Turkeys vs Swans, with Imanuel Portalatin


Photo: Imanuel Portalatin

Interviewer: Cameron Boozarjomehri

Welcome to the latest installment of the Knowledge-Driven Podcast. In this series, Software Systems Engineer Cameron Boozarjomehri interviews technical leaders at MITRE who have made knowledge sharing and collaboration an integral part of their practice. 

Despite our modern world, many systems aren’t built with uncertainty in mind. Alas, these unexpected events may become the new normal. Fortunately, Imanuel has been leading the charge to re-examine how we design and implement systems with new approaches that make them more resilient to the unexpected. Listen in as we explore how his work is helping the world prepare for the next natural disaster or global pandemic.

Click below to listen to podcast:


Podcast Transcript
Cameron: 00:15 Hello, everyone, and welcome to MITRE’s Knowledge Driven podcast, a podcast where I, your host, Cameron Boozarjomehri, have the great fortune of interviewing brilliant minds across MITRE. This week we’ll be discussing turkeys versus swans with multi-discipline systems engineer, Imanuel Portalatin. Imanuel, would you like to introduce yourself to the audience?
Imanuel: 00:34 Thanks, Cameron. I’m really, first of all, I’m really glad to be here. Yeah, as you said, I’m a multi-discipline systems engineer, so what does that mean? It really means that I have a broad base of knowledge across different disciplines, and I go deep into a few of those. Specifically, cloud and edge computing, DevOps and no SQL databases, to try to bring analytic capabilities to some very difficult problems that are not normally tackled necessarily by commercial solutions.
Cameron: 01:10 Yeah, I like to think of you as being the guy who can figure out how to make all the different little widgets and doodads that we always hear about every year work together. It’s one thing to have cameras and microphones, but being able to pull all those different technologies together to make a platform where you can share Instagram photos, it takes a lot of different knowledges and skills.
Imanuel: 01:28 That’s correct. Yeah. Yeah. Sort of take a more holistic view of technology. And also, specifically because of MITRE, our sponsors have very unique needs. We’re working in the public interest, so we need to think about things like availability, even in the case of a disaster. That’s something that is very important to many of our sponsors and also in hostile environments, which are environments where we don’t have the infrastructure necessarily that we have in CONUS [continental United States] in the contiguous United States, so things still need to work for our sponsors, even when we don’t have all these nice networks and cloud-based systems that some of the commercial technologies do have.
Cameron: 02:12 And, actually, I think this is a great spot to segue right into our first and probably the core of the discussion I want to have with you today, which is this turkeys versus swans mentality. As I’ve heard you describe it, it is this question of scarcity and resource availability and making what you have work for you. And first let me ask, what do you mean by turkeys versus swans? Because I have never heard any part of this expression before anywhere.
Imanuel: 02:37 That’s a really good question. The term comes from a Nassim Taleb, who was an options trader in Wall Street. And, basically, the options traders are the ones that are actually betting against the market. He had this mentality of kind of looking for weaknesses in the market. And he developed this whole theory around the term black swan, and just let me say that has nothing to do with the Natalie Portman movie. But, basically, the black swan comes from the fact that up until pretty much the 19th century, it was a fact that all swans are white. Nobody had ever seen a black swan. That all changed when explorers went into Australia, and they discovered that black swans are actually really common over there. The probability of a black swan existing went from 0% to a 100% basically overnight.
Imanuel: 03:33 And that’s what happens with a lot of catastrophic events or things that are really, that it could really be impactful in our lives, that we really don’t know that they could happen. And even when we know they could happen, there are a certain systematic human biases that keep us from seeing them. And this is something that I’ve also got really interested in looking at the whole black swan thing. And there’s two Nobel laureates, Daniel Kahneman and [Amos] Tversky that did a lot of work in this area. And, basically, they discovered that consistently, humans have blind spots and will act irrationally in certain contexts just because of the way that we’re wired. If you think back to our natural environment, let’s say, when a tiger is coming towards you, you don’t stop to think, oh, is this the kind of tiger that eats people or the one that doesn’t? You only do that once, after that everybody knows tigers eat people.
Imanuel: 04:29 In the modern world that we live in, there’s a lot more complexity, and sometimes it’s not as easy to know what is really the rational approach. That’s really where the turkey comes in. The turkey story is that turkeys get fed and cared for all year round. From the turkey’s perspective, farmers love turkeys. Well, that all changes on Thanksgiving when there’s a big surprise that the turkey can never see coming. If turkeys had analysts, the analyst will be telling them, “Oh yeah, this trend is going to continue. We’re going to continue to be fed and cared for and everything is going to be great.” But there’s something that is outside of the turkey’s experience that is about to change radically. And that’s really what turkeys versus swan is all about. In the context of these complex systems, not so much predict what’s going to happen, but really build ways for us to be prepared for volatility in general.
