accounting today, there’s a lot of manual work involved. We use digital tools, but there’s still a lot of auditing that needs to be done by humans, so how do you automate that piece? That’s again low-hanging fruit. Risk assessment, fraud detection. So, by domain experts, I mean professionals. We need surgeons involved. We need communications specialists involved. We need marketers. Anything you can think of where there are more tasks using intelligence as opposed to the actual physical art.
What I mean by that — a job for a physical therapist is probably not going to be as impacted by AI, because that’s more high touch. It may change the way she works by having different systems for running her business, but not necessarily the actual job itself.
Anything that uses human intelligence today, any profession that uses that level of intelligence, I think those people should be learning about AI. It also empowers these folks to shape how AI should be driving changes in their industry, as opposed to having a completely different profession driving these changes across all industries.
What we want to do is democratize AI, make it easy for people to understand these concepts. If an accountant understands NLP or machine learning and feels able to explain it in a way they can really relate to it, then they can say, “Ah. This is how I would use it in my day-to-day job.” That way, you’re using AI to make your day job better so you can focus on the more human aspects of the job. Does that make sense?
Mike Delgado: I love that. Yeah. Definitely. I’m really excited about just the progress of chatbots and seeing how chatbots can work within organizations to help us all be more efficient with our time and find information quicker for ourselves — and also voice assistance. That’s just two small examples.
Beena Ammanath: Exactly.
Mike Delgado: But that’s going to require training and data technology to learn more about our work. But it’s exciting. All this stuff is happening so quickly. Before we jumped on this chat, we were just talking about how five years ago, we were still talking about Google cars and making automated cars. Now it’s five years later, and so much is happening in the AI industry. I can’t help it, but every single time I look on the web, there’s some new article about AI doing this. You have the tabloid-type headlines about rogue robots taking over the world.
Beena Ammanath: Yeah.
Mike Delgado: And a lot of scary quotes out there.
Beena Ammanath: Yeah.
Mike Delgado: But I love your viewpoint on why we need to have more people involved in helping to shape this AI technology. Like you said, we are still in the infancy stages.
Beena Ammanath: Yeah, and a lot of the headlines are exaggerated. That’s because there is so much hype, and the fear exists. That’s why it is captivating. That’s why the media has tapped into the hype. It’s important to be aware of how much of it is the truth, and if you really read through those articles, you’ll see it’s a very specific narrow-use case. From then on, it’s somebody’s imagination that AI is taking over the world.
So, I think it’s important for more people to understand what AI is capable of, and I don’t mean they should know every little thing about how to build an AI program. But they should understand these core concepts.
Mike Delgado: I’m really curious. You got your MBA and your master’s degree and bachelor of science degree in computer science. As you were going up the ranks — and obviously we see a huge gender gap in technology roles — how did that shape you as a student going into a data science profession and now having a drive to get more diversity?
Beena Ammanath: You know, now that I think about it, there’s nobody with the title of data scientist. So I cannot say that I dreamt of being in this profession or of doing this one day. I think it kind of naturally evolved. I think computer science was something that attracted me just because of the mathematics and statistics side of it. It’s always just been natural and easy for me to understand and to grasp, and I really enjoyed it, too.
So, it was just a natural course for me to go into computer science. And within computer science, I found that databases, anything to do with data or looking at data, fascinated me. But I have to warn you that when I was looking at databases, it was FoxPro and dBase and Excel. These are antiquated software now, but I was looking at the very early stages. Today, what we have with all the big data stuff going on, things have just changed. But the core concepts still hold true, and I think that’s what has helped me in my career to be able to have a good understanding of the core concepts in mathematics or in core science.
People usually ask me what language they should learn, what will help them grow, what will help them have a great career, and I don’t think it matters. It’s about having a strong foundation. I think programming itself might not be done by humans for too long. We’re going to reach a point where it will be people like you who are describing a problem statement, and if we do AI right, AI is going to do the programming for us.
