Rethinking Intelligence with Dr. Shay Hershkovitz

(upbeat music) - This episode we're joined by Dr. Shai Kirstibans, a professor at Georgetown University and a senior analyst with the RAND Corporation. Prior to joining the Academy, Shai served as an intelligence officer at the Israeli Military, and he was an executive at several private sector, technology organizations and companies, including the Ed Spreys Foundation. He is sensitive extensively about intelligence matters, particularly how intelligence agencies need to adapt to the nature of the art. His most recent book is entitled, The Future of National Intelligence on Merging Technologies Re-shape Intelligence Communities. In this episode, we will discuss the current intelligence landscape and key topics for Shai's book. (upbeat music) - Shai Hirskowitz, welcome to Intelligence at the Edge. I am so excited that you could join us today. - Thanks for having me. - You know, your book came out in 2022, and I read it while I was still at CIA, and it stuck with me then. And when we were deciding to launch this new podcast series, you were one of the first people I thought of. I thought we had to get Shai on to talk about his book. And that book, for those of you who haven't read it, I urge everyone to do so. It's called The Future of National Intelligence. I think you wrote it in 2022, and it's a fantastic piece of work. I really wanna talk about many of the ideas that you surface in that piece. But before we do that, as I always do, I like to get a little bit of background on you, Shai, for those of you, those in our audience who haven't met you. Tell me a little bit about your early career, and what you did in Israel, and what made you decide to go over to Georgetown? - Yeah, so I'm originally from Israel, I have been living in the US for almost a decade now. Like most Israelis, I was drafted to the military at the age of 18. But even before that, I studied Arabic. It's cool, I studied English, I speak German. So languages have always been my thing. And also, I was a geeky kid, red books. I was interested in politics, and intelligence. And so you might say that intelligence was almost my calling from my teenage years. Even though, and my mom keep mentioning that, my dream growing up was to join the Foreign Service. But I ended up working with Israel Intelligence, spent a good chunk of time there. I had the opportunity, basically, to see the various facets of intelligence. So I worked for collection units. I spent a lot of time in the research and analysis division, was involved in military or intelligence operations. So, and of course, I worked very closely at some point with decision-makers, whether it's military, but also political. So I had a privilege to see the various facets of intelligence in how this giant machine, basically, plays together, all the various components. So it was kind of a natural transition from me, moving from the secret world of intelligence to the more open world of geopolitics and intelligence as part of that. - And what motivated you to write the book in the first place? - That's an interesting question. I, reading and writing has always been my thing. And I feel like the luckiest man in the world because I can do that for a living. And since I learned how to write, this is basically where I started writing, pros and or fiction and non-fiction and things like that. But I think it was when I joined the X-Prize Foundation back in 2017 or 2018 as the head of research, when I really started researching various topics, technological topics, under the title of the future of. So I did a lot of studies titled the future of forest, the future of this, the future of banking. And then I started thinking, hey, maybe I should write something about the future of intelligence. And then that was the beginning and then everything went from there. - So in your book, the future of national intelligence, you argue that Western intelligence services around the world really need to come to grips with the need for structural reform. Briefly can you outline the reasons for that? - Yeah, I think that people have been talking about the need of intelligence communities to change pretty much since the early 2000, with a seminal work by Rand, called the revolution in intelligence affairs and so forth. But I think today is kind of the perfect storm, if you will. So we live in mentality of austerity on one hand, right? We need to cut down on costs and so forth, including spending on national security. We cherish efficiency as part of that austerity mentality. We are also, I think the government is now more open to tech, maybe too much even, but still this openness is also an opportunity. And we also, we question all the certainties, all the assumptions, sometimes for better and sometimes for worse, but this is the zeitgeist. And given how volatile the world has become and this Gutenbergian revolution that we're experimenting, experiencing with AI and data and we'll touch on that. I think this is the perfect storm. And so we have to make major changes in the intelligence community. But I would argue that the cornerstone of that reform or change is they need to adopt or to take concepts, processes and technologies from the tech sector and recreate the profession called intelligence by adopting these new approaches. And it's not just making adjustments, it's a revolution altogether. - So