Imanuel: 05:26 What all these events have in common is that there’s a radical change. It’s not so much useful to try to predict what’s coming, as to make sure that the systems that we build can really deal with volatility in general. And that concept is something that Taleb calls anti-fragility, which is when things actually get better with volatility rather than worse. And there’s not a lot of examples of technological systems that do that. There’s a lot of biological examples, humans and organisms in general, when they are exposed to a virus, they developed some immunity. Unfortunately, right now we’re looking at a new virus where there’s no immunity and there’s no vaccine. That’s kind of the big problem with the COVID-19 crisis. But in a sense, this COVID-19 isn’t a black swan. It’s a white swan. We could see it coming, but in general, if we prepare for volatility, we will be protected not only against white swans like COVID-19, but we’d also have some level of protection against the black swans, which are the ones that we never see coming.
Cameron: 06:33 Yeah. Let me make sure I distill this down into its most pure version based on what you described. Before, we liked to think of it as black swans, white swans, where black swans were more or less random, we never really knew when or how they might ever appear. You’re saying, we’re not really dealing with swans, we’re dealing with these turkeys where, yeah, to the turkey in its individual context, it might feel completely random. But with all the data and all the knowledge that we see over time, especially when you’re operating with the amount of data and at the levels that the government and our other sponsors are, you really have the ability to see these trends coming and not just plan for them but make networks that go beyond being resilient to the point where they can actually accommodate these outliers.
Imanuel: 07:19 Yes and no. I think in a way, yes, we can sort of see the trends coming, and we need to prepare for those. But what the turkey story is illustrating is that it can be sometimes impossible to really prepare for everything that is coming. Building anti-fragility in general what it means is that we are prepared for change. Not just prepared for a specific event. And I’ll give you an example. This COVID-19 crisis also obviously has impacted the whole world in a new way. I think the tendency from now on is going to be to prepare for another virus that is precisely like COVID-19. That is airborne, that affects the respiratory system. And I have seen the other day somebody sent me something about another…some research company that was already advising their customers to prepare for the touchless economy.
Imanuel: 08:15 And, yes, that could be something that we could undertake just in general, but at the same , there’s no guarantee that the next virus is going to be exactly like COVID-19. I think being prepared for change in general, and that’s something that a lot of our sponsors really understand. Whether we can predict the black swans or not is a thing that is very difficult for us to grasp. A lot of people cite Taleb inside the black swan and then they’d go on to say that you can predict the black swan. But Taleb would say, “No, you can’t. If it’s a real black swan, you cannot predict it.” But a lot of, for example the military, is used to dealing in an environment of constant change, so they have very good systems in place for dealing with this kind of thing. I think, in general, we need to adopt more of that mentality that things just change constantly as opposed to a mentality of more just forecasting. Yeah, forecasting is useful, but we also need to prepare for volatility in general.
Cameron: 09:16 I think this has been an excellent segue because this actually leads us into the meat of what I want to discuss with you today. I love hearing things like we want more resiliency in IT and being able to do disaster recovery, but there’s so many different things that it can mean. It could be everything from table topping and creating business impact assessments, to something that actually manifests itself as a technical solution. And your work, your work is very technical. You find lots of ways to use a mix of, as I understand it, cloud computing and edge computing. Using the typical AWS, Google Cloud as this primary driver behind how data aggregates and is used. But then you also have edge computing, these specific devices that live at the edges of a network. The things that we as individuals are typically interacting with that you’ve found fairly novel ways of making work in all sorts of scenarios. And I am hoping we can transition into a discussion about what do you think about when you’re building these resilient networks? And how do you make those things play together in a useful way?
Imanuel: 10:16 Yeah, that’s correct. And you’re totally right. Disaster recovery, disaster preparedness means different things for different people. And traditionally, a big driver of moving to the cloud is that. That you have your data and your capabilities that are deployed across different zones and therefore are more resilient. And that’s great. But the problem is when you don’t have the connectivity between the clients and the capabilities that are running in the cloud. And that’s really what I’ve been trying to focus on. And when you ask about, how do we tie that into this whole black swan? There’s a very important concept that comes out of this work that is optionality. You need to have different options as to how you process data. And that’s how when I was doing research at UVA for my graduate capstone research, I looked into this whole issue of how do we improve capabilities for hostile environments, which is the environments that, as I said, sponsors, they operate in day to day.
