In the world of AI, are you team accelerate or decelerate?

In this Fail Faster episode, we speak to Bryan Lane, a writer, speaker, and technology advisor focused on using data and AI for the public good.

Bryan is currently serving as the chief of business intelligence at the FDIC. Prior to that, he co-founded the government-wide AI Center of Excellence at GSA, where he supported organizations like the DoD Joint AI Center. While in government service, he has developed nearly $1B in AI and data-related contracts and deployed enterprise-wide capabilities to hybrid cloud platforms. Prior to joining the federal government, Bryan led product management for a software firm that was strategically acquired for $250M and was an analytic consultant for IBM. He also served as an active duty Marine and Arabic linguist, with deployments to both Iraq and Afghanistan during the Global War on Terror. He routinely writes on AI, data, and society and has been featured in widely distributed Medium publications such as HumanParts, OneZero, The Generator, and DATA XD.

Podcast transcript

Khushboo: Hi, Brian, welcome to the Fail Faster podcast. How are you today? 

Bryan: I’m doing very well. Thank you. Thanks so much for having me. 

Khushboo: I am super excited to be having you here and share your stories and journey around some interesting topics, you know, like talking about AI, which is the talk of the town. And also, you know, like some interesting topics that we chose for this episode, specifically, like how to bridge the gap between tech connected and tech disconnected and also some regulatory stuff. So we’re just very excited to be in this space. Now, before we get into these topics and talk more about that, why don’t we start with a quick introduction, quick background about you? Like, tell us where were you born and raised? What was childhood like? And what are your family dynamics? 

Bryan: Well, it’s a long story. So I’ll do my best to make it interesting. So I was born in Northwest Indiana, just outside of the Chicago area. I come from a blue collar family. My mother is a steel worker, and my father is an iron worker. I have a big family. I have four brothers and sisters, so three younger sisters and one younger brother. And being the oldest, I started babysitting at a very early age. I was, you know, babysitting from the time I was 10 until when I moved out of the house at 17. 

We had a very, you know, stereotypical blue collar family life, you know, it was it was people were always working, were always working, and then just trying to make it through school. And, you know, shortly after that, after after school, I joined the Marine Corps, and I spent five years in the Marine Corps as an Arabic linguist from from 2003 to 2008. And then eventually made my way into the world that I’m in now for technology and data and artificial intelligence and all of that. 

Khushboo: Awesome. That’s a great story. And how did you get interested? So when you were growing up as a teenager, and when you had that option of choosing a career for yourself, or building a career, you know, like, what was the start? So my question is, like, what inspired you or what fascinated you to be in technology? 

Bryan: Oh, that’s a really great question. And I had, I had a very nonlinear pathway to get where I am today. Right. And so as a teenager, I had a lot of hopes and aspirations to be a superstar DJ and, and sound engineer. And so for, you know, my late teens, and even into my early 20s, I was traveling between Wisconsin and Chicago and in the places in Indiana and Ohio, to do DJing and live sound. And that was really, that was really an interest of mine. And, you know, I had planned on becoming a sound designer. 

And through a chain of unpredictable events, I found myself in a situation where I needed to make a decision on what my future was going to be like. Yeah. And so I decided I decided to join the Marine Corps. And it’s actually a funny story, because in my hometown, the Armed Services Recruiting Station, all the services were in the same office. So the Army, the Navy, the Air Force, the Marines, the Coast Guard, they were all there. Right. And this is funny and thinking about it. And so the Army has a very unique tactic of recruitment, right? They, I would describe it as much more of a, you know, much more of a hard sell, they want to throw money at you and get you down to the enlistment, enlistment processing facility as fast as possible. And they knew that I was talking to the Marines as well. And so the Army showed up at my house with someone who had used to be a Marine, and they, they jumped ship and went to the Army as an inter-service transfer. 

And they told me all the reasons I should join the Army. And, you know, I talked to all the services, I talked to the Air Force, and they try and make everything sound like college, right? They said, Oh, you’re going to live in a dorm. And you know, you probably won’t have to go to war. So don’t worry about that. And we’ll pay for your college. And I talked to the Navy. And, and when I met the Marine recruiter, I mean, he was, he was a picture perfect Marine. He was six foot one, flat top, big hunk of chew and his, and chewing tobacco in his mouth, even to this day, I guarantee you can still do 20 pull ups and run a 15 minute, three mile run. 

