Inclusive AI – Ep.95

With the rise of artificial intelligence impacting our everyday lives, I wanted to revisit my previous conversation with the co-founder of Deep Labs, Scott Edington, about AI’s influence on inclusion.

Melyssa Barrett:  Welcome to the Jali Podcast. I’m your host, Melyssa Barrett. This podcast is for those who are interested in the conversation around equity, diversity and inclusion. Each week I’ll be interviewing a guest who has something special to share or is actively part of building solutions in the space. Let’s get started.

Hi everyone, it’s Melyssa Barrett here. I figured since everyone is talking about artificial intelligence and how it is being incorporated into all sorts of things with ChatGPT and everything else that it might be kind of fun to repost an episode that I did really early on in, I think it was 2020. Scott Edington, who is the CEO of Deep Labs, and I just thought it might be interesting to go back and hear some of what we were talking about when it comes to artificial intelligence.

Scott is the CEO of Deep Labs and his career spans two decades of creating next generation technology in the payments defense and intelligence sectors. I got to know Scott, we both worked at Visa for a period of time, and he was Global Head of Research and Development at Visa and really founded the Visa Labs. Now they have an innovation center and all of those things, but he was in the groundbreaking states of shifting Visa in many ways when he was there. He began his career at Booz Allen Hamilton. He’s been an adjunct professor at Georgetown University and George Mason University. And he’s a guest lecturer. He holds degrees from the University of Virginia and Johns Hopkins. I should obviously be calling him Dr. Scott Edington, but he has many industry certifications in security, strategy, and network computing.

He is all around persona based artificial intelligence at Deep Labs. And they use that persona based artificial intelligence to enable businesses to solve complex problems, and they’re continuously assessing risk in real time. That was something we knew very well working at Visa because that’s the way you end up being able to reduce fraud, identity theft, optimize your customer experience, and make better business decisions, and therefore more revenue. With that, I will let you hear an excerpt from episode seven of the Jali Podcast. Check it out.

Dr. Edington, thank you so much for joining me for this wonderful conversation. I am ecstatic to talk to you about all of the things that you have accomplished with respect to now you being the CEO and Co-Founder of Deep Labs. And maybe we could just start there and you can talk about what is Deep Labs.

Scott Edington:  Sure. Well number one, thanks for taking the time to spend with me this afternoon. I really appreciate the opportunity. Deep Labs, in a 30 second [inaudible 00:04:46] view, it’s a company that’s pretty much focused entirely on context where artificial intelligence capabilities. What that means for the non-geek out there effectively is a thinkable world where computer systems really fully appreciate what persona you might be exhibiting at any given moment in time. The easy example I always use is, Scott Edington sitting in Oakland, California pre-pandemic is a little different than Scott Edington having shelter in place for two and a half to three months in Oakland, California. Yes, from an identity perspective it’s still Scott, but obviously the context and everything around all of us has shifted. And because of that having systems that are intelligent enough to understand that yes, still Scott, but the personas that Scott will now exhibit will be dramatically different or will have shifted based on context, based on surroundings, based on everything that’s happening in 2020.

Melyssa Barrett:  Yeah, that’s interesting that you call them personas because in the world of diversity and inclusion, and of course I’ve worked in the area of identity, it’s all about your identity. And it’s almost like we have multiple identities all over the place, whether you’re at work or at home, or what your shopping looks like, or any of that kind of stuff. Those personas, I’m assuming, can really engage a person in different ways depending on what they’re trying to do.

Scott Edington:  No, I think you nailed it, Melyssa. Obviously given your background, you fully appreciate all the interesting things that have happened in the industry in the last 10, 20, now 30 years in our respective careers. And the reality is that circa 2005, 2010, simply using static attributes, like mom’s maiden name, social security number, that was fine frankly. And that’s our most keyed off of understanding is it really Scott or is it not, right? But as we all know, unfortunately due to some well known data breaches and other things that have occurred now in the world, those static signals are frankly very much antiquated or, in my vernacular, decayed. And they’ve decayed because everyone has access to them now. You can Google where I went to high school, you can Google where I was born. It’s freely available and most of time it’s pretty accurate. And if it’s things you can’t Google, you can certainly find on the dark web for next to nothing.

Because of that, as we built on those practitioners, there were a number of point solutions providers that came up specifically focused around things like address verification or device signals and device intelligence, which again, were absolutely fantastic, absolutely fantastic as a way to patch the holes. But unfortunately like anything else, the advanced threats that exist out there, whether it be from nation state actors, whether it be from folks who are individuals trying to do nefarious things, these signals have also decayed. What we’ve really been focused on is securing a mechanism and capabilities now we have to effectively create a full context rich understanding about an actor.