Imanuel: 11:20 They’re operating in as first responders when there’s disasters, when there’s an emergency, they’re operating, the military’s continually operating in environments that are contested. There’s jamming, there’s different things. This is very important for our sponsors. What I try to do is look for systems that have optionality in the way that they process data, and really, edge computing, it’s a great tool that is really coming into being now because of the whole thing of the device mesh that is around the user. Nowadays, in our homes, we have tablets, we have phones, we have smart TVs, we have watches and stuff like that. Commercial providers are realizing, hey, we need to make sure that we tie all that together so that those devices interoperate and then we can create more value for our customers.
Cameron: 12:09 So if I can just expand on what you’re saying, you’re saying that there are lots of situations where we go into hostile environments, and you’d like to think you can just use the internet anywhere. But obviously if you’re in a place with bad actors, the expectation is they’re probably trying to own their network and you need to, in a weird way, bring your own entire infrastructure. And so you found ways of making it so that instead of this very centralized kind of internet we’re all used to, there are lots of ways for those end points, you could think of it almost as the routers and the different pieces that help communicate. That the towers and the stations themselves are able to do a certain amount of the processing themselves and then push that aggregate information along until it gets to whoever’s responsible for command and control.
Imanuel: 12:50 That is correct. And that’s, as I was saying where optionality comes in. And the idea would be that, eventually, any device that you have that you own and that you have secure, could be part of your network and could be used both as a data aggregation device and as a relay device to relay data from other devices that are working in that area. And this was really the subject of a big coordinated effort that my division, which is the Data and Human-Centered Solutions division within MITRE, did a few years back. We wanted to pool our resources across all the different departments that joined, that formed this division, all the different departments that formed this division, which includes social computing experts, it includes geographic experts, data engineers like myself.
Imanuel: 13:45 And we wanted to tackle a problem that was really relevant to our sponsors. And disaster recovery was really prevalent in the news, so we decided to pool our resources and come up with this concept of an end-to-end solution of how people, even on a degraded environment could still report immediate needs and they could be aggregated and presented in a way that was useful to command and control. And we wanted to use very, very basic, very cheap hardware because if you’re in a situation like this, you don’t want to have to go out and buy all the specialized stuff. You don’t want to have… you wanted to make it cheap so that even a locality could very easily deploy this kind of system.
Imanuel: 14:27 We based it on Raspberry Pi’s, which are a very simple prototype of computer that costs, $60, $70, depending on the kit that you get. And we wanted to be able to deploy both an application and also a mesh network protocol to these devices. And, basically, the mesh network is where each device knows where its nearest neighbors are and can relay information to each one of those neighbors so that if two nodes cannot communicate, then there’s different paths to where the information can flow.
Cameron: 15:02 If I could give an alternative example, I guess most of the people listening to this, you’re familiar with the fact that you’re probably listening to this podcast on a phone or a laptop and each one of those devices is able to connect to a wifi network. But what you probably don’t realize is that technically when you would make your phone a hotspot, your phone is the wifi network. You could have it where your phone, instead of sending a message directly to T-Mobile, it sends its message through a bunch of other phones until it gets to the phone that’s closest to the T-Mobile network and then can actually send it off to the actual carrier. And in this same way, it would…sounds like what you’re doing is you’re building these, and first of all, I absolutely love the DIY nature of what you’ve been describing because I think a lot of us, when we think of government or any big organization trying to do stuff, we expect them to buy multimillion dollar, several-thousand-dollars-a-piece servers that they have to set up in these large rooms and all this stuff.
Cameron: 15:58 And you’re bringing a very, like, hackathon DIY approach to it, which is we can go and get Raspberry Pi’s, which if I recall, they range in price from, I think the most expensive version I’ve seen recently is $35, and the cheapest version is maybe $5. And they have the same hardware in them that you would have in a smart phone. And they’re able to do, when I’m saying that you could have your phones passing along messages. That’s exactly what the system sounds like it’s doing. It is taking these fairly cheap computers, putting fairly sophisticated software on them and using them so that they can communicate from someone in the field to the specific person or the specific end point that lets you use your bigger infrastructure.
Imanuel: 16:42 That’s perfectly correct. And that’s a great example because essentially what our goal was, like you said with the example of the phone as acting as a hotspot, is to get these Raspberry devices to act as local hotspots that people could then go in, maybe a kind of secure code that will connect them to the hotspot and they could interact with it and instead of going to Facebook or anything like that, they go to a very targeted platform that allows them… that the main purpose of this platform is to gather information about immediate needs.