He was just like that, like the poster child of the Marine Corps. And as soon as he showed up to my house, I knew that that was the decision I was going to go with, because they talked to me in terms that I could understand, right? I wanted a challenge. I wanted to do something meaningful. I wanted to, you know, I wanted to not only change my life, but also to help others around me and more importantly, help my country. And so through that, that was my first, my first significant decision on what would eventually guide me towards a career in technology. You know, and so as part of the Marine Corps, we worked with a number of technology and analytic tools when we were doing translation. 

This is before machine translation was even a commercially available tool. So we were still doing a lot of manual translations of digital media and physical media, you know, everything from radio broadcasts to, you know, to, to pamphlets and magazines and business cards and, and everything you can think of, right? And the work never ended. And so when, when I got out of the Marines, I had made a decision to never translate anything again, but for my career, I got into the world of acquisition and procurement or technology. And so initially, this was still supporting the Department of Defense for a organization called the Joint IED Defeat Organization. 

And they had a very noble mission, which was to find roadside bombs before they blew up on the battlefield to protect both soldiers and civilians alike. And we, you know, we really evaluated and bought a lot of technology to help with everything from, you know, bomb sniffing dogs to very advanced radar systems and everything in between. So I really got a very broad exposure of technology that created data and got to work alongside users that consumed many, many different kinds of data to make risk assessments on where they thought bombs might be, which for its time, you know, I date myself in, in Python versions. 

So this was, you know, Python 1.2 era where, you know, open source technology still was not widely adopted. In many cases, we didn’t even have it in government organizations. And a lot of the big data at the time big data analysis was done in a pretty manual process. We didn’t have a lot of automation. We didn’t have a lot of smarts. We didn’t have a lot of, you know, models and methods to, to kind of take a shortcut and through the analytic process or augment the analytic process. So after a number of years working in the, you know, in the analytics space around trying to identify roadside bombs, I eventually moved over to IBM. 

I was a consultant there and we did these very large scale technology evaluations for the Pentagon. And then eventually, I worked for a software product management company that specialized in large scale geospatial data management. And so we’re talking, you know, 10s or hundreds of petabytes on a daily basis. And then eventually, that company was acquired by a, by another firm for 250 million, I think, in 2018. And, shortly after that, I joined Federal Service, I was at the General Services Administration, I helped stand up the Artificial Intelligence Center of Excellence in 2019. And I partnered with the DoD Joint Artificial Intelligence Center, where we focused on building the first department wide hybrid cloud AI development platform, which was very exciting. 

And then found my way over to FDIC, where now I lead and manage the business intelligence services team that with supporting the development of enterprise, corporate data and advanced analytics and AI adoption and all the good stuff. 

Khushboo: Yeah, all the good stuff. And what a journey like it’s beautiful, like how you’ve like, you know, how you started to, you know, where you are today. So now again, you know, taking back from where you started to fast forward you as you know, Chief Business Intelligence Services at FDIC. Now, if you have to tell me like, and in all those like, you know, you started a company and all that great stuff. If you have to tell me one or two stories that are truly your biggest wins, something that you’re really proud to share today on that, on this platform that you know, this gives you like, very like, this is a proud moment for you to share this story from your experience so far. 

Bryan: Yeah. Oh, Khushi, that’s such a great question as far as biggest wins. Okay. All right. So I would say, one of my most significant accomplishments or achievements or efforts that I had been a part of in government, when we were at the General Services Administration, and partnered with the Joint AI Center, one of our very first meetings with the executive leadership team at the JAIC was, they had a technology portfolio where they had invested in humanitarian assistance and disaster, disaster response, artificial intelligence technology. And so this was technology that would, that could be used to monitor fires and flood damage, as well as doing damage assessment, and, you know, very sophisticated computer vision tools or for search and rescue at sea and many, many different humanitarian assistance activities. But this was a case in which because the technology investment was funded under DOD authorities, it wasn’t available to state and local governments or, or first responders on the ground. 

And so in this meeting, you know, the I think the JAIC leadership, they made a really astute observation, which was we’ve invested a ton of money into these capabilities, and the people that need to use them the most on an annual and seasonal basis, don’t have access to them. So we want to figure out how to give them away. And so this was a combination, Khrushchev, working the, you know, the authorities and the policy that go along with a very large like government bureaucracy, and also, you know, doing some technology education for the first responders that may not have access to the latest and greatest analytic tools from a data and AI perspective. 