So, it’s more important for us to understand the core concepts and build out the foundation, which really helped me in my career. But throughout my career, one thing I’ve stayed true to is the data part. Every role I’ve done at different companies … I’ve worked across different domains, financial services, telecom, retail, e-commerce, IoT manufacturing and now services. So I’ve just worked in different domains, but it’s always been about data.
I’ve seen when data was more on an OLTP or a transactional database side, where storage was super expensive and you had to think of ways of normalizing the data. So you stored it in a very compact fashion. Then came the phase of data warehouses and business intelligence, where the way you organized the data was different, whether it was Starsky model, Snowflake or so many different types, and that was more on learning how to run our reports faster. How do we look at the data to see what happened, to be able to look at trends? Then came the wave of big data. With the advent of Hadoop and the newer technologies, we are able to look at unstructured data. We are able to look at data and combine it in a way that could not be done before.
And now we can do that. So, I’ve just seen the whole evolution of the data space. Honestly, I think I got lucky that I grew up in the data space because this has truly changed so much in the past few decades, and it has helped me grow, because everything is built on the previous data.
What I mean by that is when data warehousing and BI came into the picture, there was a fear that the transactional databases would go away. And they didn’t. You still need them for your record-keeping. There’s a place for it. When the big data technology started taking off, there was a fear that the whole business intelligence or data warehousing space was going to go away, and it didn’t. Every step is kind of adding on to how we look at data, we drive more insights from it and use it to drive positive business outcomes.
Mike Delgado: One of the other core missions of Humans For AI is the importance of diversity. Why do you think diversity is so important for the future of AI technology?
Beena Ammanath: It goes back to what I was saying. A lot of what we’re trying to do with AI is so close to the intelligence that exists. The classic example is a program that was written by two programmers: an AI program to identify shoes, an image recognition program. It could recognize boots. It could recognize flip-flops, but it could not recognize women’s heels. That it would not classify as a shoe. The reason was the model was not trained with a diverse set of data. [inaudible 00:18:34] It was built by two male programmers, and it just did not cross their minds to train the model on that data so the model could recognize these kinds of shoes as well.
That’s a simplistic example, and that almost seems trivial, but when you think about it, if you do not put enough diversity in the training data that you provide, you do not bring … and when I say diversity, I’m not just talking about gender or racial diversity. It’s also diversity of thought, diversity of your background, your experience. For AI to thrive and reach its full potential, you need that diversity at the forefront. Otherwise, it’s going to be these incomplete AI programs, which just won’t help AI’s own growth. So, for AI’s own growth, I think diversity is needed.
Mike Delgado: And I think that speaks to when data science leaders are looking to form their teams, having a diverse group of people who have different backgrounds, whether it’s a Ph.D. in mathematics, statistics, physics, biology … people who have different academic backgrounds can come to problems and will attempt to solve them in their own way.
Beena Ammanath: Yeah. But let me stop you right there. Even saying that you need a Ph.D. in diverse fields is not diverse. You need people at different levels to bring that diversity. So it’s diversity from every aspect.
Mike Delgado: Yeah, that’s totally true. I’ll share with you a quick little story. We have a data lab in San Diego, and we’re always recruiting different data scientists in lots of different fields. One of our latest hires graduated with her bachelor’s degree from UCLA, I think two years ago, at age 18. Stellar student, brilliant, her degree was in statistics. As a hobby, she just played around with computer science. UCLA every year has a computer science competition. On a whim, she decided to enter this competition. She won. She’s not even a computer science major. She has a strong stats background and then as a hobby she would do computer science on her own. So she ends up winning this competition, and our data lab snatched her immediately. Because, like you said, it doesn’t require a Ph.D.
Beena Ammanath: Doesn’t require a Ph.D. Doesn’t require computer science. I actually know an actress who’s building a chatbot.
Mike Delgado: Oh, really?
Beena Ammanath: Yes. So, it’s true that diversity can come in so many different forms. But going back to what you were referring to when companies are building a data science team. To think that a data science team needs only data scientists is not going to help you succeed, because you need people who understand the business aspect of it.
You need great storytellers, visual designers, who can communicate the findings from the models. You need to be able to tie it in with the business problem and show how it all fits together.