Imanuel: 17:16 And also I want it to jump in what you said about the DIY nature of this project, which yeah, I agree. It’s great. And it’s been a theme with MITRE in trying again to provide optionality to the way that our sponsors operate. They’re obviously, again, are in very challenging environments. There’s fiscal constraints. There’s all these different constraints. And one of the areas that we’ve been trying to develop this system further on is by using drones. Using unmanned area vehicles both as a relay node and also as a data gathering platform. The DIY nature just adds to the optionality, as I was saying before, that makes this more anti-fragile because you have different options in the way that you can deploy this. You have different options in the way that you put the hardware together, and it just makes it a lot more usable by  sponsors, by a broad range of sponsors.
Cameron: 18:10 Yeah. I think the important distinction here is normally when people think of MITRE’s work, they think of it being us helping government, but this is an example of us helping almost everyone because if I’m a hospital, I need to communicate this specific piece of information back to government, back to someone else who needs to be collecting this information to get a, I don’t like to use the example, but we’re right now we’re dealing with a pandemic. I’m trying to do reporting of numbers or I’m trying to understand people’s symptoms in different areas. Those people probably don’t have access to the same equipment that a larger hospital or a larger organization would have. And so being able to tell them, you can spend a small amount of money in the double digits to buy this piece of hardware, put this piece of software on it that you can get from us for free, and then use that to handle the reporting.
Cameron: 19:02 We have that kind of trust so that people believe that when we tell them this stuff, we’re not trying to do something malicious, but at the same time we’re trying to give people more opportunities. People that you don’t even think of needing that kind of resiliency and opportunity, but the people who are the most likely to make the most significant change on the ground. In this case, I’m referring specifically to maybe first responders, maybe those in, like I mentioned, more rural hospitals or healthcare providers who aren’t near large cities or major metropolitan areas.
Imanuel: 19:32 And this is like you said, this is an instance where MITRE is doing independent research, but that is on behalf of the public interest that can benefit a large number of sponsors. And again, yeah, I know we don’t want to focus only on COVID-19, we get enough of that in the news every night. But one of the conversations that I’ve been having and that we are trying to put a proposal together, I reached out to a project lead that I have worked with on the health side of MITRE, on the health FFRDC, and we were talking about this exactly that problem that you’re talking about. There is a big difference between rural healthcare and healthcare in the major urban centers, and there’s a lot of people even in rural areas, that are elderly, that are probably at greatest risk, and it’s really hard to get that information from them.
Imanuel: 20:27 Exactly what you’re talking about is one of the areas that we hope to be able to expand this platform on. MITRE, for example, has this capability called Sara, which is a way to track the symptoms and sort of assess the risk to patients and the people around them. And we want to take that same concept and refactor it in a way that it can be deployed to the edge nodes and to edge computing. And this will probably also be even on mobile devices themselves as you were describing earlier so that we can, again, pre-process, gather that information, pre-process it at the edge. And then if there’s anything, if there’s any analytics that show that maybe we need to alert local authorities and things like that, then we can share those with the local authority. Yeah, that’s a perfect example of the kind of thing.
Imanuel: 21:26 Another aspect of this is that if hurricane season is coming, and it is probably going to be pretty active again, unfortunately. Now if we are still dealing with a pandemic like this, and we’re having to enforce social distancing, and all of a sudden you have a hurricane and you lose communications. And by the way, let me just mention that part of the driver for this was my own experience. I’m from Puerto Rico, and in Puerto Rico, we had a hurricane, we had a really bad category five hurricane a few years ago. And let me tell you, if you think that cell towers and all that stuff can withstand a category five, it can’t. Basically, the transmitters are just ripped off of the tower. Many of the other utility poles that were older, they were down. That’s why the island was really out of communication, basically. It was like a blind spot in communications for weeks, pretty much. And then it took months before normal service or somewhat normal service was restored.
Cameron: 23:33 I think a lot of people in America can actually sympathize with those hurricane situations. I, as an ex-Floridian, I completely understand what it’s like to be in hurricane season, where you’re expecting that you’re going to get power back soon but for some reason you’re the last street in your entire county to get power back. I think what I want to understand right now is, let’s say I am dealing with a hurricane, I’m dealing with any crisis, and you need to deploy this network. I want to quickly just understand, what is the process like for rollout? And then, I am the end user. I want to actually report some piece of information. Maybe I’m telling you that my house has been damaged. Maybe I’m telling you that I am sick with these symptoms. Am I talking to FEMA? Am I talking to a specific agency? What happens from the moment I open my web browser, to the moment they find out that I need help?