let's start on the technology side of things. In your book, you're outlined several and we talk about many of these all the time here at the Special Competitive Studies Project, but there's quantum, there's AI, there's machine learning, et cetera. Which are the ones that you think are the most important for intelligent services to focus on in terms of transformation and why? - Yep, that's a great question. So in my book, I cover various technologies. I start with internet of things and of course related to that edge computing. I talk about how we transmit data. So 5G, 6G and of course the next generation, how we store and process data. So data is of course one of the major elements of this revolution. And then we have artificial intelligence and then we have quantum computers who will take everything pretty much to the next level. And of course we have blockchain which has the potential to redefine how we share information and control information and secure information. And if we define intelligence in the context of data and information and knowledge, we see that this revolution needs to happen all across the board. So take for example an analyst in intelligence analyst. Think of them as working without the internet. And now we are in the new internet moment in terms of the revolution that AI and big data and all other technologies basically bring. So when people say we shouldn't use LLMs for example, I would argue that it's like saying that we shouldn't use internet, you know, if we go back in time 25 years, then imagine an intelligence analyst working without the internet, right? Without databases. So to me, of course we don't need to adopt technologies as is, we have to always be critical. But we cannot ignore that and we should harness that and we should tailor how we work, how we think, work processes, structures and so forth in light of that revolution without ignoring that. But to your question, if I only have to choose one technology, it will probably be AI because I think AI is the cog of the machine. It has implications to how we collect information, how we process it, also of course the application layer, how we generate data, how we generate knowledge, and everything basically falls under the AI umbrella. But the main point here to emphasize is not whether a single technology is more important than another one. It's about the convergence of these technologies. Exponential progress is usually made by the convergence of various existing and emerging technologies. So when you bring together IoT and 6G and AI and data and blockchain and all of that, that's basically what creates the big revolution I'm talking about. So it's not an incremental change. This is a sea change moment. - Yeah, I couldn't agree more. And I'm really pleased that you're mentioning blockchain because I think a lot of people just think about blockchain is related to cryptocurrencies. But it's so much more than that. I mean, it's gonna basically create a new secure way of exchanging information that intelligence services can both benefit from and also will need to figure out how to deal with as our adversaries make use of that technology as well. So technology is one driving force. The other that you refer to are some of the new non-traditional threats that Western democracies are having to confront disinformation being one of them. But can you talk about what others are out there and why they're important? - Yeah, so that's a really important question because what we're seeing isn't just new threats. It's a new threat environment altogether. And we see the rise of the cognitive domain as another domain in the battlefield. So it's not just a peace war dichotomy and it's not just limited to one specific domain, air, land, space, and so forth. The cognitive domain basically informs decision-making all across the board. It determines how we see reality and how we act based on our understanding, whether it's in the air domain, the land domain, cyber and so forth. And so traditional intelligence methods were basically built for a world where adversaries were mostly states, secrets were hard to get and truth was relatively stable, relatively stable. But now we're in a world where any actor and it could be state, non-state, even individuals, they can shape the information space. They can manipulate public perception. They can deploy AI tools that could be used for various purposes, essentially focusing the struggle or the combat, if you will, in the cognitive domain. Take, for example, disinformation. It's a fast, decentralized, and it doesn't require or doesn't respect borders. And it spreads through civilian platforms, not necessarily secure channels. And it's not just about lying, it's about eroding trust, it's about blurring facts and fiction, and basically destabilizing open societies from within. And that is not something that intelligence agencies can counter with surveillance or covert collection alone. Because it requires intelligence services to work more openly and often to collaborate with civil society and embrace new platforms and work hand in hand with the public. So now to that, let's add the technology arms raised, which adds another layer or another complexity. So we're talking about AI, we're talking about quantum, we're talking about synthetic biology. These fields evolve much faster than governments can regulate, let alone fully