And an actor for us, by the way, could be an organization, it could be a consumer, it could be a human being, it could be a bot. The idea is effectively we want to be able to understand at multilayers, which signals should be used to identify what context that actor should be in, and then understand exactly what persona we’d expect that actor [inaudible 00:08:06] at this moment in time versus what reality is. And then effectively figure out a mechanism to see whether or not is that really who we expect Scott Edington’s persona to be right now versus we’ve actually seen this persona before, and the last time we saw it, it was associated with a bad actor. I know they presented all the correct credentials, but decline, decline, decline. Or it’s a little different than we might expect from Scott at this moment, but oh, he’s in Washington DC today as opposed to California. Therefore, let’s not introduce friction to this consumer experience. Let’s go ahead and approve that transaction. It does work both ways, either from a [inaudible 00:08:43] standpoint or an authorization standpoint, all the way to the opposite spectrum, which is advanced risk capabilities.

Melyssa Barrett:  One of the things I hear a lot about is just in terms of people of color, and I spent a lot of time obviously on financial education, there is this whole segment about financial inclusion. Essentially there are many people, specifically of color, who are not part of the financial system, but I think all economies think that they should be working or at least have a strategic plan to address financial inclusion. Can you talk a little bit about how artificial intelligence and where we’re going impacts people that maybe they don’t have a bank account or maybe they’re using a check cashing or doing some alternative financing? How does that impact some of the stuff that you’re doing with artificial intelligence and all of those discussions we’re having on privacy and other things?

Scott Edington:  Yeah, well number one, thanks for the question. It’s a topic that I’m, as you know, very passionate about, specifically the underbanked populations. Whether it be here in the US or certainly globally, the numbers are very clear, they disproportionately affect people of color. I would say first and foremost, I think the key thing is understanding, and going back to my college days, what are the influences, the why that’s happening? So that could be both macro and micro. I think there are obviously systematic reasons for some of the underbanked reasons.

But I also think from a techy perspective for a second, if you take a step back and look at how typical bank systems or financial services systems work, how they onboard new customers or clients, how they make credit underwriting decisions, how they verify things like income or verify addresses. These are all sort of things that frankly you and I take for granted given where we are, but these are real things that are hurdles and introduce friction even for something simply as I want to start a bank account for my eight year old. Or I want to start a college fund, or I just want to have that checking account so I can write checks or, not that we write checks now, but have some space in the digital economy.

When I would say, going back to my earlier statement around signals and noise, the ability to not always focus on also the antiquated signals of yesteryear, mom’s maiden name, what high school you go to, the old school static signals, but more focused on the near or the recent signals that now exist based off of where we are and normally where we would be in 2020 and hopefully now in 2021. I stress that only because whether we again talk about digital onboarding remotely because now everyone really doesn’t want to go into a branch, but how do you actually identify that is in fact, Scott? Without having a social security card, and I’m dating myself, social security card or a passport, how do you do that remotely? How do you do that securely? How do you do that in a frictionless environment, that’s not overly invasive, number one, but still has a confidence level that’s high enough that actually meets the guidelines necessary for a risk-based system.

A long answer to your very short question, the reality is that for the underbanked populations, keying off of things that are static signals of yesteryear is actually a method of disenfranchisement. It is. And so if we’re able to use modern signals, if we’re able to use the context that we talked before around personas, then in fact you can start providing banking services and banking capabilities to folks who ordinarily wouldn’t necessarily have a generic checking account. If you’re talking about, let’s say the prepaid market as an example, because as we both know, that’s something the underbanked population does use, does leverage, how do you actually still enable them to be onboarded and still meet the demands of the Patriot Act? The Patriot Act, as you know, that sometimes is difficult for some people from a hurdle perspective. Again, is it simply keying off of static signals or can you actually start to replace some of those static dynamic signals that we now enjoy? And the answer overwhelmingly is yes.

Melyssa Barrett:  Well, and it’s interesting because it seems like the more signals we get, the more models are built that provide things like credit scoring and fraud scoring capabilities. And if the underbanked population has a lack of information in those models, then it can create other challenges for whatever the objective is, which to your point, hopefully is less systemic in terms of hurdles around systemic racism and stuff like that in the past because at least it’s using information in an objective manner. But if the data is not there, then I guess you have other issues. How do you account for things like signals that aren’t really there?