Imanuel: 24:32 Pretty much the workflow for this capability, the prototype capability that we developed is that you would typically go and find a hotspot. You would scan a QR code in that hotspot. That will give you access to a local area network where you can actually interact with the capabilities. Then you would open your browser, and it would automatically redirect you. As I said, the idea for this is not to let people necessarily post pictures and put Facebook and stuff like that. But really to communicate immediate needs. You would be redirected to a web app, and the web app that we have created right now, right at the top you would see a map, and right away you can start interacting with the map. And for example, in your case where you said, “Oh, my house has been destroyed.” You can drop a pin, try to look the proximity of your house or your neighborhood and say, “Hey there’s flooding here. Houses have been flooded,” or things, trees have been knocked down, roads are blocked. Whatever it is, you can report it in there immediately.
Imanuel: 25:35 But also if you continue to scroll down, there’s this very comprehensive survey that you can fill out and has been put together by some of our colleagues over at social computing that have a lot of experience dealing with disasters. And it has a series of questions that allows command and control to assess needs very, very accurately. Once it realizes that there’s new data that needs to be sent up, it’s going to publish it over the mesh network. And that will find its way through to what we call a master node, which is a node that has good connectivity to the internet. And that master node has a master database that has all the pins that have been dropped, all the cyber responses. And we can use that to do analytics on the data and kind of figure out, where is the most immediate need?
Imanuel: 26:26 They could actually configure the system to do an aggregation of those vulnerable populations and sort of build a heat map. Okay, here are areas where they’re going to need more help. And also in the pandemic scenario that we’ve been talking about, we can also try to identify what are the populations that are most at risk and that they need to be given priority. That’s really kind of the end-to-end. One other thing that I should point out is that it can also aggregate data from other sources, social media, things like that. One of the characteristics of these types of scenarios is that a lot of data gets generated at the edge. If you’re doing mining on social media and things like that, that stuff is good for what was happening up until the event happened. But then what this is doing is actually enhancing that data with data that is from the edge that is more timely so that they can do a more broad-based analysis.
Imanuel: 27:23 And another thing that we could also do is that we could also have a more targeted capability just for first responders so that if first responders are out in the field, they can actually find a kiosk or an s-node, or maybe there’s going to be one that is mounted on their vehicle and is constantly looking for other nodes to be able to relay data, and they can enter on their device. And their device could have even more capabilities because we can configure that beforehand, and they can enter even more information and give you even better intelligence about what’s going on in the field.
Cameron: 28:00 Well, that is a lot to take in. And I’ve got to say that it’s nice to know that there’s this much breadth and depth and value being brought. And so I guess now’s the most important time to ask, Imanuel, where can we find out more about your team? Where can we find out about your work? How can people get in touch if they want to learn more about this kind of effort?
Imanuel: 28:19 Yeah. You can contact me directly, Imanuel Portalatin. We have some materials that have been already publicly released, so I can actually share even outside of MITRE. Like I said, I’ve given a talk on this very topic at a national conference. That’s MongoDB World, which is the database platform that we’re using for this effort. I can share some of those resources even outside of MITRE so that we can, my wish is that we can sort of establish partnerships to continue this work so they can do the most good. Feel free to contact me, and we’ll have the information in the show notes: iportalatin@mitre.org.
Imanuel: 28:57 I’ll be happy to route you to other team members that have worked on this effort because, like I said, I want to stress this was an effort by the Data and Human-Centered Solutions division that involved a bunch of different people. I was really, as we were talking in the beginning, I was more kind of the glue that was trying to bring these capabilities together, but there’s a lot of different things, different pieces in there, different work that was done that is very, very impressive. And I want to make sure that I point that out.
Cameron: 29:28 Well I must say that is absolutely amazing, and I’m sure everyone listening will be more than happy to get in touch. Before we go, I’d like to say thank you to MITRE and to the Knowledge Driven Enterprise for making this show possible and a very big thank you to you, Imanuel. I know these are very trying times for all of us, but I appreciate you being patient with me and helping me set up this interview and finding the time to share these incredible insights.
Imanuel: 29:51 Thank you for the opportunity. Like I said, I’m passionate about this, and I’m really glad for the opportunity from you and MITRE to have given me to be able to do this work and to let people know about it. Thank you very much.