understand. And I see that time and time again, working with governments around the world. And intelligence services now, I would argue, have to monitor not just adversary capabilities, but also emerging tool use technologies, for example, that may not be weaponized, but could shift power dynamics very quickly. So there's another layer to that, which is the technology arms raised, which adds more complexity. Because we're now talking about AI and quantum computers, and we talk about synthetic biology and other emerging technologies. And these fields move much faster than governments can regulate or even understand. And intelligence services in that context, they have to monitor not just how adversaries are using these technologies, but also emerging tool use technologies that may not yet be weaponized, but could shift power dynamics very quickly. And there is a, in my opinion, great risk, in let's say, scientific collaboration with foreign entities, because you can assume that you basically work on civilian technology, but that could be easily repurposed for military purposes. So I think that the core challenges here in this day and age is the speed and the complexity and ambiguity of all these threats, which basically require intelligence agencies to adapt in you way of thinking, kind of departing from traditional approaches and embracing new ones. Yeah, and adding to their job jar, I mean, it strikes me that intelligence services still need to protect against the traditional threats be it for military power, cyber, terrorist threats, but now also need to worry about what's happening in Chinese AI labs or synthetic bio labs, right? I mean, it's an addition to. So that kind of leads us to change in processes and change the structure that you talk about in your book. And you outline the traditional intelligence cycle, which for our listeners who may not be familiar as sort of this wash, rinse, repeat cycle of you, you get intelligence requirements from your customer, you then go out and develop a collection strategy, you collect the intelligence, you analyze it and produce it in the form of finished assessments, deliver it to a customer and they provide feedback, sort of an old school approach. And you've outlined some lessons learned from the war on terror and other recent experiences to describe how that's going to change. Can you talk about that a little bit? Yeah, absolutely. So, you know, traditional intelligence cycle was designed back in the '40s and even discussed, even earlier in the '30s and even in the '20s. And it represented an assembly line-like logic, right? You know, there is a clean linear flow as you just described. So, requirements, collection processing, analysis, and dissemination, and then, you know, and the cycle starts again. But that makes sense when information was scarce, where threats were slower moving. And collection was the most valuable bottleneck. But that model doesn't work as well today. We're now in a different world, a world of, you know, information abundance, not scarcity. And threats emerge in real time through, for example, social media or open source data or AI generated content. And intelligence professionals are expected to respond faster, more collaboratively. And with much shorter distance between insight and action. And I think, as an example, to make things slightly more concrete, think about the cyber domain. Let's say you want to initiate some sort of a cyber attack or, you know, cyber collection, if you will. What is it exactly? So, when you penetrate, you know, an information system, what are you doing? Is that an intelligence operation? Are you collecting information? Clearly, you need to analyze information in real time. And that kind of-- that's an example of how the borders of the intelligence cycle, the very silos, become kind of intertwined. So what's replacing the linear cycle is something much more dynamic. It's much more iterative. And it's much more network. You know, think of it as a living ecosystem rather than a straight line. Because analysts, you know, they collect their own data. And policymakers, they interact earlier and much more often. And AI tools continuously process streams of open and classified data. And so the feedback loop basically happens in near real time. And sort of a blurring of the lines between collector analyst and policy consumer, right? Everyone's sort of in each other's patch. With policymakers, both acting as collectors, as analysts, collectors having to analyze the collection environment that they're operating in, and analysts having to also think like a policymakers actually maybe take proactive steps to drive collection ways that they haven't before. I definitely felt that, definitely in the latter half of my career. So you've also talked about the need to, I think, open up the closed intelligence system. And I think you were kind of touching on a couple areas, sort of the relationship between intelligence services and Western style democracies and the public. But then also private industry and academia. Can you-- why is this important? Yeah. Well, it's important because the nature of the threat is evolving and everything becomes much more immediate. And the threats themselves kind of flow through-- let's call them civilian or non-military channels, such as social media and so forth. And in my book, I focus on three aspects of