And so, you know, with half a dozen, half a dozen meetings with lawyers, and, you know, a lot of brainstorming with with research labs and civilian agencies, and, and many, many partners and stakeholders, industry partners from a number of big tech companies, academia, first responder organizations like CAL FIRE, and the Pacific Disaster Response Center, we got this whole group together, and we established a plan for as a way to to give away this technology to be managed by a civilian agency with access to first responders and the state and local governments in time for, for the seasonal wildfire threat the next year. And so this was very significant, like, we had a disaster response summit at the White House, and we brought all you know, it was a challenge to bring all these groups together with different priorities and authorities. 

And, you know, in some cases, competing interests, but to convene everyone, and to get them to agree that the first responders on the ground should have access to technology was, it was a significant accomplishment for me, and just a way to demonstrate, you know, for one, not everything is a technology problem. And for two, for the things that are really important, you can get the groups, competing stakeholders together, or stakeholders that are not aligned together, and find common ground to do something good for the whole country. I think that was a significant win for us. If you want, I can give you a second one that’s a little bit more technology based. 

Khushboo: Sure, let’s go. Let’s talk about that. 

Bryan: Okay. So when I was working in geospatial data management, we were one of the first people to, or one of the first companies to use the Amazon Snowballs as a disconnected cloud service capability. And so this was probably, I don’t know, 2016 or 2017. We had a host of customers that had very large scale imagery servers, you know, in countries all over the world, and, you know, in the Middle East and in the Pacific and, you know, in Europe and South America. And when our customers would go out to different parts of the world, they were taking these massive, like server racks in a shipping container, right? 

And so they would like load up the shipping container with everything that you need to run your own private data center. So hardware, power, networking, cooling, everything, and they’d put it in a shipping container and drop it into, you know, whatever country they were going to, and publish maps, maps and imagery data from that data center as kind of the local, the local server. And so when that when the Amazon Snowballs came out as a, you know, as a disconnected, disconnected solution for very large storage that had cloud native services built in, we immediately we started benchmarking and with geospatial, you know, when you’re doing things like maps and globes and many different imagery layers, you know, you take a collection of files, but as you create, as you create zoom levels at different resolutions, your file storage grows exponentially. 

And so this was our first real use case of how to use object storage, and cloud native services to serve up, you know, petabytes of data. And so we did a ton of like, very, very, very technical benchmarking of like, what is the optimal size of each object, because there’s, you know, there’s some fixed IO cost with, you know, each individual object being opened up, and then, you know, and then how do you stitch them all together in a mosaic that you can serve over web services on a local network. And it was a really cool time for us to do some very, very hardcore and engineering and benchmark testing on a new capability that hadn’t really been tested for this particular use case. 

And I think ultimately, we were able to kind of shrink down that the big shipping container data center into a couple of network together snowballs, you know, for a few $1,000 a month subscription, as opposed to, you know, a multimillion dollar package that comes along with, you know, and engineers to manage the whole data center and the logistics of shipping thousands of pounds of gear to an operating in another country. So it was a very, very, very cool time for us to do something innovative and forward thinking for the first time as, as they were rolling out these new disconnected cloud capabilities. 

Khushboo: Wow, that’s amazing. And you know, you you mentioned disconnected clouds. And that reminds me of one of the topic that we wanted to talk about, which is gap between tech connected and disconnected, pretty much. So why don’t you share your thoughts around that? 

Bryan: Yeah, oh, well, now we’re getting into something that I think is, you know, part part of the reason that I really put a lot of emphasis on being a civil servant and focusing on how technology can benefit the public. And so that disconnected can mean a couple things, right. 

And I think in the last use case, it can mean, you know, you’re literally disconnected from a network and from resources, you know, compute and storage resources, and you’re kind of operating autonomously. And then there’s this idea of, you know, the tech disconnected slices of, of the US the population that does not live and, you know, things like the Silicon Valley bubble or, or in the IT ecosystem. 

And so, you know, I really, I really think about this cushy a lot in terms of, you know, how the changes to technology today are impacting the people that are not following it every step of the way, like, like IT professionals, and like, you know, engineers and everything. And so, you know, I have, I guess, kind of a goal to raise awareness of how these changes could impact people’s lives, as they’re happening. And then what I’m finding is like the number one, the number one thing we have to address first is make sure making sure that everybody has a common baseline knowledge of things like, you know, how these artificial intelligence tools are working, and, and what what it really means to deploy something at, you know, at massive billion person scale, because a tiny, a tiny change can have a big impact, not only across America, but across the world. So that’s really what I think about in terms of the tech disconnected. 

Khushboo: Awesome. And, you know, talking about AI, like, you know, which definitely is a turning point, but not only as a tool, but for regulation. So tell us, like, how can we make the public’s life efficient and better with use of risk and governance? 