Cameron Boozarjomehri is a Software Engineer and a member of MITRE’s Privacy Capability. His passion is exploring the applications and implications of emerging technologies and finding new ways to make those technologies accessible to the public.

© 2020 The MITRE Corporation. All rights reserved. Approved for public release.  Distribution unlimited. Case number 20-1141

MITRE’s mission-driven teams are dedicated to solving problems for a safer world. Through our public-private partnerships and federally funded R&D centers, we work across government and in partnership with industry to tackle challenges to the safety, stability, and well-being of our nation. Learn more about MITRE.

See also:

Justin Brunelle: Lessons from MITRE’s Innovation Program

Rachel Mayer on the Fight Against Maternal Mortality

Interview with Julie McEwen on why privacy is key

Dan Frisk and Paula Randall on bringing innovation to government

Marcie Zaharee and MITRE’s Open Innovation Challenge

Theodore Wilson: Thinking Like a Turtle

Awais Sheikh on Deciphering Business Process Innovation

Jackie Morin on Her Journey from Intern to Senior Engineer

Jay Crossler on Why Passion Is the Key to Success

Dan Ward, Debra Zides, and Lorna Tedder on Streamlining Acquisitions

Dr. Philip Barry on Blending AI and Education

Dan Ward, Rachel Gregorio, and Jessica Yu on MITRE’s Innovation Toolkit

Tammy Freeman on Redefining Innovation

Jesse Buonanno on Blockchain

Dr. Michael Balazs on Generation AI Nexus

Dr. Sanith Wijesinghe on Agile Connected Government

Is This a Wolf? Understanding Bias in Machine Learning


Pin It on Pinterest

Share This