collaboration with the general public. And the general public could be academia. It could be ordinary citizens. So one is crowdsourcing. And the other one is presumption or consumer, the combination of production and consumption. So we have many examples in recent history where crowds were able to better analyze, predict, and sometimes even forecast certain events. Because the aggregated wisdom of the crowd is usually greater than the wisdom of a single or even a smaller group of people. And that has proven time and time again in very scientific experiments. The Good Judgment Project, Philip Tetlock, Super Forecasters is one such example. And there are others as well that are covered in the book. And I truly believe that there is a way to bring the public into the intelligence realm, or at least the domain in terms of the problems that intelligence agencies are dealing with. And as per public to assist, so that's one thing. And then of course, you have the presumption or consumer, which is a term that was coined some 40 years ago, mainly in the business world. But that basically the idea is to take ordinary citizens and evolve them in the production of the public good, i.e. the production of intelligence, or the stuff that intelligence agencies produce. And there are various ways to do that. But I would argue that the main thing that needs to happen here is full transparency. You don't want to start operating spies. You don't want to be portrayed as someone who's working from within society. You're harnessing the power of the crowd to promote your understanding of reality and to perform forecasts and predictions and so forth. - Boy, it strikes me that that's going to be difficult for at least the US intelligence committee, but perhaps it's going to be different in other countries. But the culture of secrecy is strong, right? And so the very notion of being transparent and open about even what the questions are that you're interested in is going to require quite a bit of a sea change in the culture of the intelligence committee. To me, it sort of speaks to the need to rethink or refocus on what is the purpose of an intelligence service. And I think you used a word earlier that I really like and that's delivering insight. And so if the goal is to deliver insight, the means can vary, and perhaps some of those means I'll do include crowdsourcing or some of these more transparent techniques. And I think that's a really valuable insight. - Yeah, and I think that maybe one of the most critical things that intelligence agencies should work on is how they define secrecy. Of course, there are things that you need to protect. Certain assets that you need to protect. But in this day and age, I don't think it's about secrets. It's more about mysteries. So complex questions. Wicked problems as the literature sometimes refer to. So wicked problems where factual information is less relevant or it's relevant in context. But it's not about the secret that you're trying to protect. It's about the mystery that you're trying to solve. And with these mysteries, you need to bring as much brain power as possible because these are complex issues. Big strategic, geopolitical changes. There's no one single source of truth. One source that will give you the answer whether the Iranian regime is stable or not. Okay, it's about sense making. It's not about secrets. And it's not about revealing sense of information. - And the idea that, you know, perhaps, a single senior analyst at CIA or somewhere else is going to be the source of definitive assessment as to whether Iran will remain stable or not. That seems part of this, right? That's going to change too. Policy makers won't necessarily be persuaded by just one analyst saying so. Also machines. I mean, that's the other thing that's happening. You've talked about it already. You've mentioned it quite a bit in your book. Now, you wrote that in 2022. And at the time, you were talking about strengthening. I think you used the word connection. The connection between the humans. And the machines, I think, referencing specifically AI. Since then, the term human machine team has sort of been coined and has gained currency. Is that what you were talking about? I mean, is, or is it something different? And then how do intelligent services go about doing this? How do they create the human machine teams in this connection? - So that's the million dollar or billion dollar question. So yeah, so in my book, obviously, I talk about human machine teaming, but that is part of a kind of a bigger approach or a wider approach, if you will. And I coined the term the five Cs, as in the letter C, that basically, in my opinion, should characterize the future intelligence professionals and especially intelligence analysts. So the five Cs are connected, intelligence professionals need to know how to work with machines, not just in terms of the interface itself, but also how the machine actually works. So statistical reasoning and inference and so on, they need to understand what the machine can do for them, but also what it cannot do for them. So that's the first C. The second C is collaborator. I think that the intelligence professionals need to work closely as part of their day-to-day routine, if you will, with civilians, with their peers, with machines, of course, with academia, with startups and so forth. They cannot no longer work in