Fed Gov Today’s Post

Transcript

John Gugasian, great to see you again. Thanks for your time today to terms that people are really talking about pretty actively in the ocean space right now, authenticity and reliability, how is achieving certainty on each of those evolving over time with the proliferation of data volume increasing and so on? Yeah, I think there's a couple different ways you can look about look at this. Um, one, um, you want all data, whether it's, you know, authentic and authentic meant to be biased and by someone because information is meant to influence someone. Once you're influenced, you act and action is reality. So by removing some of those potentially inauthentic sources, you can bias your analysis unto itself and this things. Now that said, really you want to be able to tag that information as well. I want to be able to say this is, you know, authentic information. This is potentially a bot. This has a bias toward this type of approach of something so that users are the Olsen data can. Factor that into their analysis. But by looking at the whole corpus of knowledge, we're going to get a much more accurate view of the world and then be able to filter out what's reality, what potentially isn't, and give all aspects so that you can give what if scenarios. Again, we don't want to bias ourselves because we think something else is biased. And then again, we poor information. We skew analysis. We provide information to decision makers that could be erroneous. The way that we're approaching that is both. Your combination of human tagging, of information analysts looking at information, looking at sources, are these reliable sources? Are they government controlled sources? Are they? Leaning one way or the other from a source level. And then there's also some promising technology out there that can automate some of this. I haven't seen that golden nugget yet that says, hey, we're going to automate the whole thing. That's where teams of experts come in and a lot of it to help as well. But there we are making advances in that that I think will be more viable in the next year, 18 months or maybe a little bit longer. I want to talk about what you think the over the horizon view is in a few minutes, but I wonder what kind of complexities are. Introduced into what you just described by the fact that government organizations are using more open source data than they ever have before, but they're also able to collect more proprietary information than they have before. And so they're combining that stuff at a greater rate than they ever have. What does that look like from both the government perspective and the industry perspective about whether it's necessary to keep them separate while still combining them, if that makes sense, and how one tags information and so on. So a number of years ago. I had a meeting with some folks from the Director of National Intelligence and they said, hey, John, what would be your kind of perfect view of how you want to alley things? So I very much said almost exactly what you just said. I wanna run open source analysis purely the same algorithms on open source data, on classified data and then combine the two. Because I think really by that combination of things is really where you going to find truth, if you will, right and true knowledge to bring that information together. I think the promise. Here is again tagging all the information, whether it's again from an open source solution, how it is again biased or nonbiased, including that in with classified and other proprietary data sources to come up with what that real answer is, what that real view of the world is. As a former intelligence community officer running analysis as well, I know oftentimes the intelligence doesn't write about happy things. We're looking at things that are potentially problems for the country. And problems for our security. So by combining that other view of the world as well, you get a much more fulsome picture. And today with the ability to gather the open source information at a much higher rate, to be able to tag that information automatically and its speed and the ability to move that across domains and integrate that with proprietary or classified information. I think the promise of where we're going is the promise of again, more true knowledge into the future is just. That much greater, I mean it's exciting time to be in this of business. All right, as I mentioned, let's look over the horizon. What do you see future wise, either technology that can be applied that maybe exists or but isn't mature today or maybe doesn't even exist, but you know would have potential use in the future, culture changes that you may see on the horizon, whatever. How does this space look different in your view, 2 years out, five years out, maybe even 10 years? Let's talk the culture change first. Maybe, uh, well, that's harder. It may be easier to explain. You know, culture changes is tough. I think 2 submittal documents have come out like we all know over the past six to nine months, the DNS open source strategy and also the Dodds open source strategy or commercial strategy. Really, you're setting the tone, I think to drive the culture that open source information is the first source of information. It always wasn't that way. We know that it was. Kind of look at this thing off to the side that maybe I can collaborate some of my classified intelligence with that. So the cultural change of that mindset change where I think, and that will take time, right? That'll take time to sync in. That will take time to get policies around and get funding around it as well. But I think it's that first step that we're on that over the next five years, you're going to see more and more of that being used from a technology standpoint. Kind of where where there's a large focus, rightfully so on large language models and technologies like ChatGPT impart, which have tremendous potential to revolutionize not just this. Uh, discipline, but you know, number of disciplines focusing, honing that the developing the right type of prompt engineering models around that will help, I think greatly speed the analysis that we can do on open source information. But that's not the end. There are other things that need to be developed that look at how do you look at clustering algorithms? How do you look at things like? You're more traditional natural language processing and beyond that, how do you bring together both structured and unstructured data to look at together? Again, large language models are that based on language, based on, again, written word. How do we look at those things that maybe aren't, that are more of a structured data set that you can bring together into these as well? A lot of research left to be done on leveraging all this data together so that you can answer. Things answer questions from content that is both, again, your traditional open source information, news and social media and things like that. But then how do you bring that together in a seamless environment with things like ad tech, with things like breach data from dark web and things like that, and developing the approaches and the algorithms to really look at all of it together to present information to analysts so that they can incorporate that with their other content? Also, John, it's great to talk to you. Thanks for your time. Thanks again. Good to see you.