So, data science teams are a group of people who collaborate well. I think just having data scientists is not going to give you that complete view or success in a way that a diverse team could give it to you.
Mike Delgado: Beena, if you were looking to hire on a new data science team … You just mentioned all the different diversity you would want. What would be the perfect team for you?
Beena Ammanath: I’ve done that a few times, and I’m actually satisfied right now, so I think hiring a data scientist, obviously, you know, one who understands machine learning, who has … I would look at different levels of experience, like somebody who has a Ph.D. in machine learning and has done it for several years, somebody who is just fresh out of college, that part, then there’s definitely need for data engineers because a lot of data science work is data janitorial work, cleaning up everything.
Mike Delgado: I like that, data janitorial work.
Beena Ammanath: Yes. It is janitorial it’s messy. About 80 percent of the work to build a good model is in that janitorial work, where you have to be able to source the data from the right sources, clean it, make sure the data quality is up to mark, prep it, make it ready before the data scientists can take over and start building their model. And then QA is another big function of it, you know, being able to test, train, not only train the model, but test it on different data sets. That’s crucial. And then UX, user experience designing in a way where it’s not just PowerPoint, much beyond PowerPoint. That’s crucial as well. So I would look for people with different skill sets who can … you know, the core is still data science, but you’re able to deliver an end-to-end solution that encompasses all this.
Mike Delgado: Beena, what advice would you give for somebody who is applying for a job to join a data science team? They’re interviewing with you. What would be some things you would be looking for and maybe some advice you’d give them?
Beena Ammanath: I usually look more for the human skills first. You know, the team fit, the culture fit is very important for me. Somebody who can come in and gel with the team and has the right attitude. The other thing I look for is curiosity. How curious is this person to keep digging through the data and find that nugget of information, that valuable piece that can help us drive more productivity or bring better value? So, curiosity and collaboration, being able to work in a team environment, respect each other’s talent and the value you bring to the table, those are the three skills that I look for from a soft skill side.
Obviously, it’s a given that if you’re interviewing with me, you already passed the technical screen. You know how to and you’re a ninja at coding in Python or R or whatever your coding language of choice is and are very savvy with the current technologies that exist.
Mike Delgado: I know we only have a couple minutes left. Before we went on air, we were chatting about our kids, and I want to talk a little bit about, just quickly, how you are preparing your kids for the future of AI. Because I’m wondering this right now for my kids.
Beena Ammanath: They do not go for any programming lessons or programming classes. I tell my son, who is 14, to focus on understanding the concepts. Try to and always think about how you would solve this problem differently. How would you solve it if it was … yeah, just really leverage using your own human brain, not try to depend fully on a computer or a calculator. They can be aids, but to build out the foundation, I think you need to run it through your own brain to get there the first time.
Mike Delgado: I love that. I’m going to put up the URL for everyone to see. HumansForAI.com. Beena, what will people find when they get there?
Beena Ammanath: They will find a website that is currently being rebuilt, but what we are trying to do is build out a community where anybody can come in and learn more about AI and see how it applies to their profession. Our community has not only data scientists and the AI experts, but also marketers, lawyers and accountants. We want to make this a community that is very open and welcome where AI doesn’t feel like something that’s accessible to only a privileged few. It’s somewhere everybody can come in, ask their questions, and have honest, open conversations on how we shape this technology.
Mike Delgado: Wonderful. Well, Beena, I want to thank you so much for your time. I want to encourage our data science community to check out HumansForAI.com. If you’re interested in joining up, please get in contact with Beena or others there. There is a wonderful community of people there who are helping to bring data science to the masses. Make sure to check out HumansForAI.com. If you’re new to Data Talk, you can always learn more about our weekly show and our podcast by going to experian.com/datatalk. I want to thank everyone for watching. Thank you for your hearts, your shares, your comments here on Facebook Live. We’ll see you all next week.
Beena Ammanath is the Founder and CEO of Humans for AI and the Global VP of Data, Artificial Intelligence and New Tech Incubation at Hewlett Packard. She also serves as a Board Advisor at iguazio, Predii, and Cal Poly University.
Check out our upcoming live video big data discussions.