silos. The third C is critical and that is maybe the most important thing in the age of black box AI. Don't take AI output as is. Only if you understand how the machine works, can you become critical, okay? Being critical also is expressed in how you frame questions and then feed that to the machine. Then the fourth C is creativity. You have to think creatively about how let's say technology can help you, how it can make you better, how you can make better intelligence products, how to creatively ask different questions from various angles. And last but not least, the fifth C is content expertise. So let's say you're an analyst who covers a Middle Eastern country. You need to know the culture. You need to read Arabic or Hebrew or Farsi or whatever. And so you can't really say, okay, I'm gonna rely on machines. I don't need to be, you know, a subject matter expert. You have to be probably even more than ever. - I am so glad that you raised that because it's something that comes up a lot with students that I work with who are sort of interested to know what is it that they need to focus on. And one of the things I tell them is, think about what the machines won't be able to deliver. You know, it's that context, it's that cultural knowledge. It's the human stuff that you can bring to that human machine team. So I think that's really critical. - One thing I wanna add to that, because I think it's, it's, it adds some spice. If that's okay. - So I think there's a lot to be learned from the tech sector and for intelligence organizations basically to bring new concepts, new approaches. And when I look, you know, at the various roles in the tech sector and I think how can we translate that into the IC, you know, language, if you will. So take, for example, you know, NLP natural language processing researchers or computational linguists, that they could become narrative code breakers, right? In intelligence speak. So they basically decipher global propaganda or threats or sentiment shifts across languages. Or if you have engineers who specialize in knowledge graphs, which is the backbone of much of the LOM, of the LOM that we use. So you can turn them into link intelligence cartographers that basically they can map hidden relations between actors or events or assets across, you know, data silos. And you have synthetic scenario architects who are basically, you know, scenarios or simulation and war game experts. They can craft, you know, alternate futures, they can test policies, they can expose blind spots. So the bottom line here is that we need to take great ideas from out there, from the world, if you will, the external world outside of the IC, bring them into the IC and make these changes, create new professions, create new processes. It's not, you know, adaptation and the margins. It's a massive, you know, change of the DNA of how these agencies work. - Great point. And let's end it there, but I don't wanna let you go without asking you what advice you give your students, who are interested, perhaps in a career or national security or in intelligence. What pearls of wisdom do you give them as to what they should be focused on if they wanna succeed in this future world of intelligence? - So if I had to pick one thing, it's this. I think that national intelligence professionals or aspiring professionals, they must learn to master AI in AI is an umbrella term for a whole suite of technologies, but not to fear it, not to outsource it, but truly understand it. You know, we are, we're at the start of this, you know, AI revolution in intelligence. We don't even, I don't think we really understand how profound the change is. Let alone, I don't think we have any action plan in mind. But it is profound as maybe as profound as the arrival of the internet and maybe even more, okay? Now, so it basically means that students need to understand not just how to use AI tools. It means how to train analysts, how to educate them, analysts and collectors to understand what AI is good at, what it is bad at, and where human judgment still matters most. This is something that I think could be taught. They could be trained and it means building system that are transparent, accountable in our mission aligned. And I think that for aspiring intelligence professionals, they need to master that. Even if at the end of the day, they will do something that is remote from the deep tech. They need to understand how the tech works, kind of behind the scene, not at the surface level, to make best use of it. So that would be my advice. Don't fear the tech, learn it. I have a PhD in political philosophy. So as remote from tech is one can imagine, but at some point, even before writing the book, I kind of realized that if I really want to understand how intelligence works and how it should work, I need to go deep and understand the underlying technologies. And I would argue that AI is probably the most important thing because it kind of encompasses almost everything else. - Well, Shai, thank you so much for your time. I appreciate you joining us today. Professor Shai Harris-Kovitz with Georgetown University, his book is The Future of National Intelligence, how emerging technologies reshape intelligence communities. It's available wherever you buy your books. I encourage everyone to read it. Shai, thank you so much. - And JP, it was a pleasure. (upbeat music) (upbeat music)