Bryan: Yeah, that’s, so that’s a great question. I’ve seen, I’ve seen a number of technology, technology investment efforts in the federal government. I think there’s things like, some things that are very straightforward, like digitizing paperwork related processes, you know, there’s like a lot of eligibility determination services that have been digitized and turned into API driven web services. So that way, people can do things online, as opposed to going down to an office and filling out forms. There was a period of time, and I think the GSA is still doing this, they were using innovation challenges to identify use cases for artificial intelligence, and that benefit the public. 

So one of the things that we had looked at together when I was there was, we were looking at the response to the COVID, the federal government response to the COVID pandemic. And so I think our team had some, some real aspirations on how we could use AI to understand how we responded to COVID. And I think as we started interviewing, we interviewed a bunch of stakeholders from different agencies. 

And, you know, as we were conducting these interviews, we heard the same pattern pop up over and over and over again. And the pattern was, because of the pandemic, we had to change X. But in order to change X first, we had to update our policy. Yeah. And that was really interesting to us then because it didn’t matter if it was the, you know, the State Department who cared about who cared about the the office visit procedures for every embassy across the world, because they had to adjust based on, you know, based on the COVID, the COVID density and what the what the rules of the host country were and what the State Department policy was. So each embassy was a little bit different, right? That was one use case. 

But then you talk to a place like, like NOAA, the National Ocean Oceanographic and Atmospheric Administration, they basically said, you know, we had a six month gap where we had, we had no ground truth inspectors in the fisheries. And so we had a, you know, a six month window of unregulated fishing that was not verified by a by a ground truth observer. But in order for them to implement that, they had to, they had to update the policy first. And same thing with the State Department use case and all the other agencies that we talked to, they said, we were doing this response to COVID, this change in our operational procedure or whatever. 

And in order for us to do that, we need to first communicate what the new policy is. And we saw this as well, during my time supporting the COVID Task Force in the Department of Defense. And so they had policies that were related to biomedical preparedness, and responses to, you know, how they collected data in a pandemic environment or a biomedical elevated biomedical risk environment. And so each week, they would update policies and procedures. And then their concern was, okay, in 2020, it’s the COVID pandemic. Before that, it was Ebola. Before that, it was, you know, H1N1 and swine flu and all these other things. 

So how do we make sure that historical knowledge isn’t lost when the next pandemic comes so that we’re not starting from zero and, and all the lessons learned that we have are discoverable for the next use case. And so that was really interesting for me in terms of, you know, what the common thread across all those agencies are, which is, first, we change our policy, then we change our procedures. That opened up a discussion with a fantastic team. 

And there’s still people that are doing this work today in the Air Force and in the Department of Defense team for an application called Game Changer. Game Changer was a, it was, it was the most concerted and successful effort, I think, to treat policy as though it were the configuration files of an organization, right? And just how you, how you tweak your system configurations to get a certain kind of performance, you can tweak your policy, you know, your organizational policy to get changes in your performance as well. 

I think the Game Changer team was one of the first to really put hands to the keyboard on how to build an application that does that. And so, you know, in some ways, it was a data science experiment of what information can we extract out of, can we extract out of policy and relationships between different policies and different organizations. And then the important piece, which is how do we get that information to end users, you know, that are either policy consumers or policy analysts or performance analysts that are doing impact analysis across organization. 

And when you’re a large organization, like the Department of Defense, performance is very important, because those minor tweaks to your organizational system can have, you know, can have ripple effects that are tens or hundreds of millions or billions of dollars. And it also helps your, you know, helps your organization perform, you know, both more efficiently, and more effectively. 

Khushboo: Awesome. And I know we talked about, I mean, you talked about the threats and, like how to make the public’s life better, avoiding those threats, but like, what are some of the key concerns or challenges that have led to the call of increased regulation of AI technologies to fail? 

Bryan: Great, this is great. I’m so what, can I ask you a question? 

Khushboo: Yeah, sure. 

Bryan: If you had to pick a side today, yeah, would you be team accelerate or team decelerate on the artificial intelligence front? 

Khushboo: Team accelerate, I guess. 

Bryan: Okay. All right. Team accelerate. Yeah, I think there’s a battle going on right now between those two teams. Maybe it’s not a battle. Maybe it’s more of a skirmish, right? I think that there are a lot of concerns about harm. I think there are some valid concerns about systemic harm. I just had a conversation with an executive and very, very early internet founder, who I respect very much. 

She was talking to me about, you know, the different categories of harm. And, when we talk about, you know, the need for regulation and, and, you know, the potential for harm of AI systems, that kind of conjures up this idea of, you know, like the, like the singularity event where the robots take over the world. But I think, I think there’s a spectrum. 

I think the probability of that, you know, that event is very, very low. And the reason for that is because the fundamental assumption there is, it would, it would happen so suddenly that humans would not have a chance to respond, and that humans would be kind of these hapless, hapless NPCs in an environment where, you know, robots are making all of the decisions. 

I just, I don’t foresee those two things aligning at the same time in order to create this kind of existential threat to our threat to the human race. However, you know, there are some, there are some harms that we do know about, right, there are harms related to systemic bias, or, you know, filtering out groups of people or, and this goes back to, you know, this idea of tech disconnectedness, there are harms related to exploitation of, you know, people that don’t understand the technology. 

There’s harms related to, you know, the possible replacement of, of, of different kinds of jobs. You know, I’m an econ nerd by training. And so creative destruction is not very scary to me, except for in the sense of when, you know, when the affected people don’t see it coming and don’t have time to adapt and update their skills. But I think in a lot of ways, that is, you know, that that’s part of innovation, when changes happen, I think a lot of, you know, a lot of people, they adapt and, and the shape and the skill set of the workforce changes along with those innovations.

So anyways, there’s, so there’s this spectrum of what those harms could be. And I think, you know, really focusing on the ones that we know are pretty high likelihood, you know, will be an overall benefit to the public and to those affected. Now, I think the net positives, right, the net positives are, especially with some of these, you know, some of these more modern artificial intelligence tools, is, you know, you could think of it as the, you know, the overall cost of intelligence going down, because you have a virtual assistant that has, you know, a very broad range of a broad range of, of tasks and problems that can help a human solve whether it’s, you know, coding with a, you know, with a virtual agent that is based on a large language model, or whether it’s, you know, doing some kind of intelligent automation. I’ve heard people frame it as the cost of intelligence going down. 

I like to think of it as like the floor for the average baseline intelligence going up. Because, you know, as these tools become commodities and become readily accessible, it’s kind of the tide that that helps all boats to rise, because everybody is operating then on a higher baseline of knowledge that they can build upon. 

And so, you know, I, I agree with you, Kushi, I’m probably on the, on the team accelerate, you know, I’m cautiously optimistic. I think there are harms, I think that the one, you know, my one biggest fear is, is more related to language and culture. And I think that, you know, the normalization of language through commodity systems, and what language is proper versus not proper, will have an effect that, you know, that that neutralizes or, or bleaches some of your, you know, some of each individual’s culture, because they’re incentivized, and they’re continuously like, nudged and shaped into, you know, what proper languages, but I’m also very hopeful that, you know, there are people like, or, you know, there are groups of people like artists, and, you know, cultural influencers that, that understand how important it is for, for people to maintain their uniqueness and their identity. And so I’m, I’m getting less and less worried about that risk at a totalitarian kind of scale. 

But I do think every time you hit have auto complete, in an email, you know, it does something to your own unique individual style communication. I think there are there are organizations that are that are addressing this right now, I think the publishing platform medium has taken a lot of really pro human positions on on what it actually means to, to be an authentic human that is that is generating content versus versus artificial intelligence generated content. 

And I think they’ve done a pretty good job on defining, you know, the importance of the unique individual voice instead of, you know, creating content that is most applicable to the broadest audience possible. And I think we’re going to see a lot of that conversation happening, especially in the arts and culture community over the next handful of years. 

Khushboo: Yeah, I 100% agree with that. But thank you, Brian, for coming on the show and sharing these amazing insights. I did not even realize that we’ve, you know, crossed that 35 minute mark, but that’s a sign of a good conversation. So really appreciate you coming here today and sharing all these important, I would say, nuggets, you know, that our listeners can take away something from. So thank you. Before I let you go, Brian, one last thing. If people want to reach out to you, or if they want to know more about the work you are doing, where can they find you online? 

Bryan: Oh, yeah, great. Thank you so much for asking that. If you want to find me online, I can be reached at, at Bulletproof Bri on Twitter and Medium. I publish a lot of my writing on Medium. And I’m like medium. I’m moderately active on Twitter. And then by email, I can read be reached at Bryan, b r y a n, at data xd.io. 

Khushboo: Awesome. Thank you so much. And I wish you nothing but the best Brian. Thank you so much. And I really appreciated our time together today.