Scott Edington:  I’ll answer in two parts, I think first thing is let’s just fully recognize how the old ball systems work and what they go for index for, which is give your background. You know firsthand, they’re over indexed with things that frankly tend to represent certain populations, not others. If you look at it from a pure data science modeling perspective, what does the sample look like? And figuring out what should the sample be in 2020 versus maybe what it was in 1985 or 1990 when a lot of these systems were first theorized. If you eliminate the, dare I say, the lack of proper sampling, you actually fix that, that’s step one.

Step two is if you now have a proper sample or a fully representative sample, then you can start looking at other methods and other attributes and characteristics that could represent the gap that the [inaudible 00:15:23] mentioning. Again, as opposed to keying off of where you went, I’m making this up, where you went to high school, the fact that you’ve had a job for the last two years at the same place, frankly I’d rather know that.

Melyssa Barrett:  Than where you went to high school.

Scott Edington:  That’s cool, back in the early nineties. I say this somewhat flippantly, but it’s actually true. Again, if you really just think about the context surrounding each individual, human beings are remarkably great at picking up on things. You know when someone’s a little bit off that day, right? Or your gut says something’s not quite right. Up until now, those models, they didn’t really take into account instinct or gut. They would be hard, fast numbers. This is way we’ve always done it, decline. As opposed to, wait a minute, this guy’s had a job for the last two years, same place, same income level. That company’s doing well. I can tell you that company’s doing well because a thousand other factors that we have access to. You know what, he’s actually not a credit risk. That’s actually someone we want to get behind as opposed to decline because we can’t figure out where he went to high school. Who cares?

Melyssa Barrett:  Yeah, yeah. No, it’s really interesting because … where do you think we’re going now since you’re all into the artificial intelligence and cryptocurrency and all of that, what does life look like 10 years from now?

Scott Edington:  I’m glad you said 10 years as opposed to next year.

Melyssa Barrett:  It could be anybody’s guess, 2021, right?

Scott Edington:  I’ll be happy to make it to 2021, right? No, in all seriousness, I think from a financial services perspective, I think it really does come down to just hyper-personalization, but not in a creepy way where it’s like, okay, yeah, uh-huh. This is too sci-fi-ish, like 1980s the sci-fi movie-ish, I don’t want that, right? But what the customers do want, and to be fair, it’s also a little bit generational. The kids, I sound old now, the kids who grow up right now, they had a certain expectation about what an experience looks like versus someone, let’s say, my parents’ age who are like, wait, no, I want to actually type in my name. I don’t want them to know it was me who just walked in the door. And it’s slightly generation.

I would say to answer your direct question, 10 years from now, it really is all about hyper-personalization. It is about understanding not only that it’s Scott, but it’s Scott on a Friday afternoon, somewhat dressed down. And right now I can extrapolate that Scott’s sort of quasi-business Scott, but he’s sort of sliding into the weekend. So his person will have shifted. The things that Scott will expect to have as he walks into a restaurant or walks into a grocery store will be dramatically different than if he was Tuesday afternoon, full suit in New York City walking down the street. Yeah, it’s still Scott, but hey, I’m going to influence Scott to maybe purchase this item. Or don’t bother Scott at all because very clearly he has a serious look on his face, he’s carrying this briefcase, and he’s headed to a customer meeting.

Again, for me it is about total hyper-personalization. It’s about understanding exactly what persona that actor is exhibiting at that moment in time, what persona we expect the person to exhibit in a window of time. And then ultimately, how do you best create an environment that is friction free and also very secure that again creates that hyper-personalized experience such that consumer wants to use your product or that consumer wants to be part of your community? Not to totally geek out on you, but that’s coming and frankly that’s what’s already here.

Melyssa Barrett:  Yeah, well that’s really interesting because, I’m maybe dating myself now, but I still remember the Jetsons when I was growing up, and that’s what it feels like. We’re kind of moving into the Jetsons, which is pretty old when you think about what they were envisioning way back then. Very interesting stuff.

Scott Edington:  Well, what’s funny though about the, because I used to watch Jetsons and all that stuff too, is that if you really dissect those shows, all that stuff’s already happened.

Melyssa Barrett:  Yeah.

Scott Edington:  Whether it be the scooters that try to knock be over in the middle of Oakland half the time or down here in DC, that’s real. Whether it be all the sensor technology that enables things like the Segway devices to work, or whether it be even our online experiences where, lo and behold, I’m talking to you via video camera in realtime [inaudible 00:19:49] it’s almost like you’re in the same room with me. That was all, I don’t know about you, but I know when I was growing up in the late seventies and the eighties, that was all sci-fi stuff. I didn’t think it was going to happen, and that’s here now. Even all the way down to using a mobile device and being able to unlock it with your face or your fingerprints. Or to be able to transact, I send some form of currency halfway around the world, and as you know, Visa, less than a second. That’s amazing.

Melyssa Barrett:  Yeah.

Scott Edington:  That’s already here right now.

Melyssa Barrett:  Yeah, I love it. I love it. I’m going to kind of go off on a little bit of a tangent because I know now that you lead a company, Deep Labs, there has to be lots of discussion on representation and diversity. And I know you have somebody that’s dedicated to, I think they call it people and culture, if I remember her title properly. And I have been at places where people say, I don’t want to talk about company culture. We don’t have a corporate culture. What does that mean? Now we have this whole segment of people that are focused on diversity and inclusion, and I think probably every practitioner would agree that ideally those jobs should go away at some point because really we should have diversity and inclusion, belonging, equity all embedded into all of the company activities. Are there things, especially in your business I imagine with representation being challenging for STEM students, what are some things that you guys are thinking about in terms of accessing talent in those areas?

Scott Edington:  Yeah, great question. As you know well, that’s a topic I’m quite passionate about and fully believe in. I would say my mindset, to be fair, is also very much given the fact that both my parents were children of the sixties growing up in the South. And again, my grandparents, both sides were educators. So again, education is very important. It’s the building blocks of how you get ahead, so to speak, to actually achieve the “American dream”. And so with that as a backdrop, my father worked at HBCUs for most of his career. Again, understanding how HBCUs work, understanding how the network effect associated with that, whether it be when I was at Booz or at Visa, and certainly now we do tap into it from a pipeline perspective.

The good news is there’s plenty of folks out there who are super motivated, highly intelligent, and they just want an opportunity. The difficult part is figuring out how to map those folks who have all those attributes that you want into roles, and frankly also locations, where there are communities of people of color. For us at Deep Labs, the cool thing, I’ll plug my company for a second, sorry, I have to do it, is that we have locations, whether it be in the Bay Area or places like New York and Washington DC, which as you know have large communities of people of color. And so we’re actively seeking folks who want to want to join a company that is very much forward looking, forward thinking, but also very much mindful of very clear core values around things like accountability, which is important, but collaboration and integrity, and having high degree of teamwork.

And I stress that because at the end of the day, from a pipeline perspective, especially within the STEM perspective, most folks who come up in those programs, they’ve had to abide by those core values just to make it, right? You have to be accountable, not only to yourself, but to other classmates. You have to have integrity of your work. Obviously, you’re not going to get your diploma if you’re not being honest. You have to be able to show teamwork and collaboration because oftentimes you’re dealing with group projects and everything else. For us, again, really having a dedicated function within our company focused on diversity and inclusion and people and culture, we believe is a fundamental enabler to the success of our business.

And I would say having spoken with other people of color that either have co-founded or founded other companies, or senior executives at Fortune 100 companies, that is something that I believe is universally shared at this point. I will say that unfortunately for us, much like everyone else in the pandemic, it has hurt a little bit our ability to recruit, if for no other reason, just getting in front of people. But I would say for those who are listening that are actively seeking positions, there are people like me who are actively looking for people like you.

Melyssa Barrett:  Yeah, that’s awesome. Even during the pandemic, I think people thought there’s no jobs out there. And I kept trying to send them out because every other day somebody would say they’re hiring, even through the pandemic. I think it’s interesting to see how growth occurs because clearly there’s some contraction going on, but then in other areas there is still lots of growth happening. And I know we’ve seen lots of digitization growth over the last several months, so it’ll be interesting to see how things develop. I’m looking forward to seeing you in 10 years, flying a self-driving car somewhere. Or maybe back in defense somewhere, who knows.

Scott Edington:  Who knows?

Melyssa Barrett:  Well, thank you so much for being with me today, and it’s always a pleasure catching up with you Dr. Edington. I know you do a lot of teaching. I know that’s near and dear to your heart as well, so I appreciate you bringing up Historically Black Colleges and Universities because a lot of the HBCUs don’t get the credibility, the visibility that they really deserve. Kudos to your parents and your grandparents for putting so much into you that you can continue to feed it into other generations.

Scott Edington:  Yeah, I appreciate that. And I’ll definitely pass it along to them. They’ll be pleased to hear that.

Melyssa Barrett:  Awesome. Well, thanks so much for being here, Scott, and I look forward to staying in touch.

Scott Edington:  Thank you, Melyssa. Take care.

Melyssa Barrett:  Thanks for joining me on the Jali Podcast. Please subscribe so you won’t miss an